While I do appreciation of the work which was put to establish new coh2stats, I wish it was never made, because it always brings this number masturbation ppl will use to backup their statements.
When I build the site I didn't know what I will find from the data. Frankly I just wanted to know what is most played from the perspective of maps, commanders, bulletins. And I thought it's a shame that other games do have sites like this and we have nothing.
And as someone said I think the site data could be taken into consideration but is not 100% proof because it's not perfect. Because of the limited API we don't have all the data and mostly in team games there is so many factors which affects the gameplay, the AT, the team compositions and much more.
And maybe on a lower ranks the brits do have more games / better win-rate. Who knows
The only one who could do full analysis is Relic themselves with the direct access to all the data. And It shouldn't be that hard for some data science engineer. Based on LinkedIn Relic do have several data science people. But I would guess they are focused on a different games right now
I am not really convinced that top 200 analysis is good approach in 3v3,4v4 because the amount of matches which passes the specs is very little. I don't think it is enough games to produce reliable results.
OK I now about result types for matches:
0 - lose
1 - win
However I just found a match which has result type 4. That means that there should be type 2,3,4. I don't really know what those mean. I am thinking what to do with those matches.
Solutions:
1 - do not count those matches - anything from them
2 - do not count those wins/losses but count the map selection, intel bulletins, commanders and count those matches as a separate entity to wins/losses.
-------------------
Right now I am thinking solution 1 is better one. We(I) don't know what those matches mean so in case we don't include them - we can't hurt the statistics. We might be losing insight maybe into which faction is crashing most But I think that's not our concern. Relic should know about the crashes themselves.
If you add up all the wins and losses, you don't get 0. This means for some reason very few wins or losses are not counted for either Axis or Allies.
Hm you are right. I am trying to figure out what does it mean.
If you count everything together. You get 1302. 1302 / 6 (6 players) and you get 217. Which is the number of the matches correctly counted. But I don't understand why the wins/losses does not add up.
I am not sure if someone drops from the match / how it looks like in the match results - that's only thing which in my mind could be causing some issues.
I've generated it for all the past days too. However with that - march still has only < 300 games in 3v3,4v4 so I am not sure how accurate those stats are.
But we can see that axis is not dominant in the team games any more. They still have slightly higher win-rate. Will be interesting to see how it changes with more games.
Question:
Did you consinder the fact, that a 2at premade team ofen search for 4vs4 and playing it? Does your system get those players then or not?
That match would not be counted. If they are 2at team only their 2v2 games against top 200 plyaers would be counted.
I still think it should be based on a % rather than the top 200. The 1v1 Wehr ladder has nearly 3150 players, while the 1v1 Brit ladder is 1400. This means you are comparing the top 6% of the Wehr ladder against the top 14% of the British ladder.
wehr - 3371
brit - 1602
I think we would end up with even less matches. With the current calculation for the top 200, there is only ~12 brit games per day, which is way to low. Even week stat for this would be just 84 games which is not enough... I can see the stats to "stabilize" around 200 games at least.
But let's take a look, if we would take top 10% of players:
That's 160 positions of brits VS 330 positions of wehr.
Hm is there a difference in a skill? I can't really tell. For example if player ranked 160 would play against 330 ?
I am personally around rank 1k - level 10 . When I play against people 1k+ , even just few ranks worse, or level worse. I totally stomp them or usually have no problem winning. But when I play against people from top ~500 I can feel the difference immediately, sometimes someone plays really weird but usually I can feel that those players are on a completely different level them me. Would that be a case in 160 vs 330 player? Or is that more or less on a same level already?
Btw the more I think about it. It makes sense to have it based on the % in the ladder. However it would require way more work. I don't have full ladders, I would need to do additional development of lot of things.
Also another question if we applied different amount of people from each faction - it would affect the amount of games and the statistics what race is more played would be biased but on the other hand we already know this from the ladders. Hm...???
But let's say we want to have it as %, what the number should be?
1v1 - brit - 1450
1v1 - wer - 3158
1v1 - wgen - 2711
1v1 - usf - 1887
1v1 - soviet - 3158
For 2v2
2v2 - brit - 2440
2v2 - wer - 4321
2v2 - wgen - 3852
2v2 - usf - 2504
2v2 - soviet - 3757
team2 - allies - 8097
team2 - axis - 8148
Obviously we would need different values for arranged and random teams, because if we applied the same % it would be completely different amount of people which would be counted in.
I have the "raw" data for the top 200 players only.
How it works, players are split into the modes 1v1,2v2,3v3,4v4 - the faction doesn't play a role.
The match is counted ONLY if ALL players from the match are in the TOP 200 for that mode (for example 1v1, 2v2 - team players are added too).
What does it mean that faction doesn't play a role?
FYI This is the ladder data:
Amount of unique players in modes: 1v1: 472
Amount of unique players in modes: 2v2: 891
Amount of unique players in modes: 3v3: 1266
Amount of unique players in modes: 4v4: 1309
How come there is so many players in 4v4? Even though we have just 5(races)*200(positions) the teams are added too. Each team adds 4 players.
Anyway here is report for the last ~5 days ( 17k matches before filtering only top 200 players).
You can see that the win-rate is much closer to the 50% ==> Aka the game "looks balanced" . Except for Brits, it looks like they suck anyway . But I think it's sill not enough data. When I had data for just 1 day, in 3v3,4v4, it's often just 20 games and with that - the data was completely unbalanced 0.7 winrate was not uncommon. I think we will need more and more games + this statistics is not enough to have some strong conclusions. For example with the team games we should try to track the team composition which might play a big role in a winrate (for example team of 4 brits vs 2w,2wg \\ team of 1b,1u,2s, vs 2w,2wg \\ there is several combinations) + maps, I mean this is extremely complicated thing with huge amount of variables.
-----------------------------
Please let me know what you think. If you see some significant flaw in my calculations or something looks off - let me know. Thank you
(EDIT: Btw you are stating in the info section that there were 100k games per day. The site you linked shows about 110-120k games per week, if I understand it correctly. So only about 17k games per day.)
You are right, I did not noticed that the chart does aggregation. However if you count on that page all the modes/ranks I counted 244k games/week , that's 34k games.
But thanks for pointing that out!
Btw my next plans is to really filter only top 200 games - all players needs to be in that range. Will see how many games will stay. But not sure when I will have time to look into it - hopefully next weekend.
Hi guys, I Have a question. Do you think the site should be accessible from mobile devices? Because there is a lot of people (about 30% - which is hundreds) who are opening the site from mobile devices.
It's extremely hard to make those fck charts work on smaller screen...
So the question is, should we try to do something with that. Or it's more like fuck them, unless you have at least 1080p display you won't see shit on the site.
But I think that's not our concern. Relic should know about the crashes themselves. 
Relic Entertainment
Scoped 
AceHiro 
