Page 2 of 3 FirstFirst 123 LastLast
Results 11 to 20 of 24

Thread: War Weight Penalty

  1. #11
    Its good to see some differing points of view. It would be nice for SC to be a little more clairvoyant regarding the war matchup algorithms

  2. #12
    Trainee TitanToad's Avatar
    Join Date
    Apr 2017
    Location
    Sailing a sea of cheese
    Posts
    12
    Quote Originally Posted by littledoctor View Post
    Edit: sorry if I've just talked past what you said. I don't agree that they need to prove it true before we can prove it false. If they prove it true, then it's impossible for us to prove it false. Likewise if we prove it false, it's impossible for them to prove it's true. Right now it is asserted, which gives us not very much proof either way.
    My fault as I was typing fast and didn't proof what I wrote. I should have said that they would need to provide "proof" that this is exists so that we could then try to disprove it.

    Quote Originally Posted by littledoctor View Post
    One way to disprove it would be to have a war with minimax-ish #1 that succeeds in drawing a below-their-TH opponent. Back in the day, 9.5s used to often draw 9s, and it wasn't suicide to have your #1 be a 9.5, but we all pretty much agree that has changed. Has anybody seen a war recently with a non-.5 engineered base as #1 that successfully drew a weaker opposing #1? Like a 9.11 drawing a 9?

    I'm not sure how this would accurately prove or disprove the OP. Let's take for example a 10 v 10 with a 9.5 as number 1 and the other 9 whatever you choose and use the same line up for 10 wars. You will get a myriad of TH combinations for your opponents. The MM is using the weight of the clan as a whole and not matching base to base plus there's all the variables that we don't know and can only guess at. (ex. offense weight, TH spread, emphasis for top of map, protocol checklist?, etc.)
    So then after 10 wars of matches, what insight will you then have gained about the weight of your 9.5? I would think not much, although you would probably be pretty frustrated with the matches you pulled
    Last edited by TitanToad; April 21st, 2017 at 04:48 PM. Reason: Edited for quote

  3. #13
    Forum Superstar OnyxDS's Avatar
    Join Date
    Sep 2014
    Location
    Brexit Island
    Posts
    3,288
    Quote Originally Posted by littledoctor View Post
    Does anybody have observations or experiences that contradict these claims?
    The stuff they are saying about penalties doesn't really match what we're seeing, but we don't run defenceless and only rarely see enemy ones.

    However a lot of what the site says about building weights is known to be false. IIRC OP's text was copied from one of 2 sites with this info, both of which give a list of "war weights" as gospel truth, and both of which are basically the spAnser numbers with a couple of minor tweaks. But these numbers are not only known to be wrong (even spAnser has admitted they aren't literal war weights) they are so wrong they're treated as a running joke on here.

    Which leads into the analysis of the penalties. The problem is that this is quoted in terms of 10k "war weight". But those amounts in the gold stores aren't really war weights - they are a monetary bonus derived from the war weights. AFAIK everyone who has analysed building weights in detail has come to the same conclusion that the actual war weights are on a roughly exponential scale: a TH11 is something like 3 times a max TH9, not 50% more as you'd think from the 70k/110k weights.

    And then a lot of their description of the penalties doesn't so much become wrong as gibberish - it doesn't even make sense in terms of the scales that the MM system seems to use.
    TH9.5 | 1900+ war stars | Trophy record as TH9: 4579 (rank #4141, #67 in UK); as a 9.5: 5282
    XX: 5200+ | FF: 1800+ | AA 180+ | GG/EE/HH: who's counting?
    Holder of the world record for the longest ever "How to fix war matching" suggestion... Part 1 | Part 2 | Part 3 | Part 4 | Part 5

  4. #14
    Senior Member
    Join Date
    Dec 2013
    Posts
    293
    Quote Originally Posted by TitanToad View Post
    I'm not sure how this would accurately prove or disprove the OP. Let's take for example a 10 v 10 with a 9.5 as number 1 and the other 9 whatever you choose and use the same line up for 10 wars. You will get a myriad of TH combinations for your opponents. The MM is using the weight of the clan as a whole and not matching base to base plus there's all the variables that we don't know and can only guess at. (ex. offense weight, TH spread, emphasis for top of map, protocol checklist?, etc.)
    So then after 10 wars of matches, what insight will you then have gained about the weight of your 9.5? I would think not much, although you would probably be pretty frustrated with the matches you pulled
    I think you're correct. It's so frustrating trying to get good evidence when each test takes 2 days and has large error bars. You need multiple experiments where you have, say, a fairly common lineup with no engineered bases, and then this one experimental minimax, and then see how the enemy lineup changes across multiple wars, without anyone else in your clan upgrading anything in the meantime. Not likely. Or, access to a bunch of clans, and then do a ton of data entry and plug it into one of the well known formulas for solving this kind of problem. One of my coleaders is a statistician and wanted no part in doing this kind of experiment.


    Quote Originally Posted by OnyxDS View Post
    However a lot of what the site says about building weights is known to be false.
    I'm kind of steelmanning[0] what they claim: the raw data file weights are the "real" matchmaking weights, and the "gold offered" weights are a false facade used to deceive people; they correlate roughly with defense strength but use a different formula than the one used to generate the defense component of war weight. This explains all of the discrepancies that we observe between gold offered and the data file values (such as xbows being lower than splash in the data files, but increasing gold offered by more). This also explains why their spreadsheets' version of "war weight" runs from 0 to 150k instead of 0 to 110k. When they say 10k and 30k they mean relative to their 150k scale, not our 110k scale.

    Their offense + defense max TH values are:
    TH7: 40k
    TH8: 62k
    TH9: 86k
    TH10: 120k
    TH11: 150k

    They certainly make claims that I find hard to believe. One might be able to justify giving mines and collectors some war weight, because they add HP to a base, but never in a thousand years would you have them weigh more than max walls. At the end of the day though that's just rounding error; I care more about high level implications like "minimax are dead."


    Quote Originally Posted by OnyxDS View Post
    The stuff they are saying about penalties doesn't really match what we're seeing
    The steelman version of what they're saying about penalties is: if the game detects you are lopsided, it more or less rounds your weight up to the average weight of your current TH level. Isn't that approximately what you are seeing? It's pretty close to what I am seeing.

    I was trying earlier to think of an example that would show that this is not what happens. A TH11 minimax mirroring a TH8 that draws a TH9 instead of a TH10 would disprove it, because this spreadsheet says that a base with TH11 offense and TH8 defense will pull a TH10 due to the penalty. Obviously it is hard to figure out which base one is actually pulling into the war.


    Quote Originally Posted by OnyxDS View Post
    AFAIK everyone who has analysed building weights in detail has come to the same conclusion that the actual war weights are on a roughly exponential scale: a TH11 is something like 3 times a max TH9, not 50% more as you'd think from the 70k/110k weights.
    This is intriguing to me. Where can I learn more about this? Does that also account for the fact that TH11s are usually on the top of the war map, and supercell says that bases on the top of the map have an extra weight multiplier?

    If you had clans composed entirely of TH11, TH9 and TH3, then if TH11 weighed 3x TH9, one clan could have one additional TH11 and the other clan would have 3 additional TH9 (and 2 fewer TH3). That is less impressive than I imagined at first.



    [0] If you're not familiar, steelmanning is the opposite of strawmanning: it's taking the opponent's argument and making it stronger, eliminating the deficiencies you can see, for the purpose of disproving the stronger version of what they're saying. I get the impression that the author of the site was handed the spreadsheet, partly understands it, and writes like a teenager who skips doing homework to write about video games on the internet.

  5. #15
    Forum Superstar OnyxDS's Avatar
    Join Date
    Sep 2014
    Location
    Brexit Island
    Posts
    3,288
    Quote Originally Posted by littledoctor View Post
    I'm kind of steelmanning[0] what they claim: the raw data file weights are the "real" matchmaking weights, and the "gold offered" weights are a false facade used to deceive people; they correlate roughly with defense strength but use a different formula than the one used to generate the defense component of war weight. This explains all of the discrepancies that we observe between gold offered and the data file values (such as xbows being lower than splash in the data files, but increasing gold offered by more).
    I didn't remember seeing the claim about the gold being a false facade, but there are a couple of these sites and I did only skim them. It's possible the gold is bogus, but it still wouldn't make sense.

    The game needs 2 values: the actual defence weight for MM purposes; and the economic bonus given to the attacker as a reward for hitting harder bases. They could be (1) the same number, or (2) the bonus could be derived from the actual weight, or (3) they could be unrelated numbers.

    We can see the bonus, but not directly the actual MM weight. Some observations: items which cause the bonus value to increase greatly, especially eagles and infernos also seem to have a profound impact on war MM. Items which have minimal impact on the bonus seem to have a minimal impact on matching. Which implies 1 or 2 probably.

    They are suggesting (3), and that the hacked numbers are the actual matching weights, and the bonus is unrelated and calculated from different numbers. But that leads to the question - why are the MM weights in the client files where SC know they will be hacked, when they are only needed on the server? The server does the MM, and the client isn't involved. The client could possibly do some of the calculations for displaying the bonus values in the UI. They are suggesting SC put in the client installer a file of numbers that are considered secret and definitely not needed on the client. Ummmm...

    I would suggest that when wars started the numbers would definitely have been related. There was no engineering or anything back in those days. There was a quote on reddit long ago of SC being surprised at people talking about "war weights" - it seemed they hadn't considered that people would look at it.
    It would have been very odd for SC to calculate 2 different weights back then. So option (1) or (2) initially. It is entirely possible that at some more recent time SC have started using 2 different weight lists to obfuscate the real MM. But it doesn't feel like it, because the MM does seem to behave according to the changes in the bonus. Also no decent list of building weights has ever appeared online that I'm aware of, so SC probably still consider it secure enough.

    What the hacked numbers do mean is an interesting question. I tried relating them to observable base weights and couldn't find any connection. Using software to look iteratively for solutions I couldn't come up with any arrangement whereby it would give a solution linking the two. And I don't think anyone has ever come up with a way of using them that actually links them to the gold weights for unusual bases. What's interesting though is that they come out kind of close if you add them up for normal bases. I have a suspicion that the weights in the files might be entirely a spoof. It might be those that are the false facade. That SC update them giving kind of plausible values for any new items in the knowledge that spAnser will then extract them from the APK for them and put them on the web as "war weights" creating a smokescreen for SC.

    They certainly make claims that I find hard to believe. One might be able to justify giving mines and collectors some war weight, because they add HP to a base, but never in a thousand years would you have them weigh more than max walls. At the end of the day though that's just rounding error; I care more about high level implications like "minimax are dead."
    Stores and collectors certainly don't have any appreciable effect on the gold value. If they do on the actual server-side matching value that leads us back to... how are we deducing that from client side files?


    The steelman version of what they're saying about penalties is: if the game detects you are lopsided, it more or less rounds your weight up to the average weight of your current TH level. Isn't that approximately what you are seeing? It's pretty close to what I am seeing.
    It doesn't really feel like it, but we don't do defenceless and don't match enough to be sure. For example, our current war the enemy has a mostly defenceless TH11 with baby heroes but max laloon. Under this system it should be getting the full penalty and counting as a TH11 (or max TH10 or whatever it works out to). But the enemy already has more infernos, more 9.5s etc than us. Both clans have a couple of 10.5s at the top - so another base counting as a max TH11 should be significant. So it would be surprising they have a bit more of everything if this did count so high according to the penalty system.
    Last edited by OnyxDS; April 21st, 2017 at 11:48 PM.
    TH9.5 | 1900+ war stars | Trophy record as TH9: 4579 (rank #4141, #67 in UK); as a 9.5: 5282
    XX: 5200+ | FF: 1800+ | AA 180+ | GG/EE/HH: who's counting?
    Holder of the world record for the longest ever "How to fix war matching" suggestion... Part 1 | Part 2 | Part 3 | Part 4 | Part 5

  6. #16
    Senior Member
    Join Date
    Dec 2013
    Posts
    293
    Quote Originally Posted by OnyxDS View Post
    I didn't remember seeing the claim about the gold being a false facade, but there are a couple of these sites and I did only skim them. It's possible the gold is bogus, but it still wouldn't make sense.

    The game needs 2 values: the actual defence weight for MM purposes; and the economic bonus given to the attacker as a reward for hitting harder bases. They could be (1) the same number, or (2) the bonus could be derived from the actual weight, or (3) they could be unrelated numbers.
    I think it was on one of the other pages on that site ("This is what supercell wants us to believe! They laugh when we think this!" or somesuch.) For a long time (years) I assumed/guessed (1) or (2), because that seems to make the most sense from a programming efficiency point of view. But it is just a guess/assumption.


    Quote Originally Posted by OnyxDS View Post
    But that leads to the question - why are the MM weights in the client files where SC know they will be hacked, when they are only needed on the server? The server does the MM, and the client isn't involved. The client could possibly do some of the calculations for displaying the bonus values in the UI. They are suggesting SC put in the client installer a file of numbers that are considered secret and definitely not needed on the client. Ummmm...
    I had the same thought. I almost wrote out a rant at one point about how putting the weight data into separate data files is dumb design. Then I thought of a plausible reason: if they included it in a header file, they would need to recompile in order to change values, and maybe they thought that they needed, or might need in the future, the ability to tweak/playtest war weight values game-side without recompiling. Maybe they had a playtest team (or wanted to not rule out the idea of someday having a playtest team) that were not developers and couldn't recompile.

    Since early on they were kind of surprised that people were thinking about manipulating war weight, maybe they did not anticipate that they needed to keep these values server-side. That seems like a horribly exploitable system (you'd think modders could construct a program that could send a "we are starting war now" message with bogus weight values) but maybe the octopus is sufficient to prevent it.


    Quote Originally Posted by OnyxDS View Post
    I have a suspicion that the weights in the files might be entirely a spoof. It might be those that are the false facade. That SC update them giving kind of plausible values for any new items in the knowledge that spAnser will then extract them from the APK for them and put them on the web as "war weights" creating a smokescreen for SC.
    That is also a plausible idea. Though, one interesting data point is that the values in the files changed in March 2016 and one later update following the same pattern announced in the update: in March, the inferno and eagle values increased, and in the later update that buffed xbows, xbow values decreased. It could still be supercell being clever, but it is also consistent with the data files actually being used for something.


    Quote Originally Posted by OnyxDS View Post
    Stores and collectors certainly don't have any appreciable effect on the gold value. If they do on the actual server-side matching value that leads us back to... how are we deducing that from client side files?
    Stores and collectors are the least plausible part of the CK2 claim, even if you have a very low opinion of supercell developers, which I do not.


    Quote Originally Posted by OnyxDS View Post
    It doesn't really feel like it, but we don't do defenceless and don't match enough to be sure. For example, our current war the enemy has a mostly defenceless TH11 with baby heroes but max laloon. Under this system it should be getting the full penalty and counting as a TH11 (or max TH10 or whatever it works out to). But the enemy already has more infernos, more 9.5s etc than us. Both clans have a couple of 10.5s at the top - so another base counting as a max TH11 should be significant. So it would be surprising they have a bit more of everything if this did count so high according to the penalty system.
    Under the CK2 system the maximum penalty is capped, so a defenseless TH11 would draw a TH9 or so. If the CK2 system is accurate, then the game is still not properly (IMO) compensating for this kind of defenseless base.

    My massively rushed TH11, 60k gold offered, with no DE troops, seems like it may have been been drawing TH10s for a while now, not TH9s like I expected. If that's true, the CK2 system explains it.

    One test of the CK2 spreadsheets would be to enter all the bases on both sides in, over multiple wars, and see how closely they say the war weights are. Ugh.

  7. #17
    Forum Superstar
    Join Date
    Apr 2016
    Posts
    3,599
    Quote Originally Posted by Long5525 View Post
    Supercells way to nerf engineered bases:

    Supercell determines the Town Hall level required to aquire the top leveled Troop, Spell, or Hero owned on the account being weighed.

    Supercell uses the Town Hall level found in the prior step to establish what the "average" Weight of that Town Hall level is. Value is "t"

    Supercell subtracts value "t" ("average" Weight of Town Hall level required for highest leveled Troop, Spell, or Hero owned) from value "w" (total overall Weight) to establish a value "x". "w-t=x"

    If there is a difference of more than 10,000 Weight in either way, Supercell can add a "Weight Penalty" of up to 29,999 Weight. "x" must be < (less than) 10,000 to avoid any "Weight Penalty."

    Supercell does not reveal how much Weight is added to any Base. Because of the added Weight to "lopsided" accounts, this has put allMiniMax accounts on life-support.

    The only time you will not be given a "Weight Penalty" by being more than 10,000 Weight above or below value "x" is:

    1.) If your total War Weight is equal to, or less than, value "t"
    2.) Your total overall Base Weight must be more than 6 Weight above the max overall Base Weight for your prior Town Hall level.
    heres the thing, someone who thinks there is any credibility to this data, go through the logical permutations and explain how any base can ever get a penalty, because iether you will be equal to or less than t, or you will be within 10000, or you will be 6 weight above previous townhall. Please just for fun and to help me out of a logic hole, please someone explain how any base will ever recieve a penalty.

    Unless average weight t is crazy low, so if the average th11 base in clash weight the weight of say a townhall 9, then I can understand how some penalties can occur, but this would still be design leave defenceless out of the penalty.

    Challenge to all forums logitians, what bases could this penalty effect for resonable interpretarions of average base weight t.
    Last edited by Vikingchief; 4 Weeks Ago at 03:20 PM.

  8. #18
    Forum Superstar OnyxDS's Avatar
    Join Date
    Sep 2014
    Location
    Brexit Island
    Posts
    3,288
    Quote Originally Posted by littledoctor View Post
    I had the same thought. I almost wrote out a rant at one point about how putting the weight data into separate data files is dumb design. Then I thought of a plausible reason: if they included it in a header file, they would need to recompile in order to change values, and maybe they thought that they needed, or might need in the future, the ability to tweak/playtest war weight values game-side without recompiling. Maybe they had a playtest team (or wanted to not rule out the idea of someday having a playtest team) that were not developers and couldn't recompile.

    Since early on they were kind of surprised that people were thinking about manipulating war weight, maybe they did not anticipate that they needed to keep these values server-side. That seems like a horribly exploitable system (you'd think modders could construct a program that could send a "we are starting war now" message with bogus weight values) but maybe the octopus is sufficient to prevent it.
    If they were actually in a header (.h) file most of the information would disappear during compilation. spAnser has put the names of the files and the raw files on reddit at times. IIRC it is something anodyne like "buildings.csv" - and it is literally just a CSV file.

    I can believe SC would use the same data files on client and server. It just makes maintenance easier. Both ends need things like DPS, HP etc, so there is some sense in keeping all the data in one place and having one master copy. Imagine the chaos if something like the HP of a building got out of step - and live attacks stopped matching the results recorded by the server for any attack because client and server had different values for it!

    But... the problem with the 2 sets of numbers theory is that there's only 1 set of weights on the client. So if SC have lazily/sensibly kept all their data in one place and thus copied the sensitive weight calculation to the client, where have the values for the gold-in-stores calculation gone? So the hypothesis must be that SC have copied the "secret" data on the client for simplicity, but at the same time removed data which is not secret for some reason. It just does not compute.
    TH9.5 | 1900+ war stars | Trophy record as TH9: 4579 (rank #4141, #67 in UK); as a 9.5: 5282
    XX: 5200+ | FF: 1800+ | AA 180+ | GG/EE/HH: who's counting?
    Holder of the world record for the longest ever "How to fix war matching" suggestion... Part 1 | Part 2 | Part 3 | Part 4 | Part 5

  9. #19
    Centennial Club
    Join Date
    Aug 2014
    Posts
    183
    If we're talking about the same spreadsheet, it seems to be almost right, but not perfectly. Reason I'm saying is that I'm a super 9 (max warden, no IT) and until last war I always had to fight EAs. Both last and this war I'm only facing ITs. I've been busy "fixing" my base and it seems like my penalization has gone away BEFORE it should have been. Not much, just a couple of levels of a wizard tower and one level of an archer tower, but it does imply that the numbers don't seem to be correct.
    Obviously 2 wars aren't any proof, but still wanted to share the data.

  10. #20
    Forum Superstar OnyxDS's Avatar
    Join Date
    Sep 2014
    Location
    Brexit Island
    Posts
    3,288
    Quote Originally Posted by PatrickV View Post
    If we're talking about the same spreadsheet, it seems to be almost right, but not perfectly. Reason I'm saying is that I'm a super 9 (max warden, no IT) and until last war I always had to fight EAs. Both last and this war I'm only facing ITs. I've been busy "fixing" my base and it seems like my penalization has gone away BEFORE it should have been. Not much, just a couple of levels of a wizard tower and one level of an archer tower, but it does imply that the numbers don't seem to be correct.
    Obviously 2 wars aren't any proof, but still wanted to share the data.
    If you're looking at a spreadsheet, you need to try it on some weird bases. Try for example this base: https://forum.supercell.com/showthre...gineering-fail I don't think I have a link to the one andrgin posted a while back, that was even better.

    Those numbers aren't just wrong, some of them are near an order of magnitude wrong. It's interesting that they're calibrated so that on a "normal" base they come out close. But if you can test them on a very weird/rushed/engineered base they are miles out: in other words the individual items are badly wrong even if they give sort of the right total for normal bases.
    TH9.5 | 1900+ war stars | Trophy record as TH9: 4579 (rank #4141, #67 in UK); as a 9.5: 5282
    XX: 5200+ | FF: 1800+ | AA 180+ | GG/EE/HH: who's counting?
    Holder of the world record for the longest ever "How to fix war matching" suggestion... Part 1 | Part 2 | Part 3 | Part 4 | Part 5

Similar Threads

  1. Engineered Base Weight Penalty!?
    By HENTA in forum General
    Replies: 20
    Last Post: July 20th, 2016, 06:12 PM
  2. Higher loot penalty. Penalty on trophy versus lower TH.
    By dpgza in forum Ideas & Feature Requests
    Replies: 6
    Last Post: December 1st, 2015, 12:24 AM
  3. Replies: 20
    Last Post: January 15th, 2015, 08:55 PM
  4. Trophy penalty added to loot penalty
    By Cheesaholic in forum Ideas & Feature Requests
    Replies: 2
    Last Post: March 14th, 2014, 04:13 AM
  5. Change TH penalty for Level penalty
    By diondu in forum Ideas & Feature Requests
    Replies: 3
    Last Post: February 5th, 2014, 10:27 PM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •