Let's share what we know about the matchmaker

Hello,

Since there is a lot of discussion how the matchmaker should be improved with various ideas being thrown out, I think it is important that we have a basic idea of how the matchmaker works right now.

I wrote some programs that do approximate the SC matchmaker well, but those are all approximations, and I am unsure of how the actual system works.

I will write down what I think it does, and then whoever has any additional info can contribute and we can get to a decent idea about it. The idea behind is not to empower engineers or anything like that, simply to get a better understanding.

How I think the matchmaker is working now:

- You press the "search" button
- defensive and offensive weights are calculated for every member on the list
- recent win/loss points are calculated
- defensive and offensive weights of each player are being checked against each other to apply "lopsided penalty" (which is likely in the form of lower_weight = (lower_weight + higher_weight) / 2, or something similar)
- defensive and offensive weights are added up into a single weight value per player
- the weight values are sorted
- the base threshold value is added to the clan
- the clan is added to the que

- server matchmaker tick starts
- it checks all clans against all other clans in the pool
- it calculates the clan match error
- if the clan match error is below both clans current threshold values, it generates a war and removes the clans from the pool
- if no match was found for a clan, it's threshold is increased
- server waits for the next tick and then repeats the process

The match error is calculated by 3 variables:
1. absolute difference of clan totals (weighted with top player having 2x effect compared to bottom player)
2. absolute difference of per lane weights (weighted with top player having 2x effect compared to bottom player)
3. absolute difference of recent win/loss (i am assuming this has the least effect)

Those 3 variables are added up and if the result is below the threshold it is considered a match.

After a match is being found, the players are sorted by their defensive weights and then displayed on the war map in that order. Offensive weights are being used to find the recommended target for each player.

Please point out where/if I am wrong and I will update the OP.

Quote:

---

Originally Posted by dorsan

Hello,

...- defensive and offensive weights are added up into a single weight value
- the weight values are sorted...

---

I assume you are saying all the defense and offense weights of every village are adding into 1 sum? Or are you saying that some kind of list is formed out of the totals for each village (offense and defense)?

I have always thought there is a grand total that is matched with players having the freedom to disperse their clans weight how they deem the most beneficial. Then it searches for a clan with a similar total. One of the things that annoys me the most is the difference in clan structures we are matched too. Also, it seems the bigger the war, the greater the mis-match as there is more flexibility to move the weight around.

If that is what your are saying, great. you may want to make it a little more specific.

Also, somehow it seems the more wins you get you get handed more challenging matches from the MM. I don't know how it determines what is a more challenging match, other then "recent war log" performance. But as you have said, you have 4 losses on your log and have matched clans with 6 or 7 times that. My assumption on that matter is that there is some limit on the MM that allows the winning clan to be challenged by a "more average clan", but the "more average clan is allowed 110% of the winning clans weight. As the clan progresses, that number goes up until it's capped. Other wise at some point if there was no cap you would loose just by pressing the "start war" button.

I'll add that for now, but I'll keep thinking on it on my drive home from work.

why do you think the weights are added to a single value?

with the above in mind, im not sure how the above accounts for the current situation where engineers ar meeting engineers, wouldnt this indicate a double list system is already in play, just with way higher tolerances and iether no sanity check or a very generous sanity check?

Total clan weight based on geometric mean, or is that old data?

Quote:

---

Originally Posted by SuperStorm103

I assume you are saying all the defense and offense weights of every village are adding into 1 sum? Or are you saying that some kind of list is formed out of the totals for each village (offense and defense)?

I have always thought there is a grand total that is matched with players having the freedom to disperse their clans weight how they deem the most beneficial. Then it searches for a clan with a similar total. One of the things that annoys me the most is the difference in clan structures we are matched too.

If that is what your are saying, great. you may want to make it a little more specific.

Also, somehow it seems the more wins you get you get handed more challenging matches from the MM. I don't know how it determines what is a more challenging match, other then "recent war log" performance. But as you have said, you have 4 losses on your log and have matched clans with 6 or 7 times that. My assumption on that matter is that there is some limit on the MM that allows the winning clan to be challenged by a "more average clan", but the "more average clan is allowed 110% of the winning clans weight. As the clan progresses, that number grows up until it's capped.

I'll add that for now, but I keep thinking on it on my drive home from work.

---

for years people repeated, including me, clans match on total weight, based on geometric mean, however onxy pointed out there is no way to explain how close of a matchup you get in townhalls, if its simply total weight versus total weight without in lane checks. Its way to commen to have bases of very similar strength all along your rostar, for example a clan with various levels of development 4 th11, 9 th10 and 15 th9 and 2 th8 with match way more often than can be explained by coincidence to very similar clans. You very rarely see that clan match to a 25 th10 and 5 th 4. Maybe a bad example by the lane by lane comparisons seam way to similar to be coincidence, allthough i do agree the game forces you to an optimal taper for your townhall number and level, so it may simply be your clan looks similar in composition to the others as you have been trained by he matchmaker to set it up as such, and similar weight clans have been similarily trained.

great thread by the way.

100 docates.

Quote:

---

Originally Posted by SuperStorm103

I assume you are saying all the defense and offense weights of every village are adding into 1 sum? Or are you saying that some kind of list is formed out of the totals for each village (offense and defense)?

---

I changed that sentence to a more clear one. I meant the values are added up for each player.

Quote:

---

Originally Posted by SuperStorm103

I have always thought there is a grand total that is matched with players having the freedom to disperse their clans weight how they deem the most beneficial. Then it searches for a clan with a similar total. One of the things that annoys me the most is the difference in clan structures we are matched too. Also, it seems the bigger the war, the greater the mis-match as there is more flexibility to move the weight around.

---

I don't think that is how it works. I really think the 3 values are added up. But it will end with a similar result. Consider this match of 2 clans with 2 players each:

100 60
20 60

the sum of absolute differences is: 40 + 40 = 80
absolute difference of sums is: 120 - 120 = 0
if we assume that the performance is the same for both clans, we can ignore that for now

the total error value will be 80 for those 2 clans

Now let's consider a closer match:
80 81
40 80

the sum of absolute differences is: 1 + 40 = 41
absolute difference of sums is: 160 - 120 = 40

the total error value will be 81

If you look at it, in match 1, both #1 and #2 position has +-40 error, while the second match only has a +-40 error on the #2 position and +-1 error on #1 position, so logically is is a better match. But because the absolute difference of the weight totals is also added, this results in the first having a lesser total. This is why it appears that the matchmaker compares the average value first. In fact you can approximate the matchmaker by using 2 values: weighted average, weighted sum of differentials, and you will get almost exactly same matches as using this method. However, this method seems to be the one used to me.

Quote:

---

Originally Posted by SuperStorm103

Also, somehow it seems the more wins you get you get handed more challenging matches from the MM. I don't know how it determines what is a more challenging match, other then "recent war log" performance. But as you have said, you have 4 losses on your log and have matched clans with 6 or 7 times that. My assumption on that matter is that there is some limit on the MM that allows the winning clan to be challenged by a "more average clan", but the "more average clan is allowed 110% of the winning clans weight. As the clan progresses, that number goes up until it's capped. Other wise at some point if there was no cap you would loose just by pressing the "start war" button.

---

I really haven't seen enough data to conclude how the performance part works. In fact I usually ignore it completely when making approximations. You have some observations, but your data is very limited as well. It would be helpful if more people would add their observations on this.

Quote:

---

Originally Posted by Vikingchief

why do you think the weights are added to a single value?

---

Because whenever we have a huge offensive advantage, the enemies have huge defensive advantage. It has come to the point where when I see infernos on enemy #1, I know that they will have low level royals and troops. This suggest that offensive and defensive weights are being added up.

Quote:

---

Originally Posted by Vikingchief

with the above in mind, im not sure how the above accounts for the current situation where engineers ar meeting engineers, wouldnt this indicate a double list system is already in play, just with way higher tolerances and iether no sanity check or a very generous sanity check?

---

No, it wouldn't indicate a double list. It is possible they are flagging clans with lopsided bases, but what I think is more likely is that engineered clans tend to have more similar weight spread to each other than to non-engineered clans.

Quote:

---

Originally Posted by Vikingchief

Total clan weight based on geometric mean, or is that old data?

---

I believe it's a simple sum of weighted values. Like weight1 * 2 + weight2 * 1.9 + weight3 * 1.8, etc. for example.

Quote:

---

Originally Posted by dorsan

I changed that sentence to a more clear one. I meant the values are added up for each player.


Because whenever we have a huge offensive advantage, the enemies have huge defensive advantage. It has come to the point where when I see infernos on enemy #1, I know that they will have low level royals and troops. This suggest that offensive and defensive weights are being added

why would suoercell not use double list? Is it possibly they try double list forst and if they cant find a match switch to single list?

onyx seamed to think they dont currently use single list, he suggested they should, did you get the impression he believes they currently use double list, or some other method (no lists?). Maybe he thinks its total weight plus some in lane sanity checks?


No, it wouldn't indicate a double list. It is possible they are flagging clans with lopsided bases, but what I think is more likely is that engineered clans tend to have more similar weight spread to each other than to non-engineered clan

could it indicate a double list though, as a double list will match like for like? Supercell said they matching engineered ro engineered on purpose i believe, and then iether that would need a flag system OR a double list. i suppose the sometimes poor matchups indicate that double lists are not used, allthough could that not simply be that they allow high error tolerances?


I believe it's a simple sum of weighted values. Like weight1 * 2 + weight2 * 1.9 + weight3 * 1.8, etc. for example.

---

. i thought this was one of the few things supercell actually released about the matmaker, that its the square root of sum of the squareroot of base weights, to prevent rostar engineering, way back in the early days.

also, i here a lot about server side functions, that adjust the weights according to many variabler relating to the base, meaning the weight if each defence and offence structure is effected by and effects the weight of all the others.

Quote:

---

Originally Posted by Vikingchief

Its way to commen to have bases of very similar strength all along your rostar, for example a clan with various levels of development 4 th11, 9 th10 and 15 th9 and 2 th8 with match way more often than can be explained by coincidence to very similar clans.

---

So, in deference to the idea that town hall levels have any bearing on slotting within the MM my current war is such...

2 TH9, 8 TH5(defenseless) vs 1 TH10, 1 TH8, 1 TH6, 2 TH5, 1 TH4, 4 TH3.

By slotting, only 2 places have equal TH, and 4 of the slots have a 2 TH level difference.

Quote:

---

Originally Posted by SmurfKick

So, in deference to the idea that town hall levels have any bearing on slotting within the MM my current war is such...

2 TH9, 8 TH5(defenseless) vs 1 TH10, 1 TH8, 1 TH6, 2 TH5, 1 TH4, 4 TH3.

By slotting, only 2 places have equal TH, and 4 of the slots have a 2 TH level difference.

---

Sorry my example was terrible and lazy, i meant to say bases with the strength of an average th8/9 etc, i didnt mean actually the townhall level matching. It would just seam like it as if the clan was organic maxers, or balanced maxers, or permabases that are balanced, you would see similar townhall levels, but only because the most common base with the strength of a mid th9 is a th9 currently.

looking at your war it looks like some in lane checks where used, maybe on rolling average. Your top 2 bases on average even out and maybe the values were so low at the bottom 5 Spots and tolerances so high a match was possible.

Allthough i have to admit on forst glance it doesnt look like much in lane matching was going on, except maybe a sanity check in spots 1 and 2.

superstorms data can actually answer this,

who up to do the meta analysis of all his posts?


I would but its 2 45 am.

My thoughts...

Quote:

---

Originally Posted by dorsan

How I think the matchmaker is working now:

- You press the "search" button
- defensive and offensive weights are calculated for every member on the list
- recent win/loss points are calculated

---

My suggestion is that it is using win/loss streak. We've had matches against clans with wildly different win percentages, but usually when they won their previous war. But there's no way of proving it and it probably doesn't matter hugely.

Quote:

---

- defensive and offensive weights of each player are being checked against each other to apply "lopsided penalty" (which is likely in the form of lower_weight = (lower_weight + higher_weight) / 2, or something similar)

---

I'm very sceptical that there's a weight penalty applied, and certainly not at individual base level. The only form of lopsidedness that seems to give an advantage is the most extreme ones - fully defenceless. The advantage may come from a form of roster engineering (the zero weight defences giving away the 3* as easily as a 50k would, but without the weight). However if there was a substantial penalty you'd expect that to visibly hurt these bases.

SC have said they try to match engineered clans to other engineered. And I think we see this: using a load of 9.5s we get a lot of "engineered" enemies these days. But engineered for a low value of engineered: we don't match masses of defenceless, but often other clans with 8.5s or 9.5s or maybe the odd 9.11 thrown in. Maybe some defenceless TH7s at the bottom. My hunch is that "lopsidedness" is another clan metric like "recent war performance" and it tries to match similar values of lopsidedness. Again very hard to get any concrete evidence, but that's what it feels like.

Quote:

---

- defensive and offensive weights are added up into a single weight value per player

---

It does feel like this, but it's hard to be sure (and it's such a bad way to do it that I'd be hesitant assuming it)

Quote:

---

- the weight values are sorted
- the base threshold value is added to the clan

---

???

Quote:

---

....
The match error is calculated by 3 variables:
1. absolute difference of clan totals (weighted with top player having 2x effect compared to bottom player)
2. absolute difference of per lane weights (weighted with top player having 2x effect compared to bottom player)
3. absolute difference of recent win/loss (i am assuming this has the least effect)

Those 3 variables are added up and if the result is below the threshold it is considered a match.

---

it's hard to know how the different tests are combined. It could be that the sum of the errors is below threshold. It could be sum of squares of errors must be below threshold. It could be that each test must be below it's threshold. Or it could be a combination - total error below one threshold, but some/all tests have individual thresholds that could block a match.
I'm doubtful it just adds them, partly because SC have talked about a series of tick-boxes - making it sound like each one must be passed.

Quote:

---

After a match is being found, the players are sorted by their defensive weights and then displayed on the war map in that order. Offensive weights are being used to find the recommended target for each player.

---

There is almost certainly another minor step here: in-battle loot is calculated from the actual defence weight (and then win bonus is just gold in battle * 6 * perk modifier). Gold and elix are the same, but DE uses a different function.

This also has some indirect implications on the actual matching. My best estimate is that the underlying weights increase much more per TH than the gold weights. So if a max-TH11 is nominally 100, a max TH10 is only around 60 and a TH9 little more than 30, etc.

Which makes me curious about the weightings on the top lanes of the lane-by-lane check. It's possible there is no weighting based on rank and SC are relying on the exponential nature to effectively emphasise differences in higher bases. Which is similar to the issue in March 2016 when they said they were adding penalties to bases that tried to "hide" infernos/eagles by omitting lower defences. I think the weights are explainable without any penalty: rather they just greatly increased the inferno/eagle underlying weights and changed the gold-scale function to get the TH11 gold back to the same. The effect of the change was to hammer bases "hiding" infernos, but without any actual penalty. There's no way of knowing, but I'm kind of suspicious whether in the past when SC talked about increasing the emphasis on top bases they actually just increased the weights of some big ticket items (xbow/inferno/...)

Quote:

---

Originally Posted by SmurfKick

So, in deference to the idea that town hall levels have any bearing on slotting within the MM my current war is such...

2 TH9, 8 TH5(defenseless) vs 1 TH10, 1 TH8, 1 TH6, 2 TH5, 1 TH4, 4 TH3.

By slotting, only 2 places have equal TH, and 4 of the slots have a 2 TH level difference.

---

My point that Viking was alluding to, is that even in this match there is a similarity in the profiles beyond just the averages. If it wanted a clan with the same averages as yours, then probably a clan just full of TH6s or thereabouts would be most likely. But it's found a clan with 1 or 2 high bases and then basically junk. And I'd guess that's what you usually get.

Given the unusual nature of your clan it's probably struggling to get a good match, so the lane-by-lane match isn't particularly close. But it's still closer than it would be for a random clan picked with the same average or total weights.

I was sure it was confirmed by Jonas (dev) that bases are matched lane by lane, but I'm having trouble finding a source (on vacation so I'm not working hard at it either). I thought a match was found lane-by-lane, with substantial margin for error given clan/base diversity, with a second check afterwards to see if total weights aren't way out of line. Meaning (I think) its possible for a lane-by-lane search to say "good enough" for each base match, but there's another check based on total weights to make sure it hasn't gone too far astray (a lane-by-lane match might be "close enough" for a couple bases which are a th apart - if the advantage is all in the same direction the second check will start over).

Feel free to call this speculative (or outright wrong) since i'm providing no sources, but I remember a Jonas quote about a series of checkboxes. Perhaps there's an attempt at a lane-by-lane match based on individual offense/defense base weights. Then lopsidedness (however defined by the MM) is one checkbox, recent performance (I'm guessing a binary value, however the MM sees it), overall weight match But not necessarily all check boxes.

Quote:

---

Originally Posted by dorsan

I changed that sentence to a more clear one. I meant the values are added up for each player.


I don't think that is how it works. I really think the 3 values are added up. But it will end with a similar result. Consider this match of 2 clans with 2 players each:

100 60
20 60

the sum of absolute differences is: 40 + 40 = 80
absolute difference of sums is: 120 - 120 = 0
if we assume that the performance is the same for both clans, we can ignore that for now

the total error value will be 80 for those 2 clans

Now let's consider a closer match:
80 81
40 80

the sum of absolute differences is: 1 + 40 = 41
absolute difference of sums is: 160 - 120 = 40

the total error value will be 81

If you look at it, in match 1, both #1 and #2 position has +-40 error, while the second match only has a +-40 error on the #2 position and +-1 error on #1 position, so logically is is a better match. But because the absolute difference of the weight totals is also added, this results in the first having a lesser total. This is why it appears that the matchmaker compares the average value first. In fact you can approximate the matchmaker by using 2 values: weighted average, weighted sum of differentials, and you will get almost exactly same matches as using this method. However, this method seems to be the one used to me.


I really haven't seen enough data to conclude how the performance part works. In fact I usually ignore it completely when making approximations. You have some observations, but your data is very limited as well. It would be helpful if more people would add their observations on this.


Because whenever we have a huge offensive advantage, the enemies have huge defensive advantage. It has come to the point where when I see infernos on enemy #1, I know that they will have low level royals and troops. This suggest that offensive and defensive weights are being added up.


No, it wouldn't indicate a double list. It is possible they are flagging clans with lopsided bases, but what I think is more likely is that engineered clans tend to have more similar weight spread to each other than to non-engineered clans.


I believe it's a simple sum of weighted values. Like weight1 * 2 + weight2 * 1.9 + weight3 * 1.8, etc. for example.

---

Easily from our data is more like #1*10/10+#2*9/10+#3*8/10.... Till #10*1/10 in a 10v 10

Quote:

---

Originally Posted by OnyxDS

Given the unusual nature of your clan it's probably struggling to get a good match, so the lane-by-lane match isn't particularly close. But it's still closer than it would be for a random clan picked with the same average or total weights.

---

Incidently, the in-line TH levels are typically close (2 TH9, 8 TH3-5), and typically the war search takes roughly 15 minutes (this search ran 25 minutes). It turns out there's lots of players who are going about wars the same way I am. I'll be building up a few of those defenseless accounts as time allows - it'll take me months though to see if there's any real difference in war matches.

Quote:

---

Originally Posted by SmurfKick

Incidently, the in-line TH levels are typically close (2 TH9, 8 TH3-5), and typically the war search takes roughly 15 minutes (this search ran 25 minutes). It turns out there's lots of players who are going about wars the same way I am. I'll be building up a few of those defenseless accounts as time allows - it'll take me months though to see if there's any real difference in war matches.

---

Also also... this is really the weirdest matchup, truthfully. My #1 is a relatively strong TH9, 20/27 heroes, with all defense maxed but for lvl2 bows and no bomb tower... their #1 has lvl2 infernos, but lvl5 AD, and other assorted TH7/8 defenses, and barely an offense to speak of (lvl1 valks, golems, 8/8 heroes, etc).

It's entirely possible the "slot-weight" is correct, but we've gone about it entirely different ways.

Quote:

---

Originally Posted by SmurfKick

Incidently, the in-line TH levels are typically close (2 TH9, 8 TH3-5), and typically the war search takes roughly 15 minutes (this search ran 25 minutes). It turns out there's lots of players who are going about wars the same way I am. I'll be building up a few of those defenseless accounts as time allows - it'll take me months though to see if there's any real difference in war matches.

---

Agree, if match is common line up and not weird roster, then in lane match is quite accurate, but if your roster is uncommon they will lessen the in lane check and move it total weighted clan weight

Quote:

---

Originally Posted by samratulangi

Easily from our data is more like #1*10/10+#2*9/10+#3*8/10.... Till #10*1/10 in a 10v 10

---

looks a lot like you would get using geometric mean, based on likely war weights.

if base 1 weighs 120000, lets say 12 simplicity its square is 144
if base 10 weighs 40000, so 4 , its square is 16

16 times 10 is 160 so almost 144, so the tenth place has about 1/10 the effect of the 1st, not because of its place but because its low weight and use the geometric mean (sum squares of base rates then squareroot that) means that the constributions are similar to samratulangis figures.

if all the bases has the same weight however, they would all contribute the same on geometric mean.

Given the tremendous wealth of info provided by sc on the subject *cough*

It is worth taking a min to appreciate how much we've learned through reverse engineering. Brownie points for the forums!

To the point, let's not forget that our pool at search time has some unknown limits as well. Many past posts 'confirm' that clans tend to match with the same countries regularly. This very likely reduces the pool population and, if one may be liberal with some cultural traits, helps / hurts the odds of getting a fair match.

Re: the war win streak. Yeah, it's a blessing and a curse. The higher it goes the more likely you are to match with sandbaggers. We've tried curbing our numbers and including/excluding to no avail. The wws is a hulk. It just gets angrier and angrier till it stamps your green war log with a thin red line.

Presently, we're workings new element into the experiment - a limited range of ths. We'll see how it goes.

Quote:

---

Originally Posted by TankSinatra

I was sure it was confirmed by Jonas (dev) that bases are matched lane by lane, but I'm having trouble finding a source ,
........but I remember a Jonas quote about a series of checkboxes. Perhaps there's an attempt at a lane-by-lane match based on individual offense/defense base weights. Then lopsidedness (however defined by the MM) is one checkbox, recent performance (I'm guessing a binary value, however the MM sees it), overall weight match But not necessarily all check boxes.

---

Just to add my own comments :-

a) Lane by Lane

The March 3rd 2016 announcement included this statement :-

Link : https://forum.supercell.com/showthre...rs-Matchmaking

"Additionally, Clan Wars matchmaking has been fitted with new routines aimed at better balancing entire teams, rather than just individual lanes. It will place greater emphasis on matching the top of the war map, and less emphasis on matching the bottom. Total team defensive strength will also be analyzed to help filter out war matchups that are imbalanced on a map-wide scale."

The above quote suggests to me, that while MM includes plural routines (more than one) to better balance entire teams, it's still NOT excluding individual lane matching, rather adding these incremental routines to the pre existing in lane matchings.

b)Checkboxes

We know that the Youtuber Ash had a direct conversation with Jonas, and from that he published these videos. I'd suggest 5.00min onwards :-

https://www.youtube.com/watch?v=4ZiIdoFnYVk

This is the first reference I recall about 'balanced bases' being a check box, and also a reference to balanced rosters.

Other checkboxes (see 5.00min onwards again) :-

https://www.youtube.com/watch?v=33o4C6SNess

This was the video, that surprised many with the clarification of Gold storages being aprox defence weight only. Because of this, I have no reason to disbelieve some of the other objective statements made in the video, notably :-

"Think of the MM algorithm as a big list of checkmarks" , including that of recent war performance.

Any one of these checkboxes are not the "be all, and end all".

To summarise my own opinion here (based on the above, and acknowledging that many things may have changed in the last year) , I honestly believe, that regardless of individual base weights (offence and defence), there is indeed a large list of server side checkboxes, that may (or may not) effect any clans matchmaking at any point in time.

It's quite possible (as TankSinatra has suggested) that "lopsided bases", or clans with lopsided bases, may indeed be a new checkbox.

As far as I'm aware, the SC servers have not been (er... shall we use the word infiltrated?) for a very long time, and therefore there's a lot of guesswork going on in the community.

Finally,
I realise I'm probably in the minority here regarding recommended target, but in my opinion you would need a huge amount of data to formulate any guess of offensive weight.
Plus, one of the more simple server side checkboxes, may indeed be a simple , multiply that number by 0.95 this week, or 1.05 next week! Who knows!

:)

Quote:

---

Originally Posted by TankSinatra

I was sure it was confirmed by Jonas (dev) that bases are matched lane by lane, but I'm having trouble finding a source (on vacation so I'm not working hard at it either). I thought a match was found lane-by-lane, with substantial margin for error given clan/base diversity, with a second check afterwards to see if total weights aren't way out of line. Meaning (I think) its possible for a lane-by-lane search to say "good enough" for each base match, but there's another check based on total weights to make sure it hasn't gone too far astray (a lane-by-lane match might be "close enough" for a couple bases which are a th apart - if the advantage is all in the same direction the second check will start over).

Feel free to call this speculative (or outright wrong) since i'm providing no sources, but I remember a Jonas quote about a series of checkboxes. Perhaps there's an attempt at a lane-by-lane match based on individual offense/defense base weights. Then lopsidedness (however defined by the MM) is one checkbox, recent performance (I'm guessing a binary value, however the MM sees it), overall weight match But not necessarily all check boxes.

---

They have "said" a lot of things. But until there is some way to demonstrate the functionality, players ( like myself ) will continue to second guess it. concept of lane matching is great, but we have matches where it would appear this functionality is totally dysfunctional.

We know 2 things for sure.. 1... the MM is capable of nearly perfect matching. 2... the MM is capable of pretty bad matching.

Questions are ; what is the general population it is selecting from? They could give insight to this without giving too much information about the match maker. in fact, if everyone knew a certain roster was more popular, there is a good chance even matching would naturally occur as more players would migrate towards that. ( similar to FWA )

How much variance , or how bad can it actually be? What I assume is, when we have a bad match, it is probably on the outer edge of what is tolerated. This is another variable they could communicate about without telling us how it works... and tightening the parameters would lengthen searches, but this is where a notice about your roster not being conducive to a match would come in handy.

So I took our current war, and I removed most of the information to illustrate the point. ( since extra information seems to confuse those who never got past 3rd grade math )

As you can see, we are 1.68 difference on a per slot basis. Pretty fair right? Defensively spot on

Now be a champ and fill in the TH and hero levels, then explain to me how offensive weighting comes into play.

The right side has 7 TH11s, 7TH10s, 1 TH8 ( rest 9s )
Left side is comprised of balanced bases in which the weight also dictates the offensive capability. ( except for 2 9.5s being the top 2 70 weighted bases )

TH	Weight	AQ	BK	GW		TH	Weight	AQ	BK	GW
	109				0		109
	108				-1		109
	107				-1		108
	89				-1		90
	89				-1		90
	88				-1		89
	88				0		88
	84				12		72
	70				-1		71
	70				-1		71
	70				0		70
	70				1		69
	70				1		69
	70				1		69
	70				1		69
	70				1		69
	70				3		67
	69				3		66
	68				2		66
	68				2		66
	68				2		66
	68				2		66
	68				3		65
	67				8		59
	61				7		54
					42	1.68

Quote:

---

Originally Posted by ryryshouse

Left side is comprised of balanced bases in which the weight also dictates the offensive capability. ( except for 2 9.5s being the top 2 70 weighted bases )

---

Balanced? This depends on how SC and more importantly, the MM define balanced. I could be wrong, just going off the limited info you posted- the first thing I notice about your war roster is that is looks perma base heavy. Aside from those two noted .5s, you have 7 max defensive TH9s, and 6 more that are fairly close. Your TH10s- 4 of the 5 are nearly maxed. Your TH11's are again, pretty stout. Maxed or right off.

The reason I mention all this- is because I've been seeing some pretty messed up pairings when there are a bunch of maxed or perma bases involved on one side of the map. Maybe it's because the MM doesn't have a fairly large pool of matches with a bunch of maxed TH10s and 9s sitting out there. Maybe there is some variable that I can't see. I can't say its a coincidence though.

Here's my theory on how it works...

For each base in the clan the MM takes the higher of the two values -- MAX(Defense Weight, (offense weight * ~85%))

It then weighs the top bases heavier than the bottom (similar to how they do it for clan trophies rankings).

Finally it looks at lane/lane to make sure it's not too far out of tilt.

Additionally I believe the win streak is much more significant than most seem to think -- if you have a high win streak you should expect a tough war.

Quote:

---

Originally Posted by ryryshouse

So I took our current war, and I removed most of the information to illustrate the point. ( since extra information seems to confuse those who never got past 3rd grade math )

As you can see, we are 1.68 difference on a per slot basis.

---

It's worth taking into account Ryryshouse, that not everybody in the English speaking world, that frequents the forum, necessarily understands every colloquialism used in the US.
Perhaps some of the confusion may lie in translation.

For example, I know math is maths or mathematics
3rd Grade? Not actually sure how old that is. Obviously I've come across the term watching US TV/films. Without looking it up, I would guess this would equate to the third year of Primary/Infant School here in England, as this is an important year! So, 7 yrs old it is.

Edit : Googled 3rd grade, answer was 8/9... hmm not far off!

'Per slot basis' - I've never heard/read/used this term before in my life. Thankfully, having read your numbers, I can indeed figure out your meaning as being, on average 1.68 deficit per base.

My point being, sometimes in life the confusion is not always the fault of the listener or reader! Thankfully, we are all different, and we should embrace such!:)

Quote:

---

Originally Posted by cnaf

Balanced? This depends on how SC and more importantly, the MM define balanced. I could be wrong, just going off the limited info you posted- the first thing I notice about your war roster is that is looks perma base heavy. Aside from those two noted .5s, you have 7 max defensive TH9s, and 6 more that are fairly close. Your TH10s- 4 of the 5 are nearly maxed. Your TH11's are again, pretty stout. Maxed or right off.

The reason I mention all this- is because I've been seeing some pretty messed up pairings when there are a bunch of maxed or perma bases involved on one side of the map. Maybe it's because the MM doesn't have a fairly large pool of matches with a bunch of maxed TH10s and 9s sitting out there. Maybe there is some variable that I can't see. I can't say its a coincidence though.

---

Well, again, I would assert. A 100% maxed base is the most balanced base possible. Other then a base just landing on the new TH from the previous. Our 3 TH11s are less weight then the other teams... but again, that isnt the issue. See if you can put your finger on why people complain.

EDIT: I would say that right now, we see this type of match maybe 1/3 or 1/4th of the time. I contribute it directly to what is in the pool to match.. and defensive based roster match.. as it was prior to the March update.

No way lane matching has 3th11 against 7 th11 , you see what i am saying here.

Quote:

---

Originally Posted by Skrags

It's worth taking into account Ryryshouse, that not everybody in the English speaking world, that frequents the forum, necessarily understands every colloquialism used in the US.
Perhaps some of the confusion may lie in translation.

For example, I know math is maths or mathematics
3rd Grade? Not actually sure how old that is. Obviously I've come across the term watching US TV/films. Without looking it up, I would guess this would equate to the third year of Primary/Infant School here in England, as this is an important year! So, 7 yrs old it is.

Edit : Googled 3rd grade, answer was 8/9... hmm not far off!

'Per slot basis' - I've never heard/read/used this term before in my life. Thankfully, having read your numbers, I can indeed figure out your meaning as being, on average 1.68 deficit per base.

My point being, sometimes in life the confusion is not always the fault of the listener or reader! Thankfully, we are all different, and we should embrace such!:)

---

the point, you have missed it again.

Quote:

---

Originally Posted by Azrial

Here's my theory on how it works...

For each base in the clan the MM takes the higher of the two values -- MAX(Defense Weight, (offense weight * ~85%))

It then weighs the top bases heavier than the bottom (similar to how they do it for clan trophies rankings).

Finally it looks at lane/lane to make sure it's not too far out of tilt.

Additionally I believe the win streak is much more significant than most seem to think -- if you have a high win streak you should expect a tough war.

---

my theory is totally different. generate some data on your theory ? the ever ellusive offensive weight .. its like dark matter. we cant see it, but there is some possibly unexplained behavior that certain people try to attribute to it. But I have no evidence of unexplained behavior, if you know what i mean.

I am going to try to compile all the input and update the OP with the new information.

The new info I see so far are:
- checkboxes
- different possible ways the matchmaker handles the recent performance

If I missed out something, please let me know. :)

Quote:

---

Originally Posted by dorsan

After a match is being found, the players are sorted by their defensive weights and then displayed on the war map in that order. Offensive weights are being used to find the recommended target for each player.

Please point out where/if I am wrong and I will update the OP.

---

If this was true why would we a max 11 with no Eagle be placed underneath 3 mid range th 10s and a much lower 11?
Also, why would an account like that with max troops and defenses get matched with a clan who's best base is a th11 with th10 level defenses and troops? I've seen bad matchups before, but this was the worst individual one I have seen

Quote:

---

Originally Posted by jrich5768

If this was true why would we a max 11 with no Eagle be placed underneath 3 mid range th 10s and a much lower 11?
Also, why would an account like that with max troops and defenses get matched with a clan who's best base is a th11 with th10 level defenses and troops? I've seen bad matchups before, but this was the worst individual one I have seen

---

It would be placed underneath those because it's defensive weight is smaller. SC doesn't really calculate weights well.

It gets matched against a clan with much lower offense/defense because SC doesn't match clans properly. These mismatches happen.

Quote:

---

Originally Posted by dorsan

I am going to try to compile all the input and update the OP with the new information.

The new info I see so far are:
- checkboxes
- different possible ways the matchmaker handles the recent performance

If I missed out something, please let me know. :)

---

I dont think you need anything but def weight.

Quote:

---

Originally Posted by ryryshouse

my theory is totally different. generate some data on your theory ? the ever ellusive offensive weight .. its like dark matter. we cant see it, but there is some possibly unexplained behavior that certain people try to attribute to it. But I have no evidence of unexplained behavior, if you know what i mean.

---

I agree the offensive weight (dark matter lol) is very hard to get a handle on. The only in game feature that references these are recommended target and heroic attack.

My theory comes from the fact that .5 (and lopsided) has been nerfed...

https://www.youtube.com/watch?v=-Jw7xNgYQGc

At 3:05 in the video it describes an adjusted weight (i.e., penalty) that a .5 village would have. This adjustment has to be based off of some type of formula.

At 11:00 in the video she describes the following...

A base with an offensive weight of 110 and a defensive weight of 90. The defensive weight ends up getting bumped up to 100.

This basically means that this lopsided base IS being taxed with defenses it doesn't posses. This suggests the following formula which is different from my last post (a quick guess) and using 90% instead of 85%...

OffensiveWeight + Max(DefensiveWeight, OffensiveWeight * .9) = TotalBaseWeight


Showing my work using the numbers in the video you would get...

110 + Max(90, 110 * .9)
110 + Max(90,99)
110 + 99 = 209 Total Base Weight

In the video she said 210 so 90% is pretty close

I really wonder why this video doesn't get more attention on the forums -- it's basically a recipe for how to optimally upgrade for war post October update (keep your offense about 10% ahead of your defense).

@ ryry

I happen to have a large data set that does indeed suggest that offense weight is taken into account.

Also we have this little gem from wysywig from a different thread:

Quote:

---

Originally Posted by wysiwyg

so rather than speculate with our best guesses, i tried to get some info from the horse's mouth

i messaged SC about exactly this - here's how it all went down *fade into flashback*

me: Is there any way to avoid matching against lopsided clans?


answer(?): Hello Chief,

Thanks for reaching out to us. i'm always one step ahead to assis the dedicated player like you and our team never want to see our clasher unhappy.

It's really sad to hear that your clan is matched up with some higher level players. I know the hard work you put in to compete in the war. I would like to take this as an opportunity to explain how these unfair matches happen.
Every time our matchmaking system tries to match a fair battle. If many players are searching for war at the same time it's pretty difficult to find an exact match. In this case, our matchmaking system tries to match up with the nearest level opponent based on the attacking and defensive strengths.
I know how difficult to defend against the stronger players. Don't worry, Our team is continuously working to make changes in the match making system. The only thing I request from your side is to bear with us for little bit time.
Meanwhile, try to work on some strategies to defeat the higher level players. If you succeed against them you can share those to your clan mates so they can learn from you.
Your kind understands in this matter is highly appreciated. If you have anything else that you would want to share it with us, please don't hesitate to drop us a line.
We hope that we may continue to rely on your loyalty
Thank you and have a great Clash ahead!
Best Regards,

*fade back into reality*

first impression - i think this is a mostly human reply because of the numerous grammar mistakes (I copied it character for character) and the logic fail (if many players are searching for war at the same time it's more difficult?)

however a lot of it feels copy-and-pastey - the empty compliments, the play for sympathy, the closing - it's all boiler plate material.
anyhoo - i replied politely saying 'thanks for the reply, but you didn't answer my question. how can we avoid matching against these lopsided clans?'

what followed was 2 more messages from the SC rep with a lot of words, but no answer. in his final message he used the phrase "In finally, ..." which apparently meant 'i'm closing this ticket, adios'.

long story short - if there is a way to avoid sandbagging clans - we're not going to hear about it from the in-game support. if anything, it'll be buried in our collective speculation.

personally, i'm going to continue my voodoo rituals in the hopes that SC will give us fair play in CW. i'll let you know if it works outt

---

Time to post some counter evidence that offense weight is not considered if you really think that.

Quote:

---

Originally Posted by Azrial

At 3:05 in the video it describes an adjusted weight (i.e., penalty) that a .5 village would have. This adjustment has to be based off of some type of formula.

At 11:00 in the video she describes the following...

A base with an offensive weight of 110 and a defensive weight of 90. The defensive weight ends up getting bumped up to 100.

This basically means that this lopsided base IS being taxed with defenses it doesn't posses. This suggests the following formula which is different from my last post (a quick guess) and using 90% instead of 85%...

OffensiveWeight + Max(DefensiveWeight, OffensiveWeight * .9) = TotalBaseWeight

---

If you go back to my original post, you'll see that I wrote the adjustment as:

If defensive weight is smaller than offensive weight:
adjusted_defensive_weight = (offensive_weight + defensive_weight) / 2
adjusted_offensive_weight = offensive_weight

So, for this scenario and your numbers, we get
adjusted_defensive_weight = (110 + 90) / 2 = 100

The equation seems to match the data.

Quote:

---

Originally Posted by dorsan

If you go back to my original post, you'll see that I wrote the adjustment as:

If defensive weight is smaller than offensive weight:
adjusted_defensive_weight = (offensive_weight + defensive_weight) / 2
adjusted_offensive_weight = offensive_weight

So, for this scenario and your numbers, we get
adjusted_defensive_weight = (110 + 90) / 2 = 100

The equation seems to match the data.

---

Nice work! Yes your formula seems to fit it even better.

I do recall that later in the video she suggested that it was OK to be a little ahead on offense, that's what made me think it was a % thing. However going by your formula (which matched her numbers perfectly) it would seem that any offense>defense is punished.

Edit:
I wonder if the reverse is also true which would penalize defensively rushed bases? They've always been bad for war and maybe that helps explain it?

Quote:

---

Originally Posted by Azrial

Nice work! Yes your formula seems to fit it even better.

I do recall that later in the video she suggested that it was OK to be a little ahead on offense, that's what made me think it was a % thing. However going by your formula (which matched her numbers perfectly) it would seem that any offense>defense is punished.

Edit:
I wonder if the reverse is also true which would penalize defensively rushed bases? They've always been bad for war and maybe that helps explain it?

---

Yeah, but the punishment is much larger if the difference is huge.

Imagine a heavily lopsided with: o = 100, d = 20
In this case defense is bumped to 60, which is a 300% increase.

Then a slightly lopsided with: o = 102, d = 100
In this case defense is bumped to 101, which is only 1% increase

Quote:

---

Originally Posted by Azrial

I agree the offensive weight (dark matter lol) is very hard to get a handle on. The only in game feature that references these are recommended target and heroic attack.

My theory comes from the fact that .5 (and lopsided) has been nerfed...

https://www.youtube.com/watch?v=-Jw7xNgYQGc

At 3:05 in the video it describes an adjusted weight (i.e., penalty) that a .5 village would have. This adjustment has to be based off of some type of formula.

At 11:00 in the video she describes the following...

A base with an offensive weight of 110 and a defensive weight of 90. The defensive weight ends up getting bumped up to 100.

This basically means that this lopsided base IS being taxed with defenses it doesn't posses. This suggests the following formula which is different from my last post (a quick guess) and using 90% instead of 85%...

OffensiveWeight + Max(DefensiveWeight, OffensiveWeight * .9) = TotalBaseWeight


Showing my work using the numbers in the video you would get...

110 + Max(90, 110 * .9)
110 + Max(90,99)
110 + 99 = 209 Total Base Weight

In the video she said 210 so 90% is pretty close

I really wonder why this video doesn't get more attention on the forums -- it's basically a recipe for how to optimally upgrade for war post October update (keep your offense about 10% ahead of your defense).

---

Big load of nope. Regardless which team has the offensive advantage in our wars, defensive weights trend identically. So either SC implemented something using a logarithmic scale and the dude botched it, or its not using offense ( in a noticeable manner )

Quote:

---

Originally Posted by cnaf

Balanced? This depends on how SC and more importantly, the MM define balanced. I could be wrong, just going off the limited info you posted- the first thing I notice about your war roster is that is looks perma base heavy. Aside from those two noted .5s, you have 7 max defensive TH9s, and 6 more that are fairly close. Your TH10s- 4 of the 5 are nearly maxed. Your TH11's are again, pretty stout. Maxed or right off.

The reason I mention all this- is because I've been seeing some pretty messed up pairings when there are a bunch of maxed or perma bases involved on one side of the map. Maybe it's because the MM doesn't have a fairly large pool of matches with a bunch of maxed TH10s and 9s sitting out there. Maybe there is some variable that I can't see. I can't say its a coincidence though.

---

I put the line about the relative scarcity of fully (offense & defense) maxed TH9+ bases out there in bold, because it seems obvious to me that that's exactly what leads to maxers frequently being matched to a higher TH.

Evidence? Count the number of those bases you find while farming or looking at random bases on global vs the number of bases that rushed in some form or another.

Quote:

---

Originally Posted by SuperStorm103

Time to post some counter evidence that offense weight is not considered if you really think that.

---

I have repeatedly. Read. What I actually suspect is that SC's offensive method is just terrible. So it is there, but it is so stupid and wrong that its contribution is nothing or works opposite the intended way.

The situation is ;
defensive contribution - observable and fair
offensive contribution - not observable and broken. Likely less then 50% of the factors. As we are able to achieve wars with even defense but often 25-40% of the roster being of a lesser TH offense.


What this means is that; balancing your base and roster is the wrong advice. The advice should be offensive slanting your bases while having anchors bases up top.

Quote:

---

Originally Posted by ChiefTuk

Evidence? Count the number of those bases you find while farming or looking at random bases on global vs the number of bases that rushed in some form or another.

---

Anecdotally - I farm in crystal and low masters on 2 TH10s. I see more 9.5s now than defensively rushed trash. Wasnt that way until maybe Oct or Nov.

I was in champs for a bit.. Even more evident there. Saw practically no defensively rushed bases, all were either maxed 10s, mostly maxed 11s, or 9.5/10.5

To me, this is the most striking thing.. I truly think that there is a glut of 9.5s out there. good bad and otherwise.

Quote:

---

Originally Posted by ryryshouse

Anecdotally - I farm in crystal and low masters on 2 TH10s. I see more 9.5s now than defensively rushed trash. Wasnt that way until maybe Oct or Nov.

I was in champs for a bit.. Even more evident there. Saw practically no defensively rushed bases, all were either maxed 10s, mostly maxed 11s, or 9.5/10.5

To me, this is the most striking thing.. I truly think that there is a glut of 9.5s out there. good bad and otherwise.

---

And , we see more 9.5s in farm wars then maxed 9s. Career 9.5s as they typically have 100+ magmas and considerable amount of lab done.

This contributes to my theory offensive isnt a factor.

Quote:

---

Originally Posted by ryryshouse

Big load of nope. Regardless which team has the offensive advantage in our wars, defensive weights trend identically. So either SC implemented something using a logarithmic scale and the dude botched it, or its not using offense ( in a noticeable manner )

---

The defensive weight is charged to the base as far as matching is concerned. In other words we can't see this information in the war map by using gold storages but the MM took it into account.

Quote:

---

Originally Posted by Azrial

The defensive weight is charged to the base as far as matching is concerned. In other words we can't see this information in the war map by using gold storages but the MM took it into account.

---

You sure? otherwise there is no data set available in which to base a formula

Quote:

---

Originally Posted by ryryshouse

Well, again, I would assert. A 100% maxed base is the most balanced base possible.

---

That's obviously faulty logic. Theoretically, a base could be kept in balance throughout the upgrade path. Since we have incomplete information on how defense weight is calculated & almost none on how offense is calculated, it's impossible to know exactly what upgrade progression would keep you balanced, but most of us have some sense of when something is off.

Your dedication to the notion that offense doesn't count under the current MM is badly misguided. What you're taking exception with, is how MM is evaluating offense in terms of putting your maxed bases against rushed bases with higher THs. There aren't enough fully maxed lineups out there to guarantee you'll match an equal every time you search, so either you benefit by facing equal TH with lower offense or you face rushed THs. I think you have a good argument for believing the current MM is too relaxed about how much of that you'll face in any given war.

Quote:

---

Originally Posted by ryryshouse

To me, this is the most striking thing.. I truly think that there is a glut of 9.5s out there. good bad and otherwise.

---

Maybe you're starting to see the light! If you put yourself in the pool of random matchups, your lineup is almost guaranteed to put someone at a disadvantage. I see no reason why it should never be you. If I remember correctly, you win at a very high rate, which tends to undercut your argument about a totally broken MM. TBH, if warring with perma-maxed bases against nearly identical enemy clans is what you like best in the game, Friendly Wars is the way to go.

Quote:

---

Originally Posted by dorsan

Yeah, but the punishment is much larger if the difference is huge.

Imagine a heavily lopsided with: o = 100, d = 20
In this case defense is bumped to 60, which is a 300% increase.

Then a slightly lopsided with: o = 102, d = 100
In this case defense is bumped to 101, which is only 1% increase

---

Agree. The more engineered you are the more punished you are. However this actually leads me to believe that extremely lopsided bases can indeed hide their strength. Right now that lopsided base in your example weighs 160 which is a bargain for the offense it brings in.

I think SC is doing this like in your formula. However a percentage like .85 would hurt severely lopsided more while allowing for a little flexibility as well.

Quote:

---

Originally Posted by ryryshouse

You sure? otherwise there is no data set available in which to base a formula

---

I'm reasonably sure. The MM process is done server side. The client side has no need for this information.

Quote:

---

Originally Posted by Azrial

I agree the offensive weight (dark matter lol) is very hard to get a handle on. The only in game feature that references these are recommended target and heroic attack.

My theory comes from the fact that .5 (and lopsided) has been nerfed...

---

Oh dear :( I used to like Heather's videos. My suggestion would be to disregard everything in there. Apart from the bit about clan level and player trophies not counting I'm not sure there is a single thing in that video that I could agree with. The fact she bases some of her analysis on spAnser's data isn't the worst of it!

Quote:

---

Originally Posted by ChiefTuk

That's obviously faulty logic.

---

wrong.

Quote:

---

Originally Posted by ChiefTuk

Theoretically, a base could be kept in balance throughout the upgrade path. .

---

as you pointed out in the next sentence.

Quote:

---

Originally Posted by OnyxDS

Oh dear :( I used to like Heather's videos. My suggestion would be to disregard everything in there. Apart from the bit about clan level and player trophies not counting I'm not sure there is a single thing in that video that I could agree with. The fact she bases some of her analysis on spAnser's data isn't the worst of it!

---

yea me either/. Her videos are good. math not so much

Quote:

---

Originally Posted by ryryshouse

Big load of nope. Regardless which team has the offensive advantage in our wars, defensive weights trend identically. So either SC implemented something using a logarithmic scale and the dude botched it, or its not using offense ( in a noticeable manner )

---

A couple of wars back we were 53k down on the top 6 bases, and then the next ones were all regular TH9s both sides and matching until very near the bottom. Guess which clan had the better heroes?

It's a pity I didn't record all the details of the one we had about 3 before that. 2 TH10s, and IIRC 5 TH9.5s our side. They had all 7 as full TH10s, and the top 2 with eagles - we were down a full 15k on EACH base. Then their TH9s were heavier, and even matching our TH8s. So either that match was done ignoring the defence completely, or maybe the fact they had single digit heroes had something to do with it.

Quote:

---

Originally Posted by OnyxDS

A couple of wars back we were 53k down on the top 6 bases, and then the next ones were all regular TH9s both sides and matching until very near the bottom. Guess which clan had the better heroes?

It's a pity I didn't record all the details of the one we had about 3 before that. 2 TH10s, and IIRC 5 TH9.5s our side. They had all 7 as full TH10s, and the top 2 with eagles - we were down a full 15k on EACH base. Then their TH9s were heavier, and even matching our TH8s. So either that match was done ignoring the defence completely, or maybe the fact they had single digit heroes had something to do with it.

---

We have quite a bit of data in this thread:
https://forum.supercell.com/showthre...tch-Comparison

If you go through it, you will notice that whichever clan has the defensive disadvantage by rule has an advantage at offense. So clearly either the offensive and defensive weights are added up or there is a penalty for having higher offenses that is added to the defensive weight during a search but then ignored when calculating the loot in the storages.

Quote:

---

Originally Posted by OnyxDS

A couple of wars back we were 53k down on the top 6 bases, and then the next ones were all regular TH9s both sides and matching until very near the bottom. Guess which clan had the better heroes?

It's a pity I didn't record all the details of the one we had about 3 before that. 2 TH10s, and IIRC 5 TH9.5s our side. They had all 7 as full TH10s, and the top 2 with eagles - we were down a full 15k on EACH base. Then their TH9s were heavier, and even matching our TH8s. So either that match was done ignoring the defence completely, or maybe the fact they had single digit heroes had something to do with it.

---

But you dont have the weights to show :(

I have seen wild defensive mismatches also, but there is usually 1-2 bases that manage to bring the whole thing back in line, on the bottom. this is the unfortunate part.

Heroes are tricky, as some scenarios they arent real relevant, in others they are super important. No way for the MM to account for situational play.

Quote:

---

Originally Posted by dorsan

Yeah, but the punishment is much larger if the difference is huge.

Imagine a heavily lopsided with: o = 100, d = 20
In this case defense is bumped to 60, which is a 300% increase.

Then a slightly lopsided with: o = 102, d = 100
In this case defense is bumped to 101, which is only 1% increase

---

The problem is that subjectively people seem to feel that x.5s got hurt the most and very defenceless weren't affected so much. Which this weight penalty just doesn't seem to explain. Heather somehow tries to contort the numbers to try to reach an answer where the (large) penalty applied to a fully engineered account doesn't matter, but the (small) penalty applied to x.5s should expect multiple extra eagles the other side. It just doesn't fly.

To come to the conclusion that there is such a weight penalty you'd have to start from the observation that x.5s have been only slightly affected, and defenceless TH11s have been punished with extreme prejudice and will automatically lose every war. Is that anyone's observation? We haven't matched a lot of truly defenceless recently, but I wouldn't say it's what I'm seeing.

Quote:

---

Originally Posted by dorsan

We have quite a bit of data in this thread:
https://forum.supercell.com/showthre...tch-Comparison

If you go through it, you will notice that whichever clan has the defensive disadvantage by rule has an advantage at offense. So clearly either the offensive and defensive weights are added up or there is a penalty for having higher offenses that is added to the defensive weight during a search but then ignored when calculating the loot in the storages.

---

You "offensive" advantage is your opinion. No factual basis for this at all.

Quote:

---

Originally Posted by OnyxDS

The problem is that subjectively people seem to feel that x.5s got hurt the most and very defenceless weren't affected so much. Which this weight penalty just doesn't seem to explain. Heather somehow tries to contort the numbers to try to reach an answer where the (large) penalty applied to a fully engineered account doesn't matter, but the (small) penalty applied to x.5s should expect multiple extra eagles the other side. It just doesn't fly.

To come to the conclusion that there is such a weight penalty you'd have to start from the observation that x.5s have been only slightly affected, and defenceless TH11s have been punished with extreme prejudice and will automatically lose every war. Is that anyone's observation? We haven't matched a lot of truly defenceless recently, but I wouldn't say it's what I'm seeing.

---

I have had 2 bases that were close to defenseless. I built defenses because they were quite bad for matching. I think people feel that defenseless have an advantage because after the bases are sorted by their real defensive weight those bases show up on the bottom of the map and the advantage the enemy gets for that defenseless base will be on the center of the map, where people will simply ignore the fact that they have 1 extra th9 compared to the other team and instead focus on the fact that there is a defenseless th11 on the bottom facing a th5.

Quote:

---

Originally Posted by ryryshouse

You "offensive" advantage is your opinion. No factual basis for this at all.

---

What do you mean no factual basis? He's recorded every single troop level and every single hero level. What more facts do you want?

Quote:

---

Originally Posted by ryryshouse

You "offensive" advantage is your opinion. No factual basis for this at all.

---

Maybe I am missing something here, we, like many other clans, ran .5 till the matchmaker changed. We got favourable matchups till the change, usually stronger on offence, similar defence. After the change, we iether, faced bases with similar levels of offence, so .5ers, or we faced bases with higher defence levels. The advantage was gone. This certainly seams to indicate our offence was counting much more, and balanced with iether higher offence or defence.

assuming what I and almost everyone running similar clans experienced did happen, the only other explanation, is that the matmaker applied a penalty weight to clans or bases with a unbalanced offence to defence ratio. However this is in effect having an offensive weight, in effect, as in balanced accounts will balance each other out, and offence strong accounts will get a penalty in line with how developed their offence is, which is in effect an offensive weight.

then if the matchmaker calculated weights to apply a penalty, it means it calculated offensive weights anyway, then why would it not use them?

So for me offence has an effect, wether its weight, a penalty, or a penalty weight.

i think the confusion comes in from it being very tricky to try guess how its calculated, with manicured offence possibly weighing very little, despite having the impact in war as an account with troops maxed across the board.

would you agree at the very least their is a penalty weight for having offence ahead of defence?

Quote:

---

Originally Posted by OnyxDS

What do you mean no factual basis? He's recorded every single troop level and every single hero level. What more facts do you want?

---

I dont really understand where he is coming from, iether we are all idiots, and he has access to some understanding we dont have, or he is aubitrary and agressivly (not toxic, just aggressive) defending something he is not at all sure off.

i had this exact same argument on global the other day, with 2 people who seams similarily certain this was the case (no offence weight) and were relatively aggressive at pushing this point.

I believe the evidence is their that their is an offensive weight, and that its close to 50% of total if balanced upgrading of troops is assumed, by I am not so sure that I would agressivly defend the position. Allthough im 90% sure its true.

Quote:

---

Originally Posted by dorsan

I have had 2 bases that were close to defenseless. I built defenses because they were quite bad for matching. I think people feel that defenseless have an advantage because after the bases are sorted by their real defensive weight those bases show up on the bottom of the map and the advantage the enemy gets for that defenseless base will be on the center of the map, where people will simply ignore the fact that they have 1 extra th9 compared to the other team and instead focus on the fact that there is a defenseless th11 on the bottom facing a th5.

---

As I say we haven't matched strong defenceless very recently. Our last loss wasn't exactly a defenceless clan. But the previous 2 were - brutal matches that were basically unwinnable where they had a big defensive advantage up top (say 6 strong inferno bases to our 2), PLUS multiple strong TH11 offence bases. They could dip every attack from top to bottom and we needed 10v10 triples - plenty of them - to compete.

We've had opponents with plenty of x.5s, a few with a load of 9.11s high up, balanced clans, even 1 or 2 defensively rushed clans and none of those have stood out as hideous matches. It's just the ones against clans with multiple defenceless accounts that seem to hurt us.

Quote:

---

Originally Posted by ryryshouse

wrong.
https://forum-content.supercell.com/...quote_icon.png Originally Posted by ChiefTuk https://forum-content.supercell.com/...post-right.png
Theoretically, a base could be kept in balance throughout the upgrade path. .
as you pointed out in the next sentence.

---

Lol, here's what you said: "A 100% maxed base is the most balanced base possible."

If a base is balanced evenly, then it's as balanced as your maxed base regardless of where it is in the upgrading process. I've been fairly indulgent of your mindset that maxing deserves to be rewarded, but you repeatedly fail to support your case with any objectivity.

Quote:

---

Originally Posted by OnyxDS

Oh dear :( I used to like Heather's videos. My suggestion would be to disregard everything in there. Apart from the bit about clan level and player trophies not counting I'm not sure there is a single thing in that video that I could agree with. The fact she bases some of her analysis on spAnser's data isn't the worst of it!

---

Other than noting that the metadata has been pulled from the game, but it doesn't exactly line up with the numbers we see in defense weight as measured by gold storages, it's not a factor in her analysis. She did actually get to speak with devs, which is more than I can say. Most of what she presented was pretty vague, I'm assuming there was a NDA signed... the upshot was mostly explaining why .5s were suddenly a liability in MM.

*Edit- I think she would have said some things differently after getting some feedback, but it's still useful for clans that have never given any thought to how MM works.

Quote:

---

Originally Posted by OnyxDS

What do you mean no factual basis? He's recorded every single troop level and every single hero level. What more facts do you want?

---

Oh, so you have a set of equations in which you solve for all 50 variables? no? okay then. Glad we got past that quickly.

The only 3 things you know are ; 1. your defensive weight ( and you dont know if this is what is indeed used in the algorithm ) and 2. your placement on the map and TH lvl. finally, you have this information of your opponents.

So until SC flops out the algorithm you have a pile of speculation.

Quote:

---

Originally Posted by OnyxDS

As I say we haven't matched strong defenceless very recently. Our last loss wasn't exactly a defenceless clan. But the previous 2 were - brutal matches that were basically unwinnable where they had a big defensive advantage up top (say 6 strong inferno bases to our 2), PLUS multiple strong TH11 offence bases. They could dip every attack from top to bottom and we needed 10v10 triples - plenty of them - to compete.

We've had opponents with plenty of x.5s, a few with a load of 9.11s high up, balanced clans, even 1 or 2 defensively rushed clans and none of those have stood out as hideous matches. It's just the ones against clans with multiple defenceless accounts that seem to hurt us.

---

We had some defenseless few wars back. Guess what was under 2 ?

Quote:

---

Originally Posted by ChiefTuk

Other than noting that the metadata has been pulled from the game, but it doesn't exactly line up with the numbers we see in defense weight as measured by gold storages, it's not a factor in her analysis. She did actually get to speak with devs, which is more than I can say. Most of what she presented was pretty vague, I'm assuming there was a NDA signed... the upshot was mostly explaining why .5s were suddenly a liability in MM.

*Edit- I think she would have said some things differently after getting some feedback, but it's still useful for clans that have never given any thought to how MM works.

---

That video didn't really sound like it was based on info from the devs. She talked about theories and deductions a lot, or so it seemed to me. And she did use spAnser's data, such as relating the supposed weight penalty of a 9.5 to an eagle tower.

And there's practical stuff - like she was emphatic that running x.5s high up is bad. Yet my clan does that all the time. We've even done wars with 6 or 7 9.5s up top and no TH10s (not very recently as we have more 10s now). Yes, they match infernos, which is what I'd expect. So if someone was expecting their 40/40 9.5 to match a TH9 they will be bitterly disappointed. However if you go into it not expecting an advantage, but expecting a middling TH10, then the match is in fact balanced.

Quote:

---

Originally Posted by OnyxDS

T So if someone was expecting their 40/40 9.5 to match a TH9 they will be bitterly disappointed.

---

why ? we see it all the time

Quote:

---

Originally Posted by ryryshouse

Oh, so you have a set of equations in which you solve for all 50 variables? no? okay then. Glad we got past that quickly.

The only 3 things you know are ; 1. your defensive weight ( and you dont know if this is what is indeed used in the algorithm ) and 2. your placement on the map and TH lvl. finally, you have this information of your opponents.

So until SC flops out the algorithm you have a pile of speculation.

---

What point are you making? We dont see offensive weight so its useless to discuss and theories around it?

do you belive in wind?

do you believe in gravity?

do you believe in private and commercial debt levels in an economy?

are you saying if we cant see something, or cant be 100% sure it exists, we need to never discuss it, evaluate it, look for its effects, look for its interactions, or look for its shadow.

we cant be certain of anything, we dont even know if our bases have defensive weight, many smart programmers tell you the matchmaking algorythym assings and modifies all weigh server site, and the gold and elixor amounts are simply the loot per base the opponent recieves, and that your war defensive weight, if it exists, is a totally seperate set of numbers and modifications, to prevent people from knowing the values. But even so, its very useful to try understand how it works, since we have a lot of data, and experience.

also, certain people have extracted offensive weights, I assume from the client side game files, indicating that at least their is more concrete visual evidence for its existence than gravity.

why are you being so abrasive, your argument moved from offensive weight doesnt exist, to we cant see it so we cant say anything about it.

how sure are you you are right, and what makes you so sure?

Quote:

---

Originally Posted by Vikingchief

What point are you making? We dont see offensive weight so its useless to discuss and theories around it?

do you belive in wind?

do you believe in gravity?

do you believe in private and commerciak debt levels in an economy?

are you saying if we cant see something, or cant be 100% sure it exists, we need to never discuss it, evaluate it, look for its effects, look for its interactions, or look for its shadow.

we cant be certain of anything, we dont even know if our bases have defensive weight, many smart programmets tell you the matchmaking algorythym assings and modifies all weigh server site, and the gold and elixor amounts are simply the loot per base the opponent recieves, and that your war defensive weight, if it exists, is a totally seperate set of numbers and modifications, to prevent people from knowing the values.

also, certain people have extracted offensive weights, I assume from the client side game files, indicating that at least their is more concrete visual evidence for its existence than gravity.

why are you being so abrasive, your argument moved from offensive weight doesnt exist, to we cant see it so we cant say anything about it.

how sure are you you are right, and what makes you so sure?

---

Not sure why you are mad bro. you can demonstrate gravity and wind. stopped reading there

Quote:

---

Originally Posted by ryryshouse

why ? we see it all the time

---

Funny. Once our TH9's get to 20/20 they start getting matched with th10's.

Quote:

---

Originally Posted by ryryshouse

Not sure why you are mad bro. you can demonstrate gravity and wind. stopped reading there

---

Im not mad, im confused what your angle is, and trying to figure you out.

you can demonstrate gravity, you can demonstrate offensive weight. In fact you cant demonstrate gravity, you can demonstrate that objects are drawn to one another, but you cannot demonstrate objectively that this is what we have defined as gravity, or if it is another type of force that caused it.

the results you get in the matchmaker, seam to demostrate an offensive weight, as the attration between large objects at relatively large distances seam to demonstrate gravity.

so my question to you is what are you saying?

are you saying there is no significant offensive weight and your sure of it?

are you saying we cant see offensive weights so we should not try see its effects.

are you saying we should not discuss the matchmaker because we cant be sure were right?

what are you saying, is what im saying.

Quote:

---

Originally Posted by ryryshouse

why ? we see it all the time

---

Really? TH9.5s with max TH9 defence and walls and 40/40 heroes and all relevant troops (say, golem, valk, bowler, hound, baby, hog, freeze, zap, haste) maxed at TH10 matching to your TH9s?

Quote:

---

Originally Posted by OnyxDS

Oh dear :( I used to like Heather's videos. My suggestion would be to disregard everything in there. Apart from the bit about clan level and player trophies not counting I'm not sure there is a single thing in that video that I could agree with. The fact she bases some of her analysis on spAnser's data isn't the worst of it!

---

For some reason I got the impression that she spoke with someone at SC (maybe a dev?) to get this information. Wherever I found this originally (was linked here on the forum a while back) I think I remember someone stating as much.

Quote:

---

Originally Posted by OnyxDS

The problem is that subjectively people seem to feel that x.5s got hurt the most and very defenceless weren't affected so much. Which this weight penalty just doesn't seem to explain....

...To come to the conclusion that there is such a weight penalty you'd have to start from the observation that x.5s have been only slightly affected, and defenceless TH11s have been punished with extreme prejudice and will automatically lose every war. Is that anyone's observation? We haven't matched a lot of truly defenceless recently, but I wouldn't say it's what I'm seeing.

---

To me it actually makes sense using this formula why a .5 is hurt more than a defenseless base. A defenseless TH11 with max offense can come in at 160 weight using this formula. This still makes them weigh less than a max TH10.

A TH10.5 using this formula will weigh enough to pull in a full 11. Also a 9.5 will weigh enough to pull in full 10.

So yes they are penalized more in defensive weight than a .5, but since they are so lopsided the penalty isn't enough to pull a TH with the weight of their true attacking level.

Quote:

---

Originally Posted by OnyxDS

And there's practical stuff - like she was emphatic that running x.5s high up is bad. Yet my clan does that all the time. We've even done wars with 6 or 7 9.5s up top and no TH10s (not very recently as we have more 10s now). Yes, they match infernos, which is what I'd expect. So if someone was expecting their 40/40 9.5 to match a TH9 they will be bitterly disappointed. However if you go into it not expecting an advantage, but expecting a middling TH10, then the match is in fact balanced.

---

That's been exactly our experience with a couple 9.5s at the top & gave Heather that feedback. Occasionally other offensively strong TH9.5s show up. Once in a blue moon, we've come across the advantage of having our bottom 9.5 face a maxed TH9. On paper, I'd say the clan with 2 full TH10s defenses should have an advantage over our 9.5s, but we typically beat them on stars and/or percentage at the top.

I have an anomaly here. I could be something or nothing. Need the experts to have glance.

Total-----0--4-26---------------------------------------------393-370--23--1322--194 Adv.
Total-----0--4-26---------------------------------------------254-238--08--1367--108 Adv.

Lanes with 10+ "paper advantages": #02-#04-#07-#10-#11-#12-#16-#19-#22-#25

The defenses structures are the same in this match up, confirmed by the close defense weights. However we still had quite an advantage on offense. I have never seen a match like this where defenses were so close and still allowing a big difference in offense as highlighted by the Lane comparisons. We won in other lanes as well, just not a 10 difference.


Also,

Something has felt off the last few wars I was documenting in the fact that the top few villages were allowed to have fairly large differences as compared to the normal practically identical weights. It happened in my documented clan matches #47 and a similar occurrence happened in my family clan this war. The fam clan was a bit more severe.

In my family clan this was the break down:

TH11 45/45/20
TH10.1 16/17/5 (nothing from 11 but level 5 warden)
TH10 10/13
TH9.5 15/16
TH9 12/10
TH9 11/12
TH8.5
TH8.5
TH8.5
TH8
TH9m 2/1
TH8
TH7
TH8m
TH6


Opposing clan:

TH10 25/23
TH10 17/18
TH10 12/12
TH10 17/15
TH9 11/11
TH9 13/13
TH9 10/11
TH9 10/9
TH9 10/7
TH9 10/7
TH8.5
TH8
TH8
TH7
TH7

There probably wasn't a huge difference in defense weight between #2's but #1 was crazy. Granted they had heavier th9's farther down, but it was always my understanding that SC matched the top closely and not so much the bottom. I almost ready to say we are entering a new erra in the MM. If this is true, it a shame all the work we put in recently lol, just to have it all thrown out the window.....

This is the top line from match #47

#01-10----0--2--3--7-7-7-5-6-6-7-4-5-3-4-2-6-6-5-5-2-3-2-240---31--37--00---86
#01-10.5--0--2--4--7-7-8-7-6-7-7-4-5-5-5-4-6-6-5-5-3-4-3-240---41--45--16---93

Thoughts?

The Fam clan won fyi...3 starred top 3 missed a star on 4 and then got everything else but #8 tripled. As usual lol, the "fun players" couldn't get all the 9's taken care of despite me taking out #5 #6 #7 and #10.

Quote:

---

Originally Posted by SuperStorm103

I have an anomaly here. I could be something or nothing. Need the experts to have glance.

Total-----0--4-26---------------------------------------------393-370--23--1322--194 Adv.
Total-----0--4-26---------------------------------------------254-238--08--1367--108 Adv.

Lanes with 10+ "paper advantages": #02-#04-#07-#10-#11-#12-#16-#19-#22-#25

The defenses structures are the same in this match up, confirmed by the close defense weights. However we still had quite an advantage on offense. I have never seen a match like this where defenses were so close and still allowing a big difference in offense as highlighted by the Lane comparisons. We won in other lanes as well, just not a 10 difference.

---

Did you opponents have very balanced troop upgrades per base (upgraded all teoops) while you guys were more selective?

do they have a lot more upgraded witch accounts?

total freeze level?

edit: just looked through match, not that much structural difference in offence troops, your opponents got a terrible match in terms of heroes.


it must be within some sort of allowance in the current matchmaker, reminds me of the very first clash matchmaker, where you could have 5/5 heroes facing 30/30 heroes across the board.

if they dont already, they should seperate heroe from offence and defence weights and have their own lists, or they should have a sanity check on them.

i wonder if they noticed people have stopped upgrading heroes, and have dropped the hero weighting as response?

on thing the hero weighting, as a total of iether defence or offence is very small, so im suprised that a lot more mismatches dont happen with heroes, I assume they have some check on them.

so i think your offence and defence in both clans as close, just heroes were a total mismatch.

In my commentary under match #47 I said it appears that hero weights may have been reduced.

Neither clan in the above example had selective offense. Everything was *normal* upgrades. Their 10's all had max valks though. Still if you parse through the previous matches the top bases are almost identical weight wise. This is a recent thing. #48 is documented as well.

Quote:

---

Originally Posted by SuperStorm103

In my commentary under match #47 I said it appears that hero weights may have been reduced.

Neither clan in the above example had selective offense. Everything was *normal* upgrades. Their 20's all had max valks though. Still if you parse through the previous matches the top bases are almost identical weight wise. This is a recent thing. #48 is documented as well.

---

Heres the thing

lets say max king weighs 900 defence and 900 offence, Queen 1800 defence and 1800 offence, and warden 1600 offence and defence.

now thats all max heroes, even at that weight they weigh so little relatively you would think mismatches would be very common, so iether they relaxed some sanity check, have stopped matching heroes, have increased hero weight and stopped matching or sanity check (letting the increased weight mean less sanity check is needed), or reduced the weight and their never was a sanity check.

or those hero weights are horrible wrong (im sure they are).

need some more time to see if this is a new trend.

edit: more sample needed on top base weight too

Quote:

---

Originally Posted by SuperStorm103

In my commentary under match #47 I said it appears that hero weights may have been reduced.

Neither clan in the above example had selective offense. Everything was *normal* upgrades. Their 10's all had max valks though. Still if you parse through the previous matches the top bases are almost identical weight wise. This is a recent thing. #48 is documented as well.

---

This is a way too small of a sample, but the same thing is happening in our current war.

Here are the visible defensive weights and hero levels in our current war (us vs them):
71 - 72 (30/30 - 17/17)
69 - 71 (24/25 - 10/10)
67 - 68 (10/10 - 17/16)
56 - 57 (5/5 - 6/0)
54 - 54 (10/0 - 10/0)
51 - 50 (8/0 - 8/0)
43 - 49 (5/5 - 7/0)
39 - 39 (4/0 - 5/0)
27 - 38 (0/0 - 3/0)
16 - 5 (0/0 - 0/0)
---------------------------
total weights added up: 493 - 503
total king levels added up: 96 - 83
total queen levels added up: 75 - 43

As you can see the weight advantage of the enemy team on the top is nowhere near the huge difference we are used to lately, but instead the weights are pretty close. In contrast to that our advantage in queen levels in huge. On top of that we have 3 accounts with th10 troops and 2 accounts with th9 troops, while they have 2 accounts with th10 troops and 2 accounts with th9 troops (one of those without a queen). We definitely have a significant offensive advantage without the usual disadvantage in defensive weights.

I am too skeptical to draw conclusions from only 2 samples, but a trend does appear to emerge. Can't wait to see more data.

Quote:

---

Originally Posted by dorsan

This is a way too small of a sample, but the same thing is happening in our current war.

Here are the visible defensive weights and hero levels in our current war (us vs them):
71 - 72 (30/30 - 17/17)
69 - 71 (24/25 - 10/10)
67 - 68 (10/10 - 17/16)
56 - 57 (5/5 - 6/0)
54 - 54 (10/0 - 10/0)
51 - 50 (8/0 - 8/0)
43 - 49 (5/5 - 7/0)
39 - 39 (4/0 - 5/0)
27 - 38 (0/0 - 3/0)
16 - 5 (0/0 - 0/0)
---------------------------
total weights added up: 493 - 503
total king levels added up: 96 - 83
total queen levels added up: 75 - 43

As you can see the weight advantage of the enemy team on the top is nowhere near the huge difference we are used to lately, but instead the weights are pretty close. In contrast to that our advantage in queen levels in huge. On top of that we have 3 accounts with th10 troops and 2 accounts with th9 troops, while they have 2 accounts with th10 troops and 2 accounts with th9 troops (one of those without a queen). We definitely have a significant offensive advantage without the usual disadvantage in defensive weights.

I am too skeptical to draw conclusions from only 2 samples, but a trend does appear to emerge. Can't wait to see more data.

---

i wonder if supercell noticed a dropoff in hero upgrades? I hope they dont take this too far the other way though.

Quote:

---

Originally Posted by Vikingchief

i wonder if supercell noticed a dropoff in hero upgrades? I hope they dont take this too far the other way though.

---

I think it's too soon for any conclusions.

Quote:

---

Originally Posted by Azrial

I really wonder why this video doesn't get more attention on the forums -- it's basically a recipe for how to optimally upgrade for war post October update (keep your offense about 10% ahead of your defense).

---

I guess because most of it is speculation, supposition, her own theories how MM works atm...its not a coc developer channel with facts at all.

we just found a war before, its a huge missmatch in defences, not whinging though, but id like to share what i think ive found lately.
hero levels matter alot, 9.5s with high heros will basicly pull the weight as an infernos th10. we use to run quite a few engineers ourselves but we r in the process or turning that around, we done a few wars with minimal .5s and we faced fairly normal wars, we decided to putt in the lob sided for a bigger war and to see what we might draw. and ill share that

Our team
TH11 38/44/20
TH11 35/45/20
TH11 25/40/20
TH11 35/35/20 (no eagle)
TH11 25/30/5 (no eagle)
TH10 30/35
TH11 30/45/20 (lvl1 bows, infernos, no eagle)
TH11 18/26/8 (lvl1 infernos, no eagle)
TH10 31/33 (no infernos)
TH10 30/32 (no infernos)
TH10 20/24 (no infernos)
TH10 20/20 (no infernos)
TH10 23/20 (no infernos)
TH9 27/29
TH10 25/29 (no infernos)
TH9 24/21
TH9 16/16
TH9 10/10
TH9 10/8
TH8 10
TH8 8
TH8 4
TH9 6/3 (no bows, low level everything)
TH7 1
TH6

There Team
TH11 21/28/20
TH11 15/24/13
TH11 14/20/14
TH11 28/30/8
TH11 12/15/6
TH11 19/19/7
TH11 10/9/5
TH11 12/17/5
TH11 10/11/9 (no infernos, no eagle)
TH11 14/17/8 (no infernos, no eagle)
TH10 18/18 (no infernos)
TH10 8/5 (no infernos)
TH11 9/12/5 (no infernos, no eagle)
TH11 6/5/0
TH10 10/10 (no infernos)
TH10 9/10 (no infernos)
TH10 8/12 (no infernos)
TH10 7/7 (no infernos)
TH9 7/3
TH8 10
TH8 7
TH8 5
TH8 2
TH8 6
TH5

so looking at that they have 8 eagles to our 3, alot of 9.5s and engineer setups on both sides, so yes both our fault, but what ive been noticing the past few months is hero levels play a big part, how else can there clan have 8 eagles to our 3, look at there heros, mostly all rubbish, ours are all solid in the top half. i think we will win, hopefully.
post your thought on that matchup about the recent matchmaking changes.

With the huge mismatch we came close 69-70

Quote:

---

Originally Posted by Domz69

With the huge mismatch we came close 69-70

---

So U are able to 3* 2 Eagle bases? it is a good job, but not enough to win?

So Hero vs Inferno/eagle can be disadvantage for the High Heroes Team.

It will discourage people to grind Heroes, the main income sources for SC

Quote:

---

Originally Posted by Domz69

so looking at that they have 8 eagles to our 3, alot of 9.5s and engineer setups on both sides, so yes both our fault, but what ive been noticing the past few months is hero levels play a big part, how else can there clan have 8 eagles to our 3, look at there heros, mostly all rubbish, ours are all solid in the top half. i think we will win, hopefully.
post your thought on that matchup about the recent matchmaking changes.

---

It's what we've been calling Heroes vs Infernos wars. In our experience they work out reasonably matched - but a lot depends on the base quality of the "infernos" team. We've had, for example, a war with our 2 TH10s and some 9.5s matched to 2 full eagle bases and a load of inferno bases - and got a perfect score because their bases were so bad. And then had a war facing just TH10s, no 11s and struggled to 3* any of them.

Quote:

---

Originally Posted by SuperStorm103

I have an anomaly here. I could be something or nothing. Need the experts to have glance.

Total-----0--4-26---------------------------------------------393-370--23--1322--194 Adv.
Total-----0--4-26---------------------------------------------254-238--08--1367--108 Adv.

Lanes with 10+ "paper advantages": #02-#04-#07-#10-#11-#12-#16-#19-#22-#25

Thoughts?

---

Even though the hero comparison on this match is greater, this overall matchup looks similar to match 43. This is what I like to attribute towards recent war performance. They had a streak of 7 wins, not certain if they are a team where they normally lose the 8th match or if this was their greatest streak to date. You were coming off a streak of 3 and have hit 8-9 streaks in the past. In both scenarios (match 43 and 48) I believe the MM determined to give the clans with the better recent war performance a harder match.

I can't comment on your family clan match without the additional details you normally provide.

Our new clan is still struggling to break a win streak of 1/2. We have a 40% win record, and I constantly see where if we win, the next match becomes increasingly difficult. If we lose, the matches slowly become easier and easier until we win. We just came off a 3 loss streak (our second loss was due to a player not attacking, so they were kicked since this wasn't their first infraction) and the match we were presented with was overwhelmingly easy and won by 7 stars.

Quote:

---

Originally Posted by Shadrach777

Even though the hero comparison on this match is greater, this overall matchup looks similar to match 43. This is what I like to attribute towards recent war performance. They had a streak of 7 wins, not certain if they are a team where they normally lose the 8th match or if this was their greatest streak to date. You were coming off a streak of 3 and have hit 8-9 streaks in the past. In both scenarios (match 43 and 48) I believe the MM determined to give the clans with the better recent war performance a harder match.

I can't comment on your family clan match without the additional details you normally provide.

Our new clan is still struggling to break a win streak of 1/2. We have a 40% win record, and I constantly see where if we win, the next match becomes increasingly difficult. If we lose, the matches slowly become easier and easier until we win. We just came off a 3 loss streak (our second loss was due to a player not attacking, so they were kicked since this wasn't their first infraction) and the match we were presented with was overwhelmingly easy and won by 7 stars.

---

Fyi, our best streak is 28. I totally agree with you that poor recent performance in war log returns stupid easy matches.

Have some more supporting evidence:

Match #49 Win Streak = 4 0

----TH----E--I--X--B A G G W B W H D P D M M H V G W H B Army--BK--AQ--GW---BW

#01-10----0--2--3--7-7-7-5-6-6-7-4-5-3-4-2-6-6-5-5-2-3-2-240---31--37--00---86
#01-11----1--2--4--6-6-7-2-5-6-7-4-5-5-3-2-2-6-5-5-2-2-2-240---21--28--05--103

Total-----0--4-26---------------------------------------------412-391--23--1399--251 Adv.
Total-----1--6-32---------------------------------------------185-168--15--1464---90 Adv.

Lanes with 10+ "paper advantages": #04-#09-#10-#11-#12-#14-#16-#20-#22-#25

Again we see huge hero and offence difference but only minor defense gains for the opposing clan. They are squeezing 1 eagle, 2 infernos, and 6 xbows into the weight of a non maxed th9. And the top (strictly speaking is not a close match). You can tell by the paper advantage totals (largest difference to date) that we dominated most of the lanes in a lane to lane comparison with 9 out of 25 having +10 advantage. 6 of those 9 were +15. Biggest lane Mis-match was #9 at 20 to 0. Lane #4 was 19 to 1

Quote:

---

Originally Posted by SuperStorm103

Fyi, our best streak is 28. I totally agree with you that poor recent performance in war log returns stupid easy matches.

Have some more supporting evidence:

Match #49 Win Streak = 4 0

----TH----E--I--X--B A G G W B W H D P D M M H V G W H B Army--BK--AQ--GW---BW

#01-10----0--2--3--7-7-7-5-6-6-7-4-5-3-4-2-6-6-5-5-2-3-2-240---31--37--00---86
#01-11----1--2--4--6-6-7-2-5-6-7-4-5-5-3-2-2-6-5-5-2-2-2-240---21--28--05--103

Total-----0--4-26---------------------------------------------412-391--23--1399--251 Adv.
Total-----1--6-32---------------------------------------------185-168--15--1464---90 Adv.

Lanes with 10+ "paper advantages": #04-#09-#10-#11-#12-#14-#16-#20-#22-#25

Again we see huge hero and offence difference but only minor defense gains for the opposing clan. They are squeezing 1 eagle, 2 infernos, and 6 xbows into the weight of a non maxed th9. And the top (strictly speaking is not a close match). You can tell by the paper advantage totals (largest difference to date) that we dominated most of the lanes in a lane to lane comparison with 9 out of 25 having +10 advantage. 6 of those 9 were +15. Biggest lane Mis-match was #9 at 20 to 0. Lane #4 was 19 to 1

---

We got 2 wars in a row with huge hero advantage but without enemy infernos.

In the current war we still have a hero advantage, but this time enemy #1 does have 2 maxed infernos and 3 x-bows.

edited to add: We are on 6 win streak at the moment and enemy is on 5. (although the match after the 6 wins is a draw and enemy has a loss after the 5 wins and below 50% total win rate)

Quote:

---

Originally Posted by dorsan

We got 2 wars in a row with huge hero advantage but without enemy infernos.

In the current war we still have a hero advantage, but this time enemy #1 does have 2 maxed infernos and 3 x-bows.

edited to add: We are on 6 win streak at the moment and enemy is on 5. (although the match after the 6 wins is a draw and enemy has a loss after the 5 wins and below 50% total win rate)

---

This getting to be the biggest streak we had in awhile. Our current match doesn't seem to difficult. Will most likely extend it to 6. I'm thinking by accident our current clan may be structured to accommodate the new MM in a good way.

Oh and the evidence was to my statement that hero weights have been reduced big time, and also potentially big defense weights (eagle, inferno, and xbow).

Quote:

---

Originally Posted by OnyxDS

Oh dear :( I used to like Heather's videos. My suggestion would be to disregard everything in there. Apart from the bit about clan level and player trophies not counting I'm not sure there is a single thing in that video that I could agree with. The fact she bases some of her analysis on spAnser's data isn't the worst of it!

---

I have to agree here. While I also enjoy her content, I'm actually perplexed about her most recent video. Historically, I've even pointed her towards an old Ash video that promoted spAnser's data, and the issues around it, and yet still she's chosen to use this information to formulate her content.

All this does is create more misinformation within the community!:unsure:

Edit : just to confirm, Heather has stated that the source of her offence data is indeed :-

https://spanser.net/CoCGuide

Oh dear!!:facepalm:

Quote:

---

Originally Posted by SuperStorm103

Oh and the evidence was to my statement that hero weights have been reduced big time, and also potentially big defense weights (eagle, inferno, and xbow).

---

Speculation on my part, but is it possible SC saw an issue in the upcoming huge update, and therefore snuck in a change during the March tweak of witches and dragons? They could be performing additional tests before releasing the update. I can imagine people currently beta-testing the new content complained about the uneven weights of heroes and offense with the new content, and SC tweaked it back down.

I'll collect our numbers for our match when we enter battle day and display them in your format. I no longer collect data for my roster, but the last match we won and our current match (coming off a loss) look overwhelmingly easier than normal. It's the classic heroes versus ITs, but this feels more unfair for our opponent this round.

Quote:

---

Originally Posted by Shadrach777

Speculation on my part, but is it possible SC saw an issue in the upcoming huge update, and therefore snuck in a change during the March tweak of witches and dragons? They could be performing additional tests before releasing the update. I can imagine people currently beta-testing the new content complained about the uneven weights of heroes and offense with the new content, and SC tweaked it back down.

---

Highly possible. Personally, I think they've tweaked the MM at least 4 times since the October update.

Quote:

---

Originally Posted by Skrags

I have to agree here. While I also enjoy her content, I'm actually perplexed about her most recent video.

---

Not familiar with the earlier content but the vid left me confused as well. There's a claim that info is coming from Devs, then a tidbit here and there, then what seemed to be a lot of stuff lifted from the weight tax idea on the forum with no disclaimer that we've gone from inside information to forum speculation. Unless SC implemented the weight tax?

But next it was forum ideas about how to avoid matching engineers, again no disclaimer that we've gone from dev info to forum speculation (good speculation IMO, but still). At least that's what I remember bugging me about the vid.

Quote:

---

Originally Posted by Shadrach777

Speculation on my part, but is it possible SC saw an issue in the upcoming huge update, and therefore snuck in a change during the March tweak of witches and dragons? They could be performing additional tests before releasing the update. I can imagine people currently beta-testing the new content complained about the uneven weights of heroes and offense with the new content, and SC tweaked it back down.

---

The matchmaker is 100% server side code. They don't need to tweak it during any update. They can just pause the matchmaking for a few minutes, do an update and resume it without players ever even noticing anything was changed.

Quote:

---

Originally Posted by dorsan

The matchmaker is 100% server side code. They don't need to tweak it during any update. They can just pause the matchmaking for a few minutes, do an update and resume it without players ever even noticing anything was changed.

---

I can see the actual matchmaking algorithm as server side only, but from a large distributed application stand-point, it makes more sense for the client application to be calculating all the aspects of the roster as it is being picked by co-leader/leader. For instance, determine the offensive weight of my team, defensive weight of my team, calculate number of hero levels, IT count, xbow count etc... Whatever the matchmaker currently uses for determining a match. It's less taxing on the server to perform those computations on the client application and just provide the information (especially because it's static) to the server so it can perform solely match-making.

Also keep in mind, with events, they can sneak in whatever changes they want every 3-4 days without requiring patches.

Quote:

---

Originally Posted by Shadrach777

I can see the actual matchmaking algorithm as server side only, but from a large distributed application stand-point, it makes more sense for the client application to be calculating all the aspects of the roster as it is being picked by co-leader/leader. For instance, determine the offensive weight of my team, defensive weight of my team, calculate number of hero levels, IT count, xbow count etc... Whatever the matchmaker currently uses for determining a match. It's less taxing on the server to perform those computations on the client application and just provide the information (especially because it's static) to the server so it can perform solely match-making.

Also keep in mind, with events, they can sneak in whatever changes they want every 3-4 days without requiring patches.

---

Except that... they have to assume that the client application has been entirely subverted by mod software, and nothing it sends to the server can be assumed to be true.

Quote:

---

Originally Posted by OnyxDS

Except that... they have to assume that the client application has been entirely subverted by mod software, and nothing it sends to the server can be assumed to be true.

---

A fair point, but I think it's safe to assume the client application is clean. I would think one would not degrade the overall server performance for the few that might cheat for the sake of the majority that don't. I would lean towards the developers creating a game that would be fun for all and look at preventing cheating as a secondary objective.

I've only ever known of the psychic octopus since I've started playing (last 18 months), I'm not aware if there were active measures to prevent cheating before then. Not to mention, the psychic octopus has ways of determining if a user has modified files.

The paradigm for most distributed applications is to reduce the strain on the server and put as much on the client as possible, because as the user base grows, the greater the strain on the server. The user base growing has no affect on the client application.

Quote:

---

Originally Posted by OnyxDS

Oh dear :( I used to like Heather's videos. My suggestion would be to disregard everything in there. Apart from the bit about clan level and player trophies not counting I'm not sure there is a single thing in that video that I could agree with. The fact she bases some of her analysis on spAnser's data isn't the worst of it!

---

She's the female version of Blombardo if you ask me (she speaks with such confidence!)

I don't follow her vids, but ever since Skrags recommended one of her videos that suggests matching Th vs Th as a possible solution to war MM I knew her material was not worth watching.

Quote:

---

Originally Posted by Shadrach777

A fair point, but I think it's safe to assume the client application is clean. I would think one would not degrade the overall server performance for the few that might cheat for the sake of the majority that don't. I would lean towards the developers creating a game that would be fun for all and look at preventing cheating as a secondary objective.

I've only ever known of the psychic octopus since I've started playing (last 18 months), I'm not aware if there were active measures to prevent cheating before then. Not to mention, the psychic octopus has ways of determining if a user has modified files.

The paradigm for most distributed applications is to reduce the strain on the server and put as much on the client as possible, because as the user base grows, the greater the strain on the server. The user base growing has no affect on the client application.

---

You are horribly wrong.

Game developers cringe at the idea of having anything that can impact the gameplay in an online multiplayer game on client side. You did find some games with more CPU expensive stuff going on on the client side in the past, but that doesn't happen anymore. Calculating base weights and finding matches is not resource expensive at all. We are talking about a few thousand CPU cycles, no self respecting developer would put that on the client side ever.

Quote:

---

Originally Posted by ReplicantRachel

She's the female version of Blombardo if you ask me (she speaks with such confidence!)

I don't follow her vids, but ever since Skrags recommended one of her videos that suggests matching Th vs Th as a possible solution to war MM I knew her material was not worth watching.

---

:saywhat:

I believe Rachel, you are referring to the video linked by the OP in this thread :-

https://forum.supercell.com/showthre...s-Fix-Solution

and my response to that video can be found on page 2, namely :-

"Everybody is entitled to an opinion. But I promise you that this idea has been discussed to death on the forums by many informed players. I'm a Maxer myself , but even I realise the huge flaws of Heathers suggestion. She just hasn't thought it through....and btw, I enjoy her channel"

My opinion on "matching TH vs TH as a possible solution to war MM" is very well documented here on the forums!:)

Quote:

---

Originally Posted by TankSinatra

Not familiar with the earlier content but the vid left me confused as well. There's a claim that info is coming from Devs, then a tidbit here and there, then what seemed to be a lot of stuff lifted from the weight tax idea on the forum with no disclaimer that we've gone from inside information to forum speculation. Unless SC implemented the weight tax?

But next it was forum ideas about how to avoid matching engineers, again no disclaimer that we've gone from dev info to forum speculation (good speculation IMO, but still). At least that's what I remember bugging me about the vid.

---

I agree. This is what confused me as well. I know many have commented not to trust this source but I'm still on the fence about it. If she did in fact use spAnswers for her proof then I'm inclined to dismiss just about everything she said. However if she did get this information from a dev I think it would be the holy grail for MM info (at least at the time of the posting).

What boggles me is that if it was just her opinion then why did she support it with numbers pulled out of nowhere? Even if she had "numbers" from spAnswers that wouldn't explain how she performed calculations on those numbers.

Quote:

---

Originally Posted by dorsan

Game developers cringe at the idea of having anything that can impact the gameplay in an online multiplayer game on client side. You did find some games with more CPU expensive stuff going on on the client side in the past, but that doesn't happen anymore. Calculating base weights and finding matches is not resource expensive at all. We are talking about a few thousand CPU cycles, no self respecting developer would put that on the client side ever.

---

This! :thumbsup:

Quote:

---

Originally Posted by Skrags

:saywhat:

I believe Rachel, you are referring to the video linked by the OP in this thread :-

https://forum.supercell.com/showthre...s-Fix-Solution

and my response to that video can be found on page 2, namely :-

"Everybody is entitled to an opinion. But I promise you that this idea has been discussed to death on the forums by many informed players. I'm a Maxer myself , but even I realise the huge flaws of Heathers suggestion. She just hasn't thought it through....and btw, I enjoy her channel"

My opinion on "matching TH vs TH as a possible solution to war MM" is very well documented here on the forums!:)

---

No, it was a different thread but I can't find it in search.
I even remember commenting on it.

But I didn't mean to put you on the spot lol.