Most of the matches were 3-0. All you can conclude from those was that one player was better than the other. Out of all of the matches, there were only about five that were 3/2 splits. That makes such a small sample size that it isn't possible to make any balance statements.
If anything, the large number of 3-0 matches says that balance for 1v1 is close enough that the better player usually wins.
While at face value this seems like an excellent explanation of what's happening; it's actually very unlikely, especially for the OST/Sov matchup. In fact, provided sufficient randomness in faction selection (it was a coin flip), these stats (although preliminary) essentially exclude this from being a possibility.
If a large portion (or even all, for the sake of argument) games were 3-0, that means that those games must either be 2:1 axis or 2:1 allies. Provided that faction choice was random, that means most players (who progressed) should've played each side an equal number of times - first two games as allies, then moving to the next round with two games as axis, and so on. Given this, that means that each round should assign a 66% win rate for the winning side, and 33% to the losing side.
What's important to note, is that again, those 66/33 win rates should be handed out roughly evenly to each side. 50% of the time the 'significantly better' player should've been playing axis twice, and 50% of the time they should've been allies twice. As a result, on average, the allied:axis win rate should be roughly equal, and any deviation from this would be
very unlikely - since deviation means that a 'less skilled' player won against a 'more skilled' one.
What this leaves is essentially a range of likely outcomes. If the match-ups weren't 'close', and there was always a 'much better' player, that 66/33 split should cancel itself out given enough games (there were 32 in brackets, more if we include round-robin), leaving a 50-50 win:loss ratio. However, even given the
worst possible case, where the better player
always played allies twice, the win:loss ratio could still only be at best 66:33%. However, for the case of Ost
ov, we're nearly 5% outside of that possibility.
The sample size for those OH vs SU numbers is 14 games (10 W / 4 L). Syphon sample is 18 games (12 W / 6 L).
Then you have to take into account WHO was playing each faction. For example: Luvnest was the one who played the most soviets (11 games) compared to the following players Isildur/Happycat/Jove at 4 each. Luvnest only drop a game against Von Ivan till the finals against Noggano.
Rather than panicking on small sample sizes comparisons i would rather at least take a look at overall performance and even better how most games played out.
For example: i think it's more clear how to tone down SU than doing so with USF or toning up OH.
I did mention that it was a small sample size, and would need more analysis - and also that we
shouldn't buff OST. However, even with a small sample size, the stats are still extremely strange. A +/- 5% or even 10% result isn't too surprising, +/- 21.4% is a lot. As I explained above, even in an unlikely scenario of the better player always playing Sov in the Sov/OST matchup, the stats still don't make sense.
For fun, if you calculate the standard deviation for all the match ups, you get almost exactly 11%, with a margin of error around 3.48%. That means all of the matchups are pretty unremarkable, falling within one standard deviation; in fact even the margin of error calculations gives us only a meager 0.95 - 1.3sigma; i.e. very unremarkable. OST/UKF is a bit larger at 2.87sigma, so it's sticking out a fair bit more, but still nothing crazy. Then there's OST/Sov, at nearly 2 standard deviations from normal, and at
6.14 sigma. If CoH2 balance were a scientific field, I could say with around 90% confidence that UKF was UP, but with
OVER 99.99966% confidence that Sov is OP.