Login

russian armor

coh2stats "lie" or can be misinterpreted

11 Nov 2021, 21:25 PM
#1
avatar of NorthWeapon
Donator 11

Posts: 588

I want to point out that the numbers you find in coh2stats are just numbers and don't necessarily prove anything.

  • If UKF have a lower winrate than say Soviets, it might not be that UKF is a weaker faction. It might so happen to be that there is a disproprotionate amount of players or new players playing one faction over the other.
  • If either Allies or Axis has the lower winrate in general, it is not neccessarily true that one is better than the other, it may be that one side is more popular than the other and matchmaking pairs unequal players against each other
  • And especially this: just because "Mechanized is the 9th picked USF commander" does not mean mechanized is the worst commander. It means that it is the least picked, or it is the least popular. And people can not choose a commander for a variety of reasons from what is cooler, rarity, cost, and so on.


So I think it is a red flag when people use these stats to prove some argument (especially balance arguments).

Practical example: Axis has higher winrate than Allies on 4v4 but Inka Tournament showed Allies dominating axis. Why? Could it because Allies only gets good at super high level play? Who knows.
11 Nov 2021, 21:30 PM
#2
avatar of Rosbone
Senior Mapmaker Badge

Posts: 1524 | Subs: 2

I will take PageP stats every day of the week over some tards anecdotal spewings in the shout box. But I get what you are saying.

Long live PageP!!!
11 Nov 2021, 21:34 PM
#3
avatar of NorthWeapon
Donator 11

Posts: 588

jump backJump back to quoted post11 Nov 2021, 21:30 PMRosbone
I will take PageP stats every day of the week over some tards anecdotal spewings in the shout box. But I get what you are saying.

Long live PageP!!!


You mean Sturmtigers aren't OP? cuz a few game ago some dude wiped my units and I lost.
11 Nov 2021, 21:40 PM
#4
avatar of Rosbone
Senior Mapmaker Badge

Posts: 1524 | Subs: 2

Practical example: Axis has higher winrate than Allies on 4v4 but Inka Tournament showed Allies dominating axis. Why? Could it because Allies only gets good at super high level play? Who knows.

The last tournament was dominated by allies because pathfinders and maxim spam were untouchable. Since the paths give sight for the maxims nothing could attack them. This setup can only happen in a tournament setting.

Axis are easier to tech, have early sturm/kubel pressure, and tend to do better in rando games. Unless you get all OKW vs all Brits then its a severe beating for axis.

None of this is fact (are there such things?). Just observations and listening to players much better than myself. So I tend to agree with PagePs stats in general.
11 Nov 2021, 21:45 PM
#5
avatar of Rosbone
Senior Mapmaker Badge

Posts: 1524 | Subs: 2

You mean Sturmtigers aren't OP? cuz a few game ago some dude wiped my units and I lost.

Yes this :D

I find most games where I get my ass handed to me it is because I didnt tech something I should have. But many people just get enraged, cant see their error, and spew into the shoutbox.

You will never see me talk about balance in the shoutbox unless its 4v4 okw vs Brits. Thats just a seal clubbing waiting to happen. But its a sacrifice we need for asymmetry as you talked about in your other post.

Pip
11 Nov 2021, 22:12 PM
#6
avatar of Pip

Posts: 1580

I want to point out that the numbers you find in coh2stats are just numbers and don't necessarily prove anything.

  • If UKF have a lower winrate than say Soviets, it might not be that UKF is a weaker faction. It might so happen to be that there is a disproprotionate amount of players or new players playing one faction over the other.
  • If either Allies or Axis has the lower winrate in general, it is not neccessarily true that one is better than the other, it may be that one side is more popular than the other and matchmaking pairs unequal players against each other
  • And especially this: just because "Mechanized is the 9th picked USF commander" does not mean mechanized is the worst commander. It means that it is the least picked, or it is the least popular. And people can not choose a commander for a variety of reasons from what is cooler, rarity, cost, and so on.


So I think it is a red flag when people use these stats to prove some argument (especially balance arguments).

Practical example: Axis has higher winrate than Allies on 4v4 but Inka Tournament showed Allies dominating axis. Why? Could it because Allies only gets good at super high level play? Who knows.


There are three types of lie.

  • Lies
  • Damn Lies
  • Statistics

MMX
12 Nov 2021, 04:38 AM
#7
avatar of MMX

Posts: 947 | Subs: 1

I think the most important thing about these stats is to take them with a grain of salt. Some people like to cite w/l ratios of 45:55 for certain maps/factions and use this as an indicator of blatant imbalance - yet more often than not the sample size of games considered is in the low hundreds. Even a 100 perfectly even 50:50 coinflips repeated multiple times will already have a standard deviation of 5%, so making any balance assertions based on such data alone means treading on thin ice. That of course is not meant to devaluate pageP's awesome contibutions in any possible way. Just to serve as a reminder that context matters when interpreting any form of statistical data.
12 Nov 2021, 09:18 AM
#8
avatar of Sander93

Posts: 3164 | Subs: 7

jump backJump back to quoted post12 Nov 2021, 04:38 AMMMX
I think the most important thing about these stats is to take them with a grain of salt. Some people like to cite w/l ratios of 45:55 for certain maps/factions and use this as an indicator of blatant imbalance - yet more often than not the sample size of games considered is in the low hundreds..


This is the most important red flag imo, the top 200 stats for teamgames only consist of a couple of dozen/hundred games depending on the time period which is nowhere near enough for a reliable sample size, which needs to be (licks finger) in the couple of thousands at least. The fact that both 3v3 and 4v4 top 200 stats highly deviate from the averages in the all-games stats, but in opposite ways (something like 70-30% for Allies in 3v3 and 30-70% for Axis in 4v4) should be a pretty clear indicator that these particular stats are highly unreliable.
12 Nov 2021, 11:43 AM
#9
avatar of suora

Posts: 96

There's also a large skill gap between a rank 200 Wehrmacht player and a rank 200 UKF/USF player, so I wonder about the validity of winrates for data like that. Intuitively I would think this can make the UKF/USF winrates look worse than they really are.
12 Nov 2021, 11:56 AM
#10
avatar of Rosbone
Senior Mapmaker Badge

Posts: 1524 | Subs: 2

Looking at one months win ratios for each faction makes it look like:
1) Allies are stronger than axis in a 1v1 situation.
2) Axis are stronger than allies in a 2v2 situation.

All 1v1 maps are favoring allies. No 2v2 situations ever occur.

All 2v2 maps are very even. Allies win early game when they are stronger, then axis come on hard at the end when fighting over mid VP and are in a 2v2 situation. This is the most balanced game mode currently.

3v3 maps are a mixed bag but favor axis a little. Smart 3v3 play is to push fuels then secure one fuel and two VPs if that fails. So it starts as 1v1 and transitions to a 2v2 situation. Since the maps are larger axis also tend to out range allies a little and axis do better on those maps. Large maps are also dominated by forward retreats and there could be more OKW players playing then USF for example.

All 4v4 maps favor axis. Because every game is a 2v2 situation at the start, middle, and end. Axis randos synergize better than allied randos.

To me, one month of games begin to tell a tale. No hard truths but trends become visible. There is almost always something gained from data. But I agree with all here that it is not the end all be all.

SiphonX and PageP should be praised for their efforts for adding this great insight.
13 Nov 2021, 12:31 PM
#11
avatar of pagep

Posts: 101

I wouldn't use word lie but of course misinterpreted is right word to use. The issue is that we don't have all the games which are played so we can't have finite answers. The site just provides the stats and info how they are obtain and it's up to you to interpret the data.

If we focus on win-rate someone could argue that top 200 analysis doesn't have enough game (~12k games). But it you compare the top200 win-rate to all games (~60k games) the trends are still similar (one faction winning more than other) just the % is little bit different. And I would consider 60k games enough to give some results.

Let's take a look at the span of the patches:
Current Patch: https://coh2stats.com/stats?range=range&statsSource=all&type=1v1&race=wermacht&fromTimeStamp=1631145600&toTimeStamp=1636761600
Commander update: https://coh2stats.com/stats?range=range&statsSource=all&type=1v1&race=wermacht&fromTimeStamp=1623801600&toTimeStamp=1631059200
Winter balance: https://coh2stats.com/stats?range=range&statsSource=all&type=1v1&race=wermacht&fromTimeStamp=1615161600&toTimeStamp=1623715200



I think you can clearly see the changes the patches did. The long-term trends are visible.

Here is the JSON file containing the stats data we have for every day https://storage.googleapis.com/coh2-stats-public-storage/daily-1636805161.json if anyone wans to do their visualization.
13 Nov 2021, 23:18 PM
#12
avatar of Hannibal
Senior Moderator Badge

Posts: 2593 | Subs: 2

I mean, as usual, you should not switch your brain off when reading data.

There's a lot of reasons why the data can look as it does. I am happy though that the site exists, even despite the fact that there will always be some idiots that cannot interpret data properly or only accept their own line of argumentation. But that has nothing to do with the site itself.
The site aggregates data and has some basic visualization. It doesn't mislead, although it could show the uncertainty in the data a little bit better. But this often requires some deeper knowledge in statistics.

However, the data will only get stronger the longer the site exists.
1 user is browsing this thread: 1 guest

Ladders Top 10

  • #
    Steam Alias
    W
    L
    %
    Streak
Data provided by Relic Relic Entertainment

Replay highlight

Board Info

118 users are online: 2 members and 116 guests
serg_codmod, Protos Angelus
11 posts in the last 24h
116 posts in the last week
616 posts in the last month
Registered members: 28426
Welcome our newest member, reagantaylor88
Most online: 1221 users on 25 Feb 2020, 12:03 PM