But why would you want a polluted data base when you can filter games which are way closer ranked wise? I mean, it's not like the sample size is that small.
6K games vs total 15K. Compared to what happens in teamgames (2v2 been barely relevant with 1.5K games from 23K).
DISCLAIMER: I absolutely do not agree with Vipper's suggestion of including lower-skill matches in the data, low-skill players do not provide an accurate view of balance, regardless of how many of them there are. They don't show if a tool works correctly, merely that they don't know how to use it.
Personally I think the issue with this sort of data is it doesn't quite paint the most useful picture. Raw winrates are less useful for balance discussions than the faction's winrate vs each opposing faction shown separately, particularly with the disparity of total games for any given faction. Is there a breakdown that shows the results of specific matchups? Brits being included at all (Despite their abysmal performance/limited number of games) makes the rest of the data rather misleading.
(Similarly: while a 50% winrate is the desirable outcome for each faction; A 50% winrate still doesn't necessarily mean that some balancing work isnt required, especially if that winrate is only achieved through crutching on certain units/strategies. SOV "worked" for a while, but only because the T70 carried the faction, which still warranted changes even if that resulted in a "balanced" winrate.)
With regards to 3v3/4v4 vs 1v1/2v2 winrates: I think this is further evidence that these modes need to be balanced seperately. In teamgames it's far easier to crutch on certain absurd strategies that can only work due to critical mass factors, and that some mechanics don't interact correctly in larger modes due to the comparatively static nature of maps.
I think some of that could actually be solved through mapping. I think that 3v3/4v4 maps could probably benefit from A: A reduction in the amount of resources available, B: an improvement in the layout of sectors, in order to make cutoffs/flanks/etc a little more realistic in these modes, and finally: An increase in the size of maps, while implementing more "walling" to compartmentalise these maps a little better (while avoiding the laney layout present in many popular team maps, this "lane" system causes an unfortunate number of issues)
The rest of it is the over-representation of artillery (both rocket and conventional), and in the case of axis: the strength of vehicles like the Panther when massed. There's a phenomena in games similar to DOTA called a "deathball", where one team rolls around the map as a blob obliterating everything they come across, this is similar to what ends up happening in CoH2 teamgames. The difference is that in DOTA this doesn't work vs competent opponents, because the benefits from doing so are heavily outweighed by what you lose in map control/other objectives. In CoH, having your army obliterated by the deathball puts you in a position in which you can't project yourself back onto the map, and the deathballing team ends up profiting in the long run because of this.
There's also the fact that vehicles like the Elefant, Jagdtiger, and ISU-152 are far too difficult to flank in these modes, but I think that can be solved by changes to the maps.
I find it interesting that 1v1 and 2v2 games have somewhat comparable winrates, whereas 3v3 and 4v4 have a MASSIVE disparity between each other, incidentally. Perhaps that's partially due to sample sizes, but it's interesting that it seems to imply that Allies are slightly favoured in 3v3, but axis are far, FAR ahead in 4v4. Is there something special about the fourth player?