allies tank spam problem

They don't stop playing = they spend half their playing time in the lobby waiting for a match.

The number of game played is the same for Allied and Axis.
Each game provides 1 winner and 1 loser.

Top Axis players can win all their game, that wouldn't change much because they play less games in average. Top Allied players (which are usually the same players) can lose all their games vs Top Axis players, the superior number of games available vs lower skilled players they going to win will make it up for the W/L Ratio, which will be in favor for axis, but not that much.

Now take in condition the human factor, top players avoiding themselves on ranking matches, the fact that they aren't numerous per definition and the high probabily they are unmatched in their respective time slot etc... Let's also add that if a faction is underpowered or harder to play, mostly only good players are going to play it on regular bases.

So now I'm going back to my initial statement that W/L ratio between faction isn't an indicator of balance on itself and even more at high level.

Now I get what you mean. Yes, I agree that the dampening effect sounds logical. Still, even though those top players with OP factions will find games less frequently, they will still find one eventually, even if they only play half of the games (which would mean 30 min queue times). If you look at the WR for e.g. the top 5% of each faction, you will still see that effect, because the best players of the UP faction will achieve ~50% WR, the best of the OP faction will achieve better.

The overall faction WR is a whole different issue though, because it WILL hide the fact that some factions might be strong with good micro, but difficult to learn, which will lead to top and bottom cancel each other out.
This means that a 50% WR is not a guarantee for a balanced faction. In addition, this also means that a strongly deviating WR from 50% for all games indicates a larger issue with the faction.

Which is where I think we're agreeing mostly. That's also what when I wrote "The question is rather if the smaller deviation in WR is true or statistical fluctuation".

Has anyone ever looked at ELO distribution? Median and some normalized average ELO should actually help in that case. In CoH2 we did not have access to that info.

Has anyone ever looked at ELO distribution? Median and some normalized average ELO should actually help in that case. In CoH2 we did not have access to that info.

You really would need some advanced statistics math here, given how factions are not played equally, or facing equal match-ups.

We do have match-up specific stats as well, but elo distribution alone without normalizing faction pick difference won't do any good.

Can't really throw all team games into a single bag.

4v4 always was and always will be hmg spamming artyfest.

3v3 also tends to develop a frontline with arty after early game.

2v2 is much more diverse.

2v2 is what i play the most and i still feel like its not super diverse. Ppl use a lot of different units, but the deciding factor always seems to be those late early/mid game vehicles and their timing. Or just spamming them

THe build up can be diverse but what wins the game isnt, imo

2s for sure more diverse than 3s and 4s, but thats not saying much

You really would need some advanced statistics math here, given how factions are not played equally, or facing equal match-ups.

We do have match-up specific stats as well, but elo distribution alone without normalizing faction pick difference won't do any good.

In my mind I was mostly thinking about 1v1 where at least the combinatory issue does not exist, although it technically shouldn't matter.
I am not sure how complicated the analysis would actually need to be. Maybe just looking at ELO distribution within teams should already give some info. Weak factions/team combinations would have a consistently lower ELO across the board. In case something is "easy to get into but hard to master", ELO range might be compressed more than other factions. This would not resolve factors like e.g. "double USF is good against double Wehr, but bad against double DAK". This could maybe be done by filtering for these special game setups (with all the caveats of Relic not directly introducing a ladder for that), although I yet don't know if it works out and how exactly it would look like.