Another Engineering Thread with Solution

Our clan just started a 30v30 war where we have a single TH11 and we matched 14 TH11s. Yes, they can attack all but 2 of our bases with TH11s.

Supercell has said, "Dealing with engineering is hard. We are working on it." But I've got to ask, don't you think war outcome should be determined based on skill? Clan wars shouldn't be a pre-war arms race to see who can put the most engineered bases in the lineup.

I'm all for letting people build their bases however they want. Making people max out before upgrading is not the solution. But the problem is that with engineering, war matchups just aren't fair. How do you make it fair? Here is my proposal:

The war matchmaking algorithm should calculate a projected final star outcome for each side. Pair the top offensive war weight against the top defensive war weight and pull data from the database about what the expected result of an attack is. In statistical terms, you could use the average, the mean, the 75th percentile or however you want to do it. Using this, it'd come up with some sort of expected star number for an attack (like 2.3 stars for example).

With 2 attacks per person, you could repeat with the top offensive war weight vs. the second defensive war weight. Or you could just go 2nd best offense vs. the 2nd best defense. Or use some mix in between. However you choose to do it, it would be fairer than it is now. Add the expected stars from each attack and come up with a total for each side.

Given our current example of war with 1 TH11 vs. 14, once the algorithm starts calculating their #4 offensive weight vs. our #4 defensive weight on down, the expected result is going to be 3.0 stars or really close.

Add up all the numbers and compare to see if you have a war match that makes sense. In our current war, the expected stars on their side should be somewhere around 89 - they are likely to 2-star our 11 and 3-star all of our 10s down. Our expected stars might be around 80 based on average attacks of these war weights in Supercell's database. 89 compared to 80 = no way these clans should've ever been a match.

On the flip side, using my proposed algorithm, this clan would never match up with 14 full TH11-defense bases either. They'd get fair matchups as well.

I think of this as the Accuscore method of war matchmaking. Similar to how you can find expected results on many sporting events ahead of time.

Hope someone at Supercell sees this. This is the worst matchup we've gotten, but we regularly get other engineered matchups where we are at a crazy disadvantage like this. It's deflating to your clan morale.

Thank you for moving to feature requests.

Yeah, as expected they currently have 88 stars with 16 attacks left to go. Shame on them for missing on our 2nd highest TH10 5 times with TH11 dips, I guess. :rolleyes: Boy, can our clan take pride in that.

As you said there are 14 th 11's but all the 14 th 11's can be taken down because they are engineered. Even you have the chance of winning. Excluding 1 because you also have 1 th11 max.

They missed your th10 5 times with th11?
So it is clear that they are not fully engineered.They engineered only some troops or they are really bad at attacking.

Though they have 14 th11's i think u too have a very good chance of winning

Are you dealing with base engineering or roster engineering? It sounds like roster engineering

Quote:

---

Originally Posted by Lausaj

As you said there are 14 th 11's but all the 14 th 11's can be taken down because they are engineered. Even you have the chance of winning. Excluding 1 because you also have 1 th11 max.

They missed your th10 5 times with th11?
So it is clear that they are not fully engineered.They engineered only some troops or they are really bad at attacking.
Though they have 14 th11's i think u too have a very good chance of winning

---

Are you serious? You think there is any reasonable scenario where 14 TH11s vs 1 TH11 gives our clan a good chance of winning?

Yes, they do only have 5 bases with infernos where we have 6 bases with infernos. They've been able to throw 17 attacks at these bases - 3-starring 4 of them and 2-starring the other 2.

With only one TH11, most of us don't get to use the Grand Warden and max miners, max Bowler+Witch or any of these other spammy strategies that have given them easy 3-star against our TH10s. It's much harder to run a TH10 vs 10.5 attack than an 11 offense vs. TH10. And they have all the extra chances when they miss.

Quote:

---

Originally Posted by Tiler

Are you dealing with base engineering or roster engineering? It sounds like roster engineering

---

I'm not sure I know the difference. They have engineered bases in the fact that they have TH11 max offense and varying lower levels of defense. The have some TH3s at the bottom of the war line-up as well.

I mean, I get what's going on and it happens to us routinely. They are taking advantage of the matchmaking algorithm to get super easy war wins. I'd like to see the changes I proposed in my initial post implemented to provide wars that are based on skill and not determined by who can game the system to bring in as many high offense/low defense bases as possible.

that makes sense

Quote:

---

Originally Posted by MazerFromAbove

It's much harder to run a TH10 vs 10.5 attack than an 11 offense vs. TH10. And they have all the extra chances when they miss.

---

That's their stratergy ,they gave up farming for war base.

Engineered bases are part of the game and not so easy to do. Engineered bases are equally difficult to construct like maxed out bases.

Haven't u ever been matched up with rushed out bases? Was that fair enough then?You are saying it is not fair because they have a different style of playing the game?

Some times it is tough sometimes it is easy, it is never the same.If u faced continuously maxed out bases wouldn't u be bored?

In the normal search don't u find higher th battles?Is that not fair too?
Thats how a game works.

Quote:

---

Originally Posted by Lausaj

That's their stratergy ,they gave up farming for war base.

Engineered bases are part of the game and not so easy to do. Engineered bases are equally difficult to construct like maxed out bases.

Haven't u ever been matched up with rushed out bases? Was that fair enough then?You are saying it is not fair because they have a different style of playing the game?

Some times it is tough sometimes it is easy, it is never the same.If u faced continuously maxed out bases wouldn't u be bored?

In the normal search don't u find higher th battles?Is that not fair too?
Thats how a game works.

---

Why not fix it to give fair war match ups? Why not have it close to a skill vs. skill determines the outcome of the war?

Your arguments are very strawman to me. I'm asking for fair matchups. The points you make don't even reference this:
You say it's hard to make an engineered base, we've had wars that were unfair on our side, etc, etc. And then, finish with "That's how the game works." Well, it doesn't have to work that way if they fix the matchmaking algorithm.

Quote:

---

Originally Posted by Lausaj

That's their stratergy ,they gave up farming for war base.

Engineered bases are part of the game and not so easy to do. Engineered bases are equally difficult to construct like maxed out bases.

Haven't u ever been matched up with rushed out bases? Was that fair enough then?You are saying it is not fair because they have a different style of playing the game?

Some times it is tough sometimes it is easy, it is never the same.If u faced continuously maxed out bases wouldn't u be bored?

In the normal search don't u find higher th battles?Is that not fair too?
Thats how a game works.

---

Yeah that doesn't make any sense friend. The fun of the game comes from the varying armies you can build, different strategies, and how you organize your base. Getting 14 TH11 maxed troops isnt fun. Thats not "a style of playing" thats exploiting a flawed matchmaking system.

14 rushed TH11's should fight against a similar 14 rushed TH11's. The point of competitive play is for it to be fair & may the best defense/offense win. When you have a greater offense and the same defence you obviously have an advantage and it no longer becomes a fair match.

What hes saying is to have an algorithm of defensive AND offensive matchmaking. As it stands currently only defensive health determines matchmaking. You can enginerr your base to have a TH9 level of health but have maxed bowlers when TH9 can't even access bowlers. So pit the TH9 with bowlers against another TH9 with bowlers. It becomes fair.

What you're saying is like if i were to find an exploit on a first person shooter where i'm under the map & killing people from somewhere out of sight but "its in the game & thats my playstyle. You could get out of the map too".

I am not saying it is fair to play the war that way, everything can not be perfect.There are loop holes everyone will try to take advantage of.

What u guys are saying is to compare every players offence and defence and match up war. The time taken for war match ups is around 20-30 mins on an average presently by just comparing defences and not offences.

If u try to compare the offences of each and every player mins of war search will turn into hours.

First when u start the war the system tries to allign ur postions in war based on defences not taking into considerations your offences and th level.If it takes into consideration ur offences and th level alligning the bases will be of lot tough because everyones play style is difficult. Some bases will be rushed,some will be maxed and some will be engineered. You have to depend on 1 or 2 factors for sure to start the main war search.

Let us suppose we have tried to start the war search by somehow considering all the factors. Now the real search begins, the system has to compare several opponents of the same th, same offence and defence lvl with the clan that started the war search at the time.Suppose it take a range of 1 or 2 lvl differeces in offence and defence because finding the exact opponent is impossible.

How much time do u think it will take to do all this? and what if no clan matches it?

It is for sure that it would take a lot of time to search then we would expect. Let us suppose it has somehow found a clan that almost matches your clan and war has been started.

After playing few wars say 10 from 10 different clans. U will surely start to find the same clan again and again because of the increased duration of the clan war and also because they fall in the same time zone as we do.If they start to avoid match ups with previous clan then surely the search duration will be extended and extended till u find a proper clan.

Would u like to battle the same clan again and again?Won't they know your all traps?Will it be fair enough then?

So, finally what happend after u reached the finall stage of fair game?

If i had a situation of playing against 1 clan again and again or waiting to find a opponent clan for several hours i would surely get bored of the game.

I am not opposing u guys i am just saying it is not easy to make things perfect and even if we make them we will find a difficult situation again, so it is better to accept the game as it is. Staff and other members try to make the things as fair as possible.

Quote:

---

Originally Posted by MazerFromAbove

I'm not sure I know the difference. They have engineered bases in the fact that they have TH11 max offense and varying lower levels of defense. The have some TH3s at the bottom of the war line-up as well.

---

The difference between roster engineering and base engineering is that roster engineers include maxed out bases as well as very low level bases at the bottom to counter balance the top bases and draw easier opponents. Base engineers are the type to upgrade weak defenses and powerful offenses without the need for TH3s at the bottom. I asked more out of curiosity than anything. If you're interested in discussing solutions to war matchmaking, I recommend checking out the clan wars sub. There are great discussions there

Quote:

---

Originally Posted by Lausaj

<snip>

Staff and other members try to make the things as fair as possible.

---

Staff has said outright that they believe engineering is an unfair manipulation of the matchmaker. I paraphrased a bit, but by reading any of Darian's posts in the clan wars sub, you can't disagree that that's the stance.

Offense and defense should be considered separately. Is OPs suggestion the perfect one? Probably not, but the jury's still out on what is

Quote:

---

Originally Posted by Lausaj

The time taken for war match ups is around 20-30 mins on an average presently by just comparing defences and not offences.

If u try to compare the offences of each and every player mins of war search will turn into hours.

---

Now that is a valid argument against my proposal - doing the calculations I suggested would take too much processing power or take too long. I'd rather hear that from Supercell's developers then just speculate that is the case, though.

In a sample 30v30 matchup, I think the system could immediately throw out a bunch of matchups before running projected star calculations. It wouldn't need to even calculate expected results for a clan with TH9s at the top vs. a different clan with 10s and 11s.

In then end though, the algorithm would come up with a fractional number less than 90 for each side and compare the 2. It's not that complicated. Coming up with projected star numbers is not a battle simulation, it's just pulling past results from a SQL database. In the end, it could choose to run the war if the total for each side was within 1 or 2. Or maybe raise this number as your wait goes on.

I personally would rather wait 2 hours for my matchup then get what we just had. I gladly would have clicked a button to forfeit the war, taken the loss, and immediately started a new war search.

Quote:

---

Originally Posted by MazerFromAbove

Here is my proposal:

The war matchmaking algorithm should calculate a projected final star outcome for each side. Pair the top offensive war weight against the top defensive war weight and pull data from the database about what the expected result of an attack is. In statistical terms, you could use the average, the mean, the 75th percentile or however you want to do it. Using this, it'd come up with some sort of expected star number for an attack (like 2.3 stars for example).

With 2 attacks per person, you could repeat with the top offensive war weight vs. the second defensive war weight. Or you could just go 2nd best offense vs. the 2nd best defense. Or use some mix in between. However you choose to do it, it would be fairer than it is now. Add the expected stars from each attack and come up with a total for each side.

---

If I understood your proposal, you're asking for a War Simulation to be carried out when matching. If you match clans that have equal War Simulation outcome, they'll probably be the best match for each other.

That said, Simulation requires a lot of time and resources, and there is no way Supercell (with its small team) will be able to write a good Simulation Algorithm. I'm not trying to disrespect Supercell; it is just the nature of the problem. I don't think Statistics can help a lot here. Small differences in Base Designs and/or Troops/Spells Placement can have a significant outcome on an attack. Whatever Statistics you collect will be plagued with data affected by that. Furthermore, there are a lot of army compositions (what troops at what level) and base builds (what defenses at what level) that you'll have a very hard time building that database.

Right now, I feel like it will solve the problem if it is implemented correctly. However, it is very hard to implement correctly, and it is going to require a lot of time and resources to run.

Quote:

---

Originally Posted by NormalCell

If I understood your proposal, you're asking for a War Simulation to be carried out when matching. If you match clans that have equal War Simulation outcome, they'll probably be the best match for each other.

That said, Simulation requires a lot of time and resources, and there is no way Supercell (with its small team) will be able to write a good Simulation Algorithm. I'm not trying to disrespect Supercell; it is just the nature of the problem. I don't think Statistics can help a lot here. Small differences in Base Designs and/or Troops/Spells Placement can have a significant outcome on an attack. Whatever Statistics you collect will be plagued with data affected by that. Furthermore, there are a lot of army compositions (what troops at what level) and base builds (what defenses at what level) that you'll have a very hard time building that database.

Right now, I feel like it will solve the problem if it is implemented correctly. However, it is very hard to implement correctly, and it is going to require a lot of time and resources to run.

---

See my response above. It's not a battle simulation. It's just mining their database for "What are the average results when a base with X war weight attacks a base with Y war weight in war." Very quick to get an average, a mean, the 75th percentile (if you want to base the calculation on what a "skilled" attacker can do), etc.

Quote:

---

Originally Posted by MazerFromAbove

See my response above. It's not a battle simulation. It's just mining their database for "What are the average results when a base with X war weight attacks a base with Y war weight in war." Very quick to get an average, a mean, the 75th percentile (if you want to base the calculation on what a "skilled" attacker can do), etc.

---

War Simulation is about distributing attacks and checking offense versus defense. Battle/Attack Simulation is about how you place troops/spells and at what time. I already replied to using Statistics.

Quote:

---

Originally Posted by NormalCell

I don't think Statistics can help a lot here. Small differences in Base Designs and/or Troops/Spells Placement can have a significant outcome on an attack. Whatever Statistics you collect will be plagued with data affected by that. Furthermore, there are a lot of army compositions (what troops at what level) and base builds (what defenses at what level) that you'll have a very hard time building that database.

---

Quote:

---

Originally Posted by NormalCell

Small differences in Base Designs and/or Troops/Spells Placement can have a significant outcome on an attack. Whatever Statistics you collect will be plagued with data affected by that. Furthermore, there are a lot of army compositions (what troops at what level) and base builds (what defenses at what level) that you'll have a very hard time building that database.

---

These things are what I would call "skill".

Given equal footing, which clan has more skills? Base design is a skill. Choosing your troop composition, doing your drops, and where to place spells during an attack are skill.

I don't want Supercell's war matching algorithm to try to accurately determine any war outcome based on a clan's skill or lack thereof. I just present a way I think there would be close to equal footing and skill would determine the winner.

Quote:

---

Originally Posted by MazerFromAbove

These things are what I would call "skill".

Given equal footing, which clan has more skills? Base design is a skill. Choosing your troop composition, doing your drops, and where to place spells during an attack are skill.

I don't want Supercell's war matching algorithm to try to accurately determine any war outcome based on a clan's skill or lack thereof. I just present a way I think there would be close to equal footing and skill would determine the winner.

---

I never said Wars shouldn't be determined by skill... You said data will be pulled from a database. Think about it, when will the data be entered into the database?
Supercell have to monitor attacks and start building this database. This is a lot of work of its own, but what I am pointing out is that this data will have a large variance in terms of army composition (the troops and spells used and their level) and base build (defenses and their level). My question is: how are you going to do this look-up? The offensive profile you're examining could have Archer level 4, Balloons level 6, Heal level 1, etc. What you have in your database may not match this. Now look at the build of the base you're examining. It could have 4 level 5 Cannons, no Archer Towers, 1 level 1 Inferno, etc. The combinations are many. The database will be sparse. Unless you group profiles (based on some Weight calculation), it'll be very hard to predict the outcome.
Forget about all of this, can you give me an example of 2 Clans that will be deemed a good match for each other based on your matchmaking proposal? We already know that exact matches are good matches, so can you give me a different example?

Quote:

---

Originally Posted by NormalCell

Supercell have to monitor attacks and start building this database. This is a lot of work of its own, but what I am pointing out is that this data will have a large variance in terms of army composition (the troops and spells used and their level) and base build (defenses and their level). My question is: how are you going to do this look-up? The offensive profile you're examining could have Archer level 4, Balloons level 6, Heal level 1, etc. What you have in your database may not match this. Now look at the build of the base you're examining. It could have 4 level 5 Cannons, no Archer Towers, 1 level 1 Inferno, etc. The combinations are many. The database will be sparse. Unless you group profiles (based on some Weight calculation), it'll be very hard to predict the outcome.

---

That's exactly what I'm saying. They already calculate an offensive and defensive war weight. They could search their database for average results when a certain offensive war weight attacks a certain defensive war weight. I'm not asking them to store any extra data they don't already have.

Quote:

---

Originally Posted by NormalCell

I never said Wars shouldn't be determined by skill... You said data will be pulled from a database. Think about it, when will the data be entered into the database?

---

I know at some point they have it in a database - you can access a few of your past war logs. Maybe they purge it as your war logs purge - but up until then, they know what results happen when a certain offensive war weight attacks a certain defensive war weight. With so many wars going on, there would still be a lot of data available to answer the question "What happens when offensive war weight x attacks defensive war weight y?"

Quote:

---

Originally Posted by NormalCell

Forget about all of this, can you give me an example of 2 Clans that will be deemed a good match for each other based on your matchmaking proposal? We already know that exact matches are good matches, so can you give me a different example?

---

Two clans would match up if their expected total star outcomes were close. It wouldn't have to be the same number of TH8s, 9s and 10s. It could be. But you could have a war with some engineered TH11s on one side vs max TH10s on the other side.

An example match would be: if expected war stars gained by clan A is 84.8 and expected war stars by clan B is 84.7, you got yourself a match - run the war.

Quote:

---

Originally Posted by MazerFromAbove

Two clans would match up if their expected total star outcomes were close. It wouldn't have to be the same number of TH8s, 9s and 10s. It could be. But you could have a war with some engineered TH11s on one side vs max TH10s on the other side.

An example match would be: if expected war stars gained by clan A is 84.8 and expected war stars by clan B is 84.7, you got yourself a match - run the war.

---

I'll assume what you said about the war data Supercell stores is true. What I want is an example of two rosters (5 versus 5 for simplicity) that you think will be matched. I'll give you two Clans, edit Clan B till it becomes a possible match for Clan A (not an exact match). Think about it as if you were the matchmaker and state your assumptions about attack success rates.

Clan A	Clan B
1 2 3 4 5 Offensive Weight TH10 TH9 TH9 TH8 TH8 Defensive Weight TH10 TH9 TH9 TH8 TH8	1 2 3 4 5 Offensive Weight TH10 TH11 TH9 TH8 TH8 Defensive Weight TH10 TH9 TH9 TH8 TH8

Right now Clan B has 2 TH11 attacks instead of 2 TH9 attacks.

Awesome use of tables. Thanks for engaging with me seriously on this - I appreciate it even if it seems that maybe we are at odds.

Quote:

---

Originally Posted by NormalCell

I'll assume what you said about the war data Supercell stores is true.

---

I don't know how this would be debatable that they have this data. They typically have entire war replays of every attack for your last few clan wars. The algorithm I'm suggesting just needs to know offensive war weight, defensive war weight, and final star outcome. Those 3 things should be there.

I guess it is possible that they could decide not to store the offensive war weight after the war is matched. The defensive weight is definitely still in around or at least derivable based on the loot offering. So I'd ask them to keep offensive war weight if they currently aren't, which is a small amount of data compared to troop and spell drops needed to show an attack replay.

Quote:

---

Originally Posted by NormalCell

What I want is an example of two rosters (5 versus 5 for simplicity) that you think will be matched. I'll give you two Clans, edit Clan B till it becomes a possible match for Clan A (not an exact match). Think about it as if you were the matchmaker and state your assumptions about attack success rates.

Clan A	Clan B
1 2 3 4 5 Offensive Weight TH10 TH9 TH9 TH8 TH8 Defensive Weight TH10 TH9 TH9 TH8 TH8	1 2 3 4 5 Offensive Weight TH10 TH11 TH9 TH8 TH8 Defensive Weight TH10 TH9 TH9 TH8 TH8

Right now Clan B has 2 TH11 attacks instead of 2 TH9 attacks.

---

Cool, I'll run this an example and try to be simple as well. Let's just do 10 attacks - 5 for each side.

Looks like we are assuming Clan A's keeps their offense and defense levels the same. So their best offense vs. defense would look like:
A1 vs B1
A2 vs B2
A3 vs B3
A4 vs B4
A5 vs B5

Clan B would look like:
B2 vs A1
B1 vs A2
B3 vs A3
B4 vs A4
B5 vs A5

I'm going to pull theoretical averages for these attacks because I don't know the average attack score when a TH11 hits a TH10 or a TH9 his a TH9 even. I would expect these numbers to come out of simple, quickly executed SQL statements from the database. So these are just guesses:

Expected attack results based on average offense vs defense war weight:
A1 vs B1: 1.8
A2 vs B2: 2.1
A3 vs B3: 2.1
A4 vs B4: 2.3
A5 vs B5: 2.3
Total expected results for clan A: 10.6

B2 vs A1: 2.6
B1 vs A2: 2.8
B3 vs A3: 2.1
B4 vs A4: 2.3
B5 vs A5: 2.3
Total expected results for clan B: 12.1

So in this case given these "shot in the dark" numbers, clan B would have a 1.5 star expected advantage based on those 5 attacks. That's pretty big percentage wise. I wouldn't expect a fair matchup between those 2 clans (though I've seen worse - LOL).

That's the simplistic run through. Everyone gets two attacks, it should be expanded to cover that. You could run the top best offense against the top 2 defenses, the 2nd best against 3&4 etc, but that would render the bottom half of the roster insignificant. But running everyone mirror vs. mirror doesn't capture the whole picture either, so you'd need some middle ground.

Ugh, this is a long response. You need to teach me how to table! :D

Quote:

---

Originally Posted by MazerFromAbove

Awesome use of tables. Thanks for engaging with me seriously on this - I appreciate it even if it seems that maybe we are at odds.

---

:thumbsup:
We're all trying to solve the Matchmaking problem. I don't think anyone cares about which solution is picked. Our job, as a community, is to help refine ideas and propose as many ideas as possible. The more ideas we propose, the more likely Supercell will find an idea that works with their system.

Quote:

---

Originally Posted by MazerFromAbove

I don't know how this would be debatable that they have this data. They typically have entire war replays of every attack for your last few clan wars. The algorithm I'm suggesting just needs to know offensive war weight, defensive war weight, and final star outcome. Those 3 things should be there. ...

---

I understand what you mean. What I am worried about is whether the data is easily accessible or not. For example, they may store the War Weight instead of its separate components (Offensive Weight and Defensive Weight). I'm not saying they are; I'm just saying that this is an assumption we better state. If they are stored, they will be stored as single attacks. Then, you'll have to lookup the Offensive Weight of the Attacker, and the Defensive Weight of the Defender. You need to do that for a range of attacks to form a meaningful statistic (average/mean + standard deviation). If I want to look up 12512 Offensive Weight, I will look up attacks between 12012 to 13012. It could be a lot of work or not; I don't know.

Quote:

---

Originally Posted by MazerFromAbove

Cool, I'll run this an example and try to be simple as well.

---

I didn't want you to run the War Simulation; I understood your explanation. I wanted you to give me an example of two Clans that will pass the Algorithm as a match. Basically, I have two concerns:

• The Algorithm will reject matches after a long War Simulation when it is possible to reject them using a simpler Algorithm.
• The Clans that pass the Algorithm can be determined using a simpler Algorithm. Determining exact matches is very simple and fast. If what passes your Algorithm is exact matches, then what's the point?

That aside, I think if we examine the success results of running the War Simulator, we may get insight on how to design a simpler Algorithm or insight on the Offensive advantage to Defensive advantage trade that can be done between Clans while maintaining fairness.

Quote:

---

Originally Posted by MazerFromAbove

That's the simplistic run through. Everyone gets two attacks, it should be expanded to cover that. You could run the top best offense against the top 2 defenses, the 2nd best against 3&4 etc, but that would render the bottom half of the roster meaningless. But running everyone mirror vs. mirror doesn't capture the whole picture either, so you'd need some middle ground.

---

That's what I mean when I say Supercell has to design a War Simulator. You need to make decisions on how to distribute attacks. If your proposal says the bottom half is not important Offensively, then maybe the Matchmaker should ignore the Offensive Weights of the lower half of Clans (after sorting) when matching. I think you should read my proposal.

Quote:

---

Originally Posted by MazerFromAbove

You need to teach me how to table! :D

---

If you want to add a Table, click the "Go Advanced" button next to "Post Quick Reply" button. You'll get a Tool Bar for Tables in the advanced interface.

I just thought about this. If you want the War Simulator to be more accurate, it should differentiate between the first (scout) attack and subsequent attacks on a base. This way, the average stars for scout attacks will be lower than subsequent attacks making second attacks useful to clear scouted bases.

The simulation will be done in reverse order. The lowest Offensive Weight will scout the lowest Defensive Weight and one above. If the result is not 3-stars, the next lowest Offensive Weight will clear both, or scout one higher, etc...

Quote:

---

Originally Posted by NormalCell

:thumbsup:
I didn't want you to run the War Simulation; I understood your explanation. I wanted you to give me an example of two Clans that will pass the Algorithm as a match. Basically, I have two concerns:

• The Algorithm will reject matches after a long War Simulation when it is possible to reject them using a simpler Algorithm.
• The Clans that pass the Algorithm can be determined using a simpler Algorithm. Determining exact matches is very simple and fast. If what passes your Algorithm is exact matches, then what's the point?

That aside, I think if we examine the success results of running the War Simulator, we may get insight on how to design a simpler Algorithm or insight on the Offensive advantage to Defensive advantage trade that can be done between Clans while maintaining fairness.

---

Given the expected star outputs I threw out previously, here is a 5vs5 war that would match that doesn't involve exact matches in offensive and defensive weights:

Clan A	Clan B
1 2 3 4 5 Offensive Weight TH10 TH9 TH9 TH8 TH8 Defensive Weight TH10 TH10 TH10 TH9 TH8	1 2 3 4 5 Offensive Weight TH10 TH10 TH10 TH11 TH8 Defensive Weight TH10 TH9 TH8 TH8 TH8

If I run the totals based on what I threw out before for expected offense vs. defense weight, each side is expected to get 11.3 stars based on average attacks. So start up the war and let's see who wins!

Even though one side immediately dropped x-bows and infernos before developing their offense and the other side has an engineered TH11, 10, and a 9.5, it seems like the results would be pretty even.

As someone who has done SQL and other programming, I think coming up with these averages would not be computationally intensive. I think you could throw out a lot of matchups immediately before even running these projected star totals - no reason to check a clan of TH8s vs. a clan with TH10s & 11s at the top for example.

I read your proposal too. It would work great for me as well. I think mine comes up with more matches but both are fair. And, yeah, I guess I'm assuming they have stuff in their database but with past programming experience it seems very likely they would.

Quote:

---

Originally Posted by MazerFromAbove

Clan A	Clan B
1 2 3 4 5 Offensive Weight TH10 TH9 TH9 TH8 TH8 Defensive Weight TH10 TH10 TH10 TH9 TH8	1 2 3 4 5 Offensive Weight TH10 TH10 TH10 TH11 TH8 Defensive Weight TH10 TH9 TH8 TH8 TH8

---

I'll run my own War Simulation with the hit rates shown below. Two attacks each starting with the lower bases.

Offensive Weight	Defensive Weight	Average Stars of Scout Attacks	Average Stars Subsequent Attacks
THX	THX-3 THX-2 THX-1 THX THX+1 THX+2 THX+3	3 3 3 2 1 0 0	3 3 3 3 2 1 0

Clan A:

		Attack Type	Result
A5	B5	TH8 vs. TH8 Scout	2 (B5 Scouted)
A5	B4	TH8 vs. TH8 Scout	2 (B4 Scouted)
A4	B5	TH8 vs. TH8 Subsequent	3 (B5 Cleared)
A4	B4	TH8 vs. TH8 Subsequent	3 (B4 Cleared)
A3	B3	TH9 vs. TH8 Scout	3 (B3 Cleared)
A3	B2	TH9 vs. TH9 Scout	2 (B2 Scouted)
A2	B2	TH9 vs. TH9 Subsequent	3 (B2 Cleared)
A2	B1	TH9 vs. TH10 Scout	1 (B1 Scouted)
A1	B1	TH9 vs. TH10 Subsequent	2
A1	-

Total Stars 14/15

Clan B:

		Attack Type	Result
B5	A5	TH8 vs. TH8 Scout	2 (B5 Scouted)
B5	A4	TH8 vs. TH9 Scout	1 (B4 Scouted)
B3	A5	TH10 vs. TH8 Subsequent	3 (B5 Cleared)
B3	A4	TH10 vs. TH9 Subsequent	3 (B4 Cleared)
B2	A3	TH10 vs. TH10 Scout	2 (B3 Scouted)
B2	A2	TH10 vs. TH10 Scout	2 (B2 Scouted)
B1	A3	TH10 vs. TH10 Subsequent	3 (B3 Cleared)
B1	A2	TH10 vs. TH10 Subsequent	3 (B2 Cleared)
B4	A1	TH11 vs. TH10 Scout	3 (B1 Cleared)
B4	-

Total Stars 15/15

Since my War Simulators does not give partial stars, fair matches are matches where both sides can get the same number of stars. As you can see, you can implement different War Simulators that give different results. We'll have to take people's opinions on that match to know which War Simulator they like. The point is, it is not an easy task to get this right. It is better than the current War Matchmaker for sure!