Careful I don't brush it off. The differences are tiny. Such tiny differences aren't sort of "big enough" to influence a balance discussion much (I'm talking about all games now). If they were bigger that is another story. Too many factors will usually influence such differences, especially when the matchmaking system is at work. The graph is also a bit misleading as the 2% difference looks really big there.
I might have copy pasted too much - of course we are speaking about 1v1. There are more arguments there, don't brush them off
I know how to read a graph, mate. But I am still talking about the graph that plots game rank vs win rate. This one contains by far the most information.
It actually does. There are cherry picked players and sort of the best possible match for each other there is. Of course it is not enough. But this is all we have got - no random drops due to a phone call, etc. No silly matchmaking due to waiting long for a player or no ping problems, no cheating, to name just a few factors.
No it does not. These cherry picked players are the same that you find in the top 5% of the graphs and therefore the region that I am talking about. Tournaments are additionally held on only a couple of maps which further scews the outcome in case one of them is slightly unbalanced. Additionally, the "form" of the player is way more important since they are a snapshot of only 1-2 days. If you're playing Asia/Oceania but the tournament runs until late evening in European time, the cherry picked players from Asia will have a disadvantage because they will play their games in the middle of the night. Did not sleep well that day, had a cold, a stressful day at work, argument with your family? Well suddenly the player's favourite two factions will get an additional loss. The game size is small. The only thing that tournaments have going for them is that the skill gap is - assumingly - smaller, but that's all. Most, if not all, other factors still apply (they still run via Relic's servers, there can still be a bugsplat. And quite famously, in game bugs as it happened in Luvnest vs Asiamint. All these weight actually MORE than in the 15k game sample), but with a smaller sample size and that actual RNG has more weight. You said you know your statistics, so give an actual reason as to why all this does not apply instead of sentiments and statements. The "silly matchmaking" is a fair point, but as I said: At the edged of the graph this happens to all players, therefore all factions have higher winrates at the top 10% of players. There is no good reason to assume that this happens more often to one specific faction than another.
If you know statistics you should also know that - for the most part - the tournament matches are a subsample of the data that SiphonX visualized, but slightly scewed due to the tournament environment.
They do I'm afraid. I understand You had a discussion. The conclusion and arguments seem to be wrong. The ends of the graph is actually the worst part to use for comparison. Silly matchmaking on both ends will give you silly and random results and maybe even more importantly, it will be based on a rather small sample. Drops, experiments, good players just playing for fun against weaker ones - too many variables basically. And still the differences are too tiny to be significant. And still the number of games will be different for Soviets and UKF, for example. And my main critique is that the bigger the sample the more realistic result - closer to 50% in this case. Just wrong approach I'm afraid. Completely wrong.
Alright, then give a reasoning as to why all this is wrong. Just saying "it's wrong" does not make it wrong. You also focus on the end of the graph - but don't say anything to my points regarding the top players.
Those random elements have been addressed already as above. I agree that the bigger sample size is better, that's why SiphonX's data is better. You specifically claim that the more games were played, the closer the win rate is to 50%. Let's have a look: For OST this is true, SOV has the most games from Allied factions yet differs the most, UKF and USF differ the least yet have the least games played.
Are these differences significant? You say no. The honest answer is: We don't know for sure. But it's a lot of games, and the differences in the top player department is between 5-10% with 15k games played. The drops and other random mishaps that you mention are likely a very minor fraction of all these games.
U're making too many assumptions here, and not really counting all variables/reasons to win/lose. Actually it is impossible to enumerate them tbh. For example, trying to suggest that that low end will show You how quickly a player will learn a faction is just expecting far too much from a simple graph.
If it came across as if these is the be all end all conclusion, then this was not intended. But if a faction performs the worst with both high end low skill players, it's a decent hint that this faction might be weaker than the other factions.
I'm afraid U can't see the faction strength there tbh - too many caveats. Explained by the author of the whole spreadsheet very well. Too tiny differences, too many variables, too few players for some factions with many more for other factions, etc influencing stuff. Ends of the graph, trust me, are the worst point for any analysis.
This data is not the perfect set to prove faction strength, but at some point you have to see that your initial argument is based on your personal "game experience and my tourney and cast observations" stating that "Soviets are much more forgiving and generally easier to play with than OKW". You said I can't conclude from a simple graph how easy it is to play a faction, yet you do it from what? The 100-200 games a CoH2 fan can play in a month maximum (unless you have a full month of free time) + the 50 you watched? How do your observations suddenly beat data from 15k games? You said you know your statistics, but this is fully ignoring good scientific practice. You even agreed that the matchmaking will even things out for the majority of all players. So how does your personal game experience then translate into actual balance knowledge if you additionally have to filter out that all your matches were decided by the matchmaking algorithm to give you a chance of roughly 50% (I am assuming at this point that you are not in the top 5-10% of players)?
And again: How can you say that tourney stats could be a decent basis with those few games, split across 5 different factions and a small player and map pool, but then say you agree with the spreadsheets author that the games and players in his (much, much larger) data are possibly not enough to draw conclusions?
Nothing of this adds up.