About the new method of evaluating tank performance

Hello everyone,

this post will be a bit complicated, so bear with me. Basically, a couple of Russian players came up with a new method of measuring whether a tank is overpowered or not (or even underpowered) than classic winrate with some sort of threshold. In the beginning, there was the post by Makkad. This player (LJ) referred to the recent Storm statement that T57 Heavy dropped significantly in winrate and thus doesn’t need to be nerfed. Makkad points out that it practically took a year of being OP to drop to reasonable levels:

3945592

A year to discover whether a tank is OP or not based on balance alone is of course too long a period. What is needed is an indicator, that would show the tank OP factor directly after two or three weeeks. In the following part of the post, Makkad proves the fact that such an indicator can indeed exist.

Based on this premise, Makkad makes another (possibly obviousy) statement, that the tank winrate depends on the ratio of good and bad players playing it. The quality of players playing that respective tank from its introduction shifts gradually: in the first week after its introduction, only extremely dedicated players will have the tier 10 tank, because they buy it for accumulated free XP. Here is the amount of T57′s played in time: in first week it’s only like 340 players with them, while by now there are over 130 thousand of T57′s on RU server

3982459

Obviously the first ones to unlock the vehicle are very good players from top clans, but over time, as the vehicle becomes more popular, its winrate spread (when comparing it with players) will correspond the general skill spread of the playerbase, which looks like this for RU server:

3939448

Thus it can be concluded that the ratio between good/average/bad players (furthermore referred to as GAB) does influence the vehicle winrate. That’s why the author suggests to use the corellation between the player winrate group and their actual winrate of the vehicle. For example, you take all the players with 52 percent winrate and check their average winrate on the vehicle you are trying to assess. You do that for several of the groups and the result looks like this:

3994747

The horizontal axis displays the player group winrate, the vertical one the average winrate of that respective group on the assessed vehicle. The dotted line represents the average player winrate growth, while the solid line represents the vehicle. In this case, the vehicle would need a buff, because its winrate grows slower than the winrate of the player groups. This method allows to estimate much earlier, whether the vehicle needs a buff/nerf or not.

3984507

On the picture above, the situation is not so simple. For bad players, the vehicle is hard to play and for good players, the vehicle is overpowered. In this case, the method could even help the developers pick, what to do with the tank. In order to make it easier for bad players to play it, you need to use straightforward buffs (such as armor or alpha increase) and to compensate it by limiting the maximum tank potential (which good players use), for example by nerfing DPM, camo and viewrange (those are the values a bad player will gain little benefit from).

To summarize: by looking at not only at winrate, but also at winrate/player quality corellation, you can potentially reach more accurate results. In the comments to this post, Storm stated that he actually passed the method for evaluation to his statistic specialists, as he considers it to be a reasonable idea.

The method is applied by another LJ user (Ivanerr) in his post as follows (based on public data from wot-news I assume):

T57 Heavy

Not imbalanced in general – noobs can’t play it, but it is somewhat OP in unicum hands. Ivanerr states that the point, where the two lines cross can be called the “OP point” and the lower it is, the more OP the vehicle is. Ivanerr also advises that the fact the OP point is 52,5 for T57 means that anyone with less winrate should avoid this vehicle, as that person will suck with it.

57huy

AMX40

The shape of the T57 graph (“noobs can’t, unicums can pwn”) also means one more thing: that this vehicle can potentially greatly influence the battle. AMX40 seems to be the other extreme case:

amx40

The fact that noobs pwn with it but unicums don’t means one thing: this vehicle can do very little to influence the battle outcome. It’s a noob friendly vehicle statistically, but unicums will have a hard time influencing the battle with it.

Foch

Using this method it shows that the Foch is a classic case of an overpowered tank. It has very low OP point, which means even noobs can use it to produce better results than what they usually have.

fosh

Maus

Maus is an example of a tank, that is very lackluster. As you can see from the graph, it doesn’t really have an OP point, no matter how skilled you are.

maus

WT E-100

While current winrate results according to developers do not warrant a nerf, using this method the vehicle is completely overpowered

trajer

M48 Patton

Using this method, Patton seems like a noob friendly vehicle, but it does require VERY high skills to truly pwn.

patton

KV-1S

Using this method, KV-1S is shown to be completely overpowered. It has a very low OP point, which means even total noobs can pwn with it.

kvas

Rheinmetall Borsig WT

Very, very imbalanced vehicle, it is almost completely overpowered.

borsig

T110E3

Slightly tiny little bit OP

t110e3

T95

Nearly ideally balanced

t95

T110E4

Not OP, needs a skilled player

t110e4

Jpz E-100

Lackluster, just like the Maus

pze100

Object 140

A bit underpowered, needs a skilled player

140

FV4202

Mostly underpowered except for the best players using it

fv4202

E100

Pretty fine, not imbalanced

e100

IS-4

Quite underpowered

is4

FV215b (heavy)

Pretty fine, not imbalanced

fv215b

FV215b (TD)

Again, not overpowered

fv215b183

T110E5

Not overpowered, you need to be a skilled player to use it

t110e5

AMX50B

A little bit underpowered

amx50b

Batchat

Has big potential for skilled players, but in noob hands it’s bad. Not overpowered.

batchat

Object 263

Pretty ok, not overpowered

263

E50M

Very well balanced, neither OP nor UP

e50m

113 and 121 (Chinese)

Both almost ideally balanced

113

121

And – IS-7, that is COMPLETELY underpowered

is72

Regarding the IS-7, Ivanerr has a theory about the influence of players unlocking IS-7 first while they are still noobs, but overall, the stats really don’t look so good.

Well, that’s it. What do you think?

198 thoughts on “About the new method of evaluating tank performance

  1. Practically it takes pro player less than week of gaming with the tank to estimate if it’s op or not… Same goes with maps, most of them are horribly imbalanced (for example Mines), but WG has “statistics” based on data provided by in majority brainless tomatos who could fail 15 vs 1 and treat it seriously for some unknow reason just in this one case, when in any other situation Serb basicly bashes idiots.

    • the biggest errors with the graphs – the data itself.

      the game has changed so the data used in these graphs (well, for the old tanks at least) can be considered outdated or irrelevant to the current tank balance.

      • I think that other data should be taken into consideration, like the other tanks playing in any of those battles and their owners winratio, dmg ratio, kill/death ratio and so on… It’s just to reductive to trust only two datasets when you can work with many more variables.

  2. I wish they also displayed the stats of the T-62A. Would have been interesting to compare it with the obj 140 – especially since for some reasons unicums prefer that tank yet the data is weird on it

  3. I think this method has some flaws.

    1. The IS-7 is not UP, so I’m going to have to go with the theory mentioned.
    2. This would also then, apply to Maus, but that doesn’t seem quite true either.
    3. It allows devs to nerf tanks for good players while buffing them for noobs at the same time. This is bad. The vehicles need to become harder to play, but with a higher skill ceiling. Noobs don’t particularly seem to care whether they win or not, because if they did they would try to stop sucking. Meanwhile, let the good players have battles that are actually worthy of them, rather than the mostly luck-based bullshit that dominates the game right now. This should allow worthy players to offset the noobteams a bit better. What this new method does is completely crush any such vehicles that currently exist, like the T57H, T110E5, or the russian meds for example. If my fears come true there will be very few battles actually worth playing as they will be decided by the matchmaker even more than they are now, dumping perfectly good players in a barrel of rotten tomatoes.

      • You obviously haven’t read the article if you don’t know the difference between OP and hard to play with a high skill ceiling. Please read it properly the next time before you comment.

        • And you need to learn some manners before making a response like that. He asked a question. How about responding with an explanation rather than an insult? That would go a long way to making others as informed as you imply you are.

          • You can’t expect that sort of thing on the internet, I’m afraid. Still, it’s much better around here since anonymous posting was disabled.

          • You are suggesting the poster asking you a question doesn’t want to learn… that’s quite a reach. Acting like a pretentious dick in the process doesn’t make it better either.

      • Well, as for the IS-7 theory: that can apply to any tank, comparing my current global WR with WR on some ever-present german tanks (Panther, PzIV) that I’ve just ground throught during my early WoT days (without revisiting them later) would mean that the tank utterly sucks when in fact it was me who sucked :)

        So, for a better results the formula would have to take into account this fact somehow to reflect “early noob days” vs “I started to grind this tree after 2 years of playing and now having over 60 % WR” and so on :)

        • Pretty simple fix, just use a snap shot of the last week/month of play. That means every tanks results are going to reflect the players skill, since they won’t dramatically get better over a month.

    • ” It allows devs to nerf tanks for good players while buffing them for noobs at the same time. ”

      It just gives more accurate info to consider buffs or nerfs, its doesnt MAKE devs nerf tanks. Doing changes while seeing the whole picture precisely is always better than swinging the nerfbat blindly. Having graphs like this, devs can decide if tank needs a “hard” nerf/buff to help noobs and newbs or “soft” to make them more influential for better players.

      Its actualy a great tool, because it could make nerfing/buffing much better and less annoying, like lets imagine a tank that has armor and relies on armor and has good soft stats to back that up, so its decent for low-to-mid players but good and above excell in it. Now, instead of going trying to hit it with a nerfbat like a pinata with eyes closed and hit armor, which would make this tank useless for everyone, devs would know they should nerf some soft stats. Having said that, I believe there should be noob friendly tanks and hard-but-rewarding tanks, but I wouldnt mind if they were designed like this on purpose and not by sheer luck.

    • 1. and 2. Do You have any proof or this is just your opinion?
      3. And this is bad? How it is? It is just opposite. They do not need to buff or nerf tank gor everybody. They can be more precise. That is really great.
      And your post completely makes no sense. If all the tanks would be properly balanced, then sklls would decide. That’s the point. Now, when teams have similar skills, but one drives T110E5, the other IS-7, who is gonna to win? Matchmaking decides.
      Chinese T10 are close to be perfect, but we may see, that there is lots of work for balancers.

      • The problem is how they inevitably will make tanks worse for good players and better for bad players because that’s how you keep retards playing your game and paying you money.

        • dude if you can only play good in one of those acclaimed OP tanks then I’m afraid you’re not really a skilled player you might think yourself are. unless you can make any tank work regardless, and in that case you wouldn’t care how they balance a tank.

    • Optimizing all of the vehicles for unicum level players may be great for the highly skilled players. It would allow the unicums to pwn everyone else, but it would be a bad business move. By not pandering to the noob majority, WG would see a significant drop in revenue if the game was so hard that the noobs quit in frustration at too fast of a pace. A high churn rate would not give WG enough time to properly fleece the noobs.

  4. People trying to be smart when they just aren’t. IS-7 not able to influence battle? Completely underpowered. Then lets start using FV215b in CW… NOPE.

    One more thing. You just can’t measure a player’s current skill with his overall WR. It’s just wrong. You might measure his progress a bit but not his current skill. More like 60d WR would be much more accurate.

    • Why IS-7 in CW? Hmm, because it is reasonably fast, has very bouncy armor, and the gun isn’t a problem when people use premium ammo. In the FV, gold ammo is not necessary.

      • It’s slower than the T110E5 in most terrain. Its armor is rubbish against gold rounds except for its own gold rounds. Its gun is crap, even with premium rounds. 303 pen APCR is worse than tier 10 TD standard AP rounds except for waifutragger.

        If your clan is using IS-7 in CW, I have some bad news for your clan.

      • Gun isn’t a problem?
        Worst fire control of tier X(except maybe 183) + second worst dpm of tier X is not problem?

    • Does anyone even use IS-7 in CW? Haven’t seen any (except in scrub setups) for very long time on RU cluster.

    • “Then lets start using FV215b in CW… NOPE.”

      CW is a very different game from pub matches, and so you can’t compare the two. Some balances of tank characteristics that work well in pub matches are utter failures in CW. (This is why you don’t see scout tanks in CW.)

      So: “Use success in pub matches to determine how well a tank will do in clan wars….NOPE.”

  5. Foch … that thing sure pisses me off it so needs a nerf you wouldnt believe it. I dunno about the WT e-100 and the boring. The fact hat the WT’s have completeley OP guns is du to the fact that they have compared to their respective tier no armour so they need to compensate for that. If you catch a WT off guard it is pretty much dead. If you catch a Faoch off gurad it may be able to outrun you, to outmanouver you or hide behind its insane frontal amour with tiny weaks spots… the foch is the only vehicle fear to play against in randoms… because even the biggest idiot has the chance to blow you up and leave with no scratches at all.

  6. Did he have any graphs to display the effect of nerfs over time on the results? If the results he posted are over all-time history of the vehicle it isn’t going to tell us much other than “historical OPness”

  7. KV-1s is OP we already know this…
    and to all KV-1s drivers sorry but
    T6 HT with MT mobility and 400 dmg is too much.

    • > T6 HT with MT mobility and 400 dmg is too much.
      And MT armor. The thing’s literally an upgunned medium they call a Heavy Tank.

  8. i think at some tank the unicums come out only because they use 100% gold ammo.
    the 113 for example. with the gun of WZ-120 is almost garbage without gold.

    • from all pub games I been in its mainly noobs using gold or HE, unicorns knows how to aim and know every tanks weaknes and strength of the tanks. Unicorns mainly use gold in Clan wars.

      • After using XVM and the hitlog mod, all the unicums on random battles are either spamming gold rounds or with 3 man platoon.

      • Not true. With that new mod which show, what is the type of ammo You have been hit… well, when I see player >green effi I can be 100% sure he will use gold ammo. A LOT of them use it. Fuck**g statters…

    • Problem with Wz-120′s gun is not the AP ammo.

      Problem lies in track and gun dispersion(hilariously bad comapred to T-54; which is not that great already).

  9. Great stuff but we probably never see wargaming devs using such method since this method is far better than their own . Wargaming devs dont want to use other people ideas (specially if they’re better) since this would proove they were wrong all this time, and dont know their own stuff too good ;)

      • And u think this means anything… right?. U failed to realize one simple fact – storm often says that he is passing something to developers. Problem is that this mean nothing since developers just dont want to use player ideas, we had plenty of examples in past.

        • Actually, no. It’s not often he says that. The last time he said that was about 4-6 months ago when someone proposed an alternate method of matchmaking, which is understandably not as likely to get adopted as it would require significant recoding.

    • You’re right. For example WG never implemted OTM into the vanilla UI, and they never copied the reload mods to make it part of the game either. Yup, totally close-minded communists.

      • It all had taken them abut half a year to implement… Or more…
        So we may see changes following this statistics in a year or so?

        • It’s neither front- nor backend modification. Just another datamining tool that can be thrown together in a single afternoon. I think we’ll see the impact it makes faster than you can say Llanfairpwllgwyngyllgogerychwyrndrobwllllantysiliogogogoch.

  10. I have the IS-7 at a 58% winrate over 200+ battles. My overall winrate is 50% at over 10k battles. Therefore, I am overperforming, am i not?

  11. Shocking that they haven’t done anything like this before. Take, for example, the Obj 268 or T-62A- hordes of RU pubbies bring their winrates down, but both of them scale very well with skill (great firepower for their classes, great camo, great flexibility) and the former is potentially pretty noob friendly as well (sit in a spot and click at people for 40% of their HP, armour provides lots of random bounces). When you look at almost any particular player in the blue/purple range, they will likely overperform.

  12. I don’t understand the low performance of IS-7. According to these statistics, it’s worse than Maus, which is peculiar since IS-7 is a CW staple while Maus is almost never used. How can an UP tank be demanded on CW?

    • CW and random are two very different environments. Random is where the bulk of the statistics comes from, CW is only a few hundred games every day.

    • Whats so hard to understand about it?

      IS7 is the noob magnet on the RU-Server. Every fucking moron first grinds it, something similar to the Siemka E100s on EU.

      • So this method in its current form is also flawed. Perhaps they should use data for past few months only instead of overall.

        • As long as you work with WR only to evaluate player skill, you will get BS results anyway. Furthermore you need to disregard certain players all together from your evaluation:

          - less than 5K battles: newbies have no clue how the game works. they polute your data sample
          - everyone with less than 42% WR. These are ever bots (some tanks are very popular in these circles) or people with some serious problems in other areas

          Also it would be good to finally work with single player unplatooned results. Platoon greatly influences the potential results one can achieve and often works as a multiplier, blowing things out of proportion, negative and positive.

          • No. Anyone with knowledge of statistical analysis will tell you you cannot exclude selective data of this type. Why? Because these people actually influence games. You can’t setup a random battle that insures you won’t get a bot or a tomato player. That directly reflects on the tank when in the hands of a poor or automated player. These elements are real and they happen.

            You might as well exclude all 60% players as well or players with 50k battles. That’s called cherry picking the data. And it’s totally wrong.

            However I do agree on using solely win rate as being a poor way to evaluate. A player controlled “nerf” or “buff” by using gold ammo, special consumables, high skill crews, etc. can also affect the specific performance of tanks. For example, the tier 3 Pz S35 premium totally sucks without gold ammo, can’t pen shit. Add gold ammo and suddenly it’s OP as it’s armor is great and you can rip up anything tier 3 or lower, can even pen a Matilda with some skill.

            The point is before you start excluding data you had better understand that data as you will more than likely skew the results of your analysis. Try defending that in a peer review or in a doctoral thesis, and 9 times out of 10 you find yourself incaoBle of justifying it.

            • As long as certain tanks are the favorites of bots u have to exclude them, because that has nothing to do with gameplay.

              And pls explain to me what newbies, who dont understand anything about this game bring to the table, when it comes to balancing.

              Both types of players just produce statistical noise. Balancing including these 2 types of player doesnt improve the game play experience of anyone, neither the for the 44% potato head nor for unicums.

          • I dont get all you people saying ” players with X k amounts of games dont have a clue”, you people maybe sucked until 5k games, but I know ppl who do well with 1-3k battles, myself being one of those, first account, no statspadding, very good stats
            The only real thing lacking is far more experience in certain situations then maybe players with 10k games
            And then again, there are people with 25k games who dont have clue number one

            • You can debate if 3K or 6K battles is the border, when it comes to newbies. What you cant debate is, that newbies in general dont have a clue about game mechanics. Just click random battles after you installed XVM with a good custon config, and you will see what I mean. Are there exception? Sure! But they are a tiny minority.

              • I think that “tiny” is much bigger then you think. You also have to consider that some of the starting players will actually be grinding crews from 50% to 100%. And that at a given point everyone can afford to drop credits for a 75% crew with every tank and or unlock all of the modules before they start to play it.

                Now that does have an influence on player stats and would be hard to argue it does not.

    • Totally agree on that. Hard to believe that WG doesnt work with an balancing mechanism similar to this one.

      Is there any (semi) official (not trolling) reaction to this idea? I cant remember that i have read anything in FTR.

      @SS: AWESOME Job. my daily routine: start PC, read FTR and clan forums, do something else ;)

  13. you should remeber the CW influence on CW important tanks (IS7 being one of the tanks of risky tactics which can fail more often)

    If you want to balance for random, you should just take random-stats.

    • If the tactics involving IS-7′s would truly fail more often than not then people wouldn’t be using them anymore. The IS-7 is a very good mainline tank when you need a type of shock-and-awe, rather than surigical precision like the T110E5 or the do-or-die attitude of batchat/medium rushes. Additionally the CW influence on these tanks (in terms of number of battles) is neglectible. My CW games represent currently 1.44% of all of my stats.

  14. It’s about time someone used a sensible method like this. Indeed the only tanks that strike me as surprising are the WTF E-100 and the IS-7 in this case, and to a lesser extend the Patton. The IS-7 i suppose can be explained? Personally i have 65.59% on it (61.3% account-wide average) having played 683 games with it, and taking it out very regularly.

    The WT E-100, while i don’t own it yet, is already heavily held back by the secondary things, particularly camo, as was marked here as being a thing to be nerfed to make unicums less effective with them. My experience with vehilces like the WT E-100 is that while they are perfectly capable of kicking ass, when it comes down to the wire it can be difficult to carry games in them because of their general vulnerability.

    M48A1, well, it was good when it had it’s insane DPM in the testserver, these days it’s too vulnerable against HE and HESH penetrations, while providing a pretty large and fluffy target in general. T57 Heavy is worth of a mention in the sense that while i have an INSANE amount of average DPB with it (3.8k on 249 games) the winrate of this vehicle doesn’t seem to follow suit. I have a similar problem with the AMX 50 B, but with less DPB and an atrocious winrate.

    On the whole, i do believe this is the correct method for judging tanks by. Obviously some tanks will always be UP and some will always be OP, depending on situation. The rest is up to the maps. If you were to take this same graph for the Maus, but take purely Himmelsdorf and Ensk battles i’m sure you would get a VERY different picture… but seeing as most maps are atrocious for the Maus, the general results here aren’t surprising.

  15. The method seems better than the one WG uses now.As far as the graphs shown in this post,I find them extremely accurate,WT E-100 is,in my opinion,a completely OP,unhistorical,abomination.

    Also I would be quite curious what are these graphs saying about KV-2,ISU-152,SU-122-44,Obj.268.I don`t play tier 10 battles so idk how is Obj. 268 performing but just looking at it`s stats that tank is beast: great camo,alpha,DPM,accuracy,mobility,penetration,good armor,good view range.This tank has it all.

      • That’s nice. I love mine. In the hand of capable player, 268 can be a monstrous. But noob? They die after 2 minutes battle

      • The only fact i see is that those toamtoes ran in with this tank, die and loose. but if you know what ur doing this tanks is still ridiculous. This DPM along with that trollarmor, accu, mobility and all that shit, is just too much for experienced hands. (prooved by the graph)
        would say to nerv the DPM on T110E3 level. but it truely seems to be perfectly balanced for the main playerbase which somewhat suprises me.

        Would love to see some Leo stats. Probably even more extreme than B-chat ones in the lower part.

  16. Smart yet simple Gj
    Is7 ins’t a bad tank, just has a bad gun for the present Tier X RANDOM games.

  17. Mr. Noobmeter uses that method since ages for his OP-rating. Took them long enough to “copy” it.

    Furthermore using WR to evaluate skil in that regard isnt to clever, because its easy to manipulate with platoons/company.

      • I have my doubts about that. WGs book keeping is lousy at best. For example the removed T50-2 has the same tank-ID as the new M-25 t6 scout. So when you look someone up on noobmeter.com or wotlabs.net, you cant be sure if you are seeing his T50-2 (t5) stats or M-25 stats … .

        I dont know if they are able to filter tanks/players stats internal by performance in solo random, platooned, company and cw on large scales (server wide). But that would really be necessary to balance tanks properly. And seeing how balance goes since CB I really have my doubts … KV1-S anyone?!

        Just the fact that some russian players 2.5 years after game release, brought this OP-Rating idea to WGs attention and they werent aware of it, proves the point for me that they have no fucking clue how to balance. This gets even worse by their balancing philosophy to produce a balanced tank for the 46-50% WR crowd. That is one of the main reasons why they never got it right in the big picture.

        • I would really want to know why some random player is bringing up something that WG’s (double digit) team of statisticians could not come up with ? In all honesty, I don’t think this is the case and perhaps there are graphs already used like this or potential are not used for specific reasons. Obviously there are anomalies, and they perhaps be one of the reasons.

          Just as a point of reference, WG does not release A LOT of internal data and statistics that may be even more useful in their balancing effort. Although I do believe there is reasoning to keep tanks at X level for certain reasons before tuning tanks back. They are about business first, and making huge revenue numbers.

  18. It’s a decent mechanism to evaluate the tanks performance vis-a-vis other tanks of the same tier. It is of course for WG to decide how to balance tanks that appear too OP/UP without destroying the tank and game balance.

  19. I’m wondering about one thing,what about the influence of the map on a player and even on vehicle ? It can be ignored or it should have been at statistics too ?

    For me;i manage quite well on city maps usually,however map such as Redshire or Komarin i feel not so confident especially if i have a vehicle which hit by tier bias by matchmaker (2 lower tier than usual or such)

    Also i find that information valuable and useful,will look into that,thanks for sharing that SS :)

    • That doesnt play a role, becuase it evens out ever thousands of battles. Law of large numbers.

  20. I only played tier X tanks on CW with my noob clan, so I had a 42% win rate with IS7 because IS7 was used like folder, but when I started not to play CW anymore and to use tier X in random I put the win rate of my IS7 in 52%. in random IS7 is so bouncing tanks (there is not so many gold shells, mainwhile in CW is 100% gold shells). so this method should avoid any “””””team battle mode”””” ( CW, TC, TB ) and only take care of random battles.

  21. SS, good and comprehensive article. Basically, you are right that this method is better to decide whether a tank is OP nor UP. One thing need to be adjusted, though: this method is fine, but the time-frame of 2-3 weeks to decide, whether an tank is OP or not is way too short. The final decision can be taken only after 2-3 month. The reasons are simple:

    1. For the first two weeks, less “bad” players reach this tank. It takes about 2-3 month until the majority of “red barons” reach the new tank, adjusting the global stats of this tank.

    2. The more players play a specific tank, the more numerical data can be used to give WG reliable results. Less players = less accurate data, more players = more accurate data.

    So, for tanks like Foch or KV-1S, the above stats are more reliable than the stats for the Rheinmetall Borsig WT, since the KV-1S has WAY more players/battles than the Rheinmetall Borsig. The tendency is nevertheless obvious.

    • You also forget that all the potato heads first have to learn, how to deal with certain tanks. And that takes a long time.

    • 3. I also want to add that special events play a major role. Triple exp weekends, Christmas specials, and similar events utterly change the playing field as greater or fewer players jump onto the bandwagon of certain tanks, huge masses of bad players jump on to grind through their tanks quicker, and so on, so forth. In a game that depends on teams working together, this will not only have a disproportionate effect on individual win ratios, but on the teams that the matchmaker puts together.

  22. You would have to be careful using this method in the early days of a tank, not because of those driving it, but because of those fighting it. It generally takes a while for players to work out how to kill a certain tank, after which its winrate will drop even more from the combined effect of more bad players driving it and the rest of the playerbase working out how to kill it.

  23. The FV215b have a very interesting dip on the curve, why does 59%ers do worse in it than 58%ers? And why does it turn back up at even higher base win rates?

  24. wait i’m a 53% win rate player with 8k game and i have 55% WR in 500 battles in my maus and 54% in the jpe100 i think i’m strange right? o.o

    • No, you are just annother brick in the wall. You might be on top of it, but you are still a brick.

  25. Not sure if ppl mentioned this already – the IS-7 theory (that later player performance is clouding the statistics as they tend to use it early in their wot career) is based on the idea that player skill is not static in time. A simple solution to this is to slice player performance into time periods when overall skill is relatively static – say, perhaps 2 week chunks? Feel free to select a better time chunk, as long as within that chunk:
    1) you compare player winrate with IS-7 to general player win rate
    2) the players you’re studying used a healthy selection of tanks, not just IS-7; the IS-7 winrate must have an opportunity in that time period to be compared to other tanks
    3) there are sufficient matches played for a meaningful comparison
    4) for whatever time chunk chosen, you divide the entire data set into a collection of chunks (ie 1 year of data = 26 2 week time chunks to study per player studied)

    You could probably apply the above to any tank… Let me know what you think, a little good input might allow this to be refined into something more useful.

  26. Ok I found this very interesting, but the chart is a little wrong in some cases (IS-7 extremely UP, it needs a buff, but a little), what the Devs has to is is the following:

    First, you use this method to nerf the OP tanks, after 1 month, you can check the datas of the UP tanks and you can analyze if the tank winrate go up a little, cause I am really sure that the super UP tanks are highly influented by the OP tanks, if you put a T-57 vs IS-7, guess who win, so nerf the T-57 and OP tanks first, then you check the UP winrate tanks, and you will have a better vision if the tank needs a buff and how much.

  27. I wouldn’t deduce to much information from this pictures.
    For example the downward bend occuring from 58% to 59% WR Players for the WT-E100 and from 59% to 60% for the FV215b(heavy) is an clear indication of few data points.

    Unless they give us confidence intervalls along with these data points ( see for example http://en.wikipedia.org/wiki/Boxplot) those are just pretty drawings, not more.

  28. They TOTALLY hit the nail on the head. I agree with each statement about each tank here.

    I REALLY, REALLY, REALLY, hope the Devs use this method to balance.

    • The point isn’t to make every tank an all-rounder; this obviously doesn’t make sense, as a Maus can’t stealth scout and a Chaffee can’t tank, etc. The point is that you shouldn’t be able to get an advantage from selecting tanks that, on average, inflate any player’s win rate.

      Of course there are types of tanks you will be more effective at, which you will get a higher WR than your average, but there shouldn’t be any tanks that almost every player gets a higher than their average WR. Same for less effectiveness; I would wager that most vehicles in the game are ‘UP’ and won’t represent your skills proportionately (partially thanks to the SPG nerf and the camo system).

  29. This one is completely hilarious: http://ivanerr.ru/lt/images/liz/m3lee.png How can good players completely fail to take advantage of the obscene gun? (analogous to ~12000 DPM and ~600 alpha at tier 10)

    I’m only ~56% WR average but ~63% M3 Lee.

    I guess the explanation is that high WR players quickly go past the Lee and don’t stick around to unleash its seal clubbing potential (like a sad guy such as me :-D).

    Regarding Obj.140 vs T-62A: http://ivanerr.ru/lt/images/liz/140.png vs http://ivanerr.ru/lt/images/liz/t62a.png

    What a surprise, they perform almost identically. I would have thought 140 higher top speed and gun depression vs T-62A higher rotation speed and tough turret would cause some difference. Should check back after a while when players ‘get used to’ the 140. Obj.430 will also be available to compare.

  30. Great analysis. So the spread or difference in height between the dotted and solid line shows how much playing a particular tank improves(or makes worse) a particular player at different skill levels measured by WR.

  31. Objectively, based on my experience and knowledge:
    T57 – the reload speed makes it OP, other stats are pretty mediocre
    Foch – combined burst capability, mobility and frontal angle make it overperforming, but it can die rather quickly
    Maus – the high survivability and armor are compensated by sluggish handling and before all low DPM so you can’t really carry games except on certain maps
    WT E-100 – totally broken, autoloading speed and damage per clip are ridiculous
    M48 – well balanced
    KV-1S – it’s very noob friendly and it’s easy to pwn with
    RhmB WT – ISU with low pen and a turret (borderline OP)
    T110E3 – I think it’s balanced but I somehow don’t see it often
    T95 – slowwwwwwwwwwwww but balanced
    T110E4 – balanced
    Jagdzer E-100 – it’s fine but it isn’t able to carry games often
    Obj140 – IMO even a bit better than T-62A, which has unnecessarily high DPM
    FV4202 – balanced, lackluster top speed, but I find it a great pub stomper
    E-100 – balanced, but demands much skill
    IS-4 – balanced
    FV215b – fine but tricky to play
    FV 215 183 – again tricky to play but gold ammo makes it OP
    T110E5 – well balanced
    AMX 50B – tricky but fine
    Bat 25 – same
    Obj 263 – very similar to the regular Foch – great mobility, great frontal armor, good gun but dies rather easily
    E-50M – well balanced
    113 – requires skill but rewarding to play, I consider it superior to IS-7
    121 – above average performing, DAT GUN
    IS-7 – gun stats are horrid but the platform is excellent

    • m48 totally balanced nope it really needs a buff i just cant get my 3000average dmg on it,my 2900 on it just arent enough for a tier10 medium:P

    • one more add
      the IS-7 has horrible soft stats
      no depresion
      poor terain resistance which means that 59 kph that is showed in stats can only be achieved by going downhills and ofc accelaration and mobility is slugy as hell despite the 1050 tones egine power
      121 gun is craaaazy it is the is4 gun with better stats and with more RoF while the 121 is a medi while is4 is a heavy

  32. Well, yes this method seems good and should be used but…

    isn’t it big fun to farm tanks and eventually find one that is OP and to have much fun with it against noob or common players? e.g. as VK2801 was, or Foch155, or Rhm-Borsig, or arties were?
    If they nerf every tank within a few weeks and almost all tanks are equal, where is the fun to farm tanks? We could all use one tank only if they are all totally balanced. Real fun only comes up if you are a good player in an (for them) op tank and kick others asses. :))
    I only farmed Foch of FV because they were said to be op. Otherwise i would’nt even have researched the tree…which is not the goal of wg i guess.
    Nerfs kill fun, fun kills time spent in wot, kills money earned by wg.

  33. I would like to see more SPG charts made other then the SU-26. And for SU-26 data it should be data from only the last few months. IE post nerf hammer otherwise it is skewed.

  34. This method looks pretty good, but still needs some adjustments.
    WR in a particular tank, especially new ones, is not some number you should blindly use. I believe a certain number of battle threshold should be used., like 150 battles. Take the Borsig for example: it relies too much on camo and view range to stay alive, something noobs are not able to fully abuse, according to the own article, so I wonder if the graphic discrepancy happened because of low battle count for players who recently got one.
    Besides, I don’t think overall WR is too reliable, as it doesn’t account for the player improvement over time. 60 days WR, though still not perfect, is more reliable.

    • Yes, the winrate calculation in the graphs uses the same logic.
      It assumes that you can fully understand a tank at about 100 battles.
      If you have less than 100 battles your influence in total winrate is lower proportionally.

  35. Interesting method. I mostly agree with it except for the fact that I consider WT E-100 and Borsig crap and I think IS-7 is the best tier 10 heavy. Other than that, seems good to me.

  36. Interesting. Using winrate lines is simple but it hides alot of data. Using a scatterplot where individual wr for each tank are plotted as a spot on the graph. The density of the spots shows who is using the tank and the range of variance.

    • i think you never played it and when you are fighting against it you are aiming on the upper front plate or on the turret right?

      • And I guess you never bounced 5 timees in a row @268 avg pen on LFP of IS-7?
        ~100-200m range and almost flat terrain.

        But I still wonder how those unics can hit with this gun @500m+ fullspeed ahead.
        Mine is not allowed to hit almost at all at that distance even when standing fully aimed.

        • Blind luck, for the most part. Aiming low whilst driving does help, though – do not ask me why.

  37. Hi!
    I’m the maker of the pictures.
    I must warn everybody not to take the graphs too seriously, especially for the older tanks, especially for IS-7.
    According to my data, IS-7 is the first choice for a player who thinks about buying a tier 10 heavy tank.
    http://ivanerr.ru/lt/images/liz/stat.png

    The graph for IS-7 is heavily influenced by said players. No way it is SO MUCH underpowered (but it really is a little).

    On the other hand, graphs are very accurate for recently introduced tanks such as Hellcat, Borsig and so on which are obviously overpowered.

    For older tanks the graphs show “historical” performance which could be inaccurate at present (see KV-220 beta and conventional KV-220).

    The new methodology is still making it’s first steps, but it is a giant leap forward comparing with older WG’s “average winrate” bollocks.

    I know that the creator of wot-news website who has much more data to analyze, who has the opportunity to get very recent data, is considering using this method to analyze tanks in the near future.

    I really think it is a revolution in tank balancing, but it has pitfalls that should be avoided.

    • Really nice work. I think the IS7 just logically follows from the fact the KV-1s is so easy to get (some) results with and/or the usual clan requirement to have one for company battles. Still, I find it highly noteworth that your system seems to get it so wrong with the IS7. I am not a true maths guy, but I do understand and use statistics in my line of ‘work’ as a student of history. From what I gather, it should not really matter whether or not a large amount of noobs plays the IS7 – they get in their own category.

      • > it should not really matter whether or not a large amount of noobs plays the IS7

        Actually experience means a lot in WoT.
        I’ve made two graphs using data from players who have only one tier 10 heavy tank.
        I assume that players who have only one tier 10 tank are noobs (they don’t have much experience).

        For E-100:
        http://ivanerr.ru/lt/images/liz/e100ed.png

        As you can see the graph is totally different to the graph using data from all players:
        http://ivanerr.ru/lt/images/liz/e100.png

        So, in WoT we have both skill (reaction time, spatial thinking etc.) and experience which affect your winrate.

        • E-100, IMHO, is an example of a poorly balanced tank. It can go from complete garbage to uber OP. Back in time when premium rounds weren’t available for credits, having an E-100 in the team was one step closer to defeat. Several E-100s in the team would lead to defeat 9 out of 10 times. After sprems, if a player is willing to spend the credits on premium rounds, the tank can be utterly overpowered. In my experience, sub-50% won’t use sprems, no matter what. I am tired of seeing E-100 drivers shooting HE rounds. On other hand, good players completely dominate with the same tank. Of course the latter knows how to fully use the armor, but the premium rounds makes a huge difference. And I am not sure this difference can be measured. So, imo, E-100 is a too OP in the right hands and completely useless in the wrong hands, thanks to premium rounds, not the tank itself being balanced.

          • Sorry google, but that is utter BS. If you know what you are doing – weak spots, positioning etc. – the E-100 got plenty of penetration without gold. I am so sick and tired of hearing this fairytale of the E-100 being only useable with gold rounds – 10 minutes on gamemodels3d.com should tell an intelligent player everything he needs to know to pen and cripple every damn vehicle in the game with roughly 240 pen and reasonable accuracy.

            @ ivanerr: So, despite WR, the number of battles played influences this stat? I assume that might be caused by stat padding? BTW, the IS7 is my only T10 and yet I am easily one of the top players almost every game. OK, I have got a garage full of T8 tanks and a boatload of experience playing those T8 tanks in T10 games. :P

            At any rate, I would suggest you use the WR of battles played in tanks of the same tier. (maybe -2/+2 tiers) If somebody has a 55% WR, but got most of his battles played in T1, his results in a T8 tanks might be far different from his overall WR.

            • I guess you can’t read. And I know by a fact you don’t own an E-100. Plus, the E-100 doesn’t have 240 pen and certainly doesn’t have reasonable accuracy. Leave your comments to the only tier X you own.

              • @ google: I guess that is where the test server rounds have come in. At any rate, you will see E-100 tanks do quite well without gold rounds in random games. You also see people fail hard in them, but that goes for all tanks. Even the WT E-100, which is the most OP tank in T10 as far as I am concerned. By the way, the E-100 got 235 pen. That would qualify as roughly 240, would it not? 0.4 accuracy is what I call reasonable. Up to .42 works for sniping weak spots up to roughly 300 meters, although you may miss occasionally. Your move.

                • Yeah , right. 0.4m accuracy happens to be the same accuracy, on paper, of your IS-7, which is well known to be sniper in the game, am I not right? But wait, there is more. The E-100 also packs a high velocity gun, compatible with sniper tanks. Want more? Aiming time. The best there is. Let’s not forget its superb camo. People won’t even now you are aiming at them.
                  Only someone who doesn’t own the tank to say such things. Reviewing a tank by watching random players on pub matches. The best review there is.
                  Anyway. You still can’t read. Try reading again and, more importantly, understand.

  38. Really impressive. Just one remark, the IS7 is a beast in the right hands. I find it quite odd that the stats do not reflect this, whereas the other tanks really nicely conform to what I see in game.

  39. I have T25/2 in my garage, and I am really curious about this tank, could someone please show me its stats? thanks :)

    also if someone would provide me stats for aufkl. panther and/or pz 38 nA I would really appreciate it.

  40. Ok a nice way to rvaluate tanks but have they considered the use of gold rounds? I mean gold in any players hands can be deveststing , espicially talking about the german heavies

    • To be fair, the devs only read the Russian forum and I get the impression WG EU and WG Ru do not communicate too well. :P

      • it seems hellcat is very OP, the threshold is 47 winrate. I hope this be enough to make devs do something about these tank next year.. after all 2014 will be the year of balance!

        • The graphs need to be taken with a huge bucket of salt. Some people grind the same tank for 500+ battles. Some grind from a 50% crew to 100% because they cant afford a 20k per crew member “buy in”. Some just slap 100% crew members from other tanks, pay gold from the start, etc. OR, they play it 25-50 times to gett a good WR on it and then skip it with Free XP. Or skip it outright.

          So many unknowns. Also the data becomes flawed, like others noted, if it is more then 60-90 days old. Mostly this is because tanks are edited every 90+ days with the patches. Which means stats then change.

          • even before the graph hellcat is known to be OP. anyone who actually plays the game can see it, if they want to see it that is!

          • Nemo, a large enough sample would take care of the variations you are talking about, as the same variations apply to all tanks almost equally (at least within their tier). That is to say that just as many people start a Hellcat with 50%/75%/100% crews as JPIV, so if Hellcat is way more OP than JPIV, something should be done to make them more competitive.

  41. Very interesting analysis. However, this raises a question.

    I am a 52% winrate player in Korean server. However, most of my tier 9 tanks have winrate of 55~60%. If you take a look at my IS-8, it has 59% winrate. Does this mean IS-8 is overpowered? absolutely not. On the other hand, my T54E1 winrate is 46%. It is because I played T54E1 when my overall winrate was about 49% at 5k games played, and I played IS-8 when my overall winrate is 51% at 8000k played. another word, It is because my recently improved skills are not captured in my overall winrate (my recent 1000 battles winrate is about 55%).

    Of course you can argue that when there is numerous cases of players, the statistics will show the true result. However, it is also true that player’s skill generally gets better with more games played. Therefore, at what point you played the vehicle will show some distortion to above analysis.

    however, this analysis still fantastic and much much better than just capturing plain tank winrate. my applaud to the russian players who analyzed this, and I hope my comment would make their analysis more complete when they go further with this. thanks.

  42. Pingback: Statistics of all the tanks using the “new method” | For The Record

  43. Excellent method, I used that for a long time looking at my own tanks and seeing which ones I would keep for new CW type.

    GJ!!!

  44. Would it be possible to include this kind of graph into the Tank Inspector? This kind of info would help people to select tanks that would fit their skill level.

    Im curious about a graph for T-150.

  45. Pingback: Wot-News launches the “new method” application | For The Record