GCS2 Citadel faction win rate

I don't think UKF is too hard to fix provided the decision makers abandon its "crutch unit" philosophy. Compensating for lacking important tools by making one or two units crazy strong for their cost hasn't worked well in practice.

I give you the point for the map choices but in terms of player skill it basically can't get much better

It's the qualifier, there is a wide range of player skill in this right now and for the top brackets it's a noob stomp.

18/32 of the matchups were 2:0 wins. These threads are interesting but people need to wait for more data before jumping to conclusions.

Most of community agrees ukf is underpowered. So it is possible that community will agree on some set of changes aimed to change that. These changes will probably be implemented and the situation will change in some way. That is all certain.

On the other hand, there is no reliable way to estimate effect of such changes, just as there is no good way to measure if they were for the better. This is very important to acknowledge, as some community members try or may try to use small, overinterpreted set of statistical data to prove one of possible directions is universally right.

Biggest thing is, brits were a faction that lacked a lot of basic stuff(mortars, non doc mobile rocket arty, vehicle snare, no reliable light if you went for bofors to name a few) and relied on gimmicks to make up for it.

Modders decided it would somehow make them balanced if they have removed or nerfed to the very ground every singular thing brits relied on without giving them anything from the basic mandatory kit.

Couple the two and you have a perfect recipe for underpowered trash faction.

Biggest thing is, brits were a faction that lacked a lot of basic stuff(mortars, non doc mobile rocket arty, vehicle snare, no reliable light if you went for bofors to name a few) and relied on gimmicks to make up for it.

Modders decided it would somehow make them balanced if they have removed or nerfed to the very ground every singular thing brits relied on without giving them anything from the basic mandatory kit.

Couple the two and you have a perfect recipe for underpowered trash faction.

That is the design standpoint and I don't disagree with that. The point is, there is no need to mix statistics into that, especially if the data set is small and biased for obvious reasons.

It's the qualifier, there is a wide range of player skill in this right now and for the top brackets it's a noob stomp.

18/32 of the matchups were 2:0 wins. These threads are interesting but people need to wait for more data before jumping to conclusions.

The skill range is clearly broader than my post suggests, but I really think, that these qualifiers are still a lot more reliable than average ladder matches. We should probably still wait until the whole thing is trough to see a trend

The skill range is clearly broader than my post suggests, but I really think, that these qualifiers are still a lot more reliable than average ladder matches. We should probably still wait until the whole thing is trough to see a trend

There are two ways to gather datasets for faction ballance problem.

One is automatch. It fails due to low awerness of game mechanics.

The other is tourneys. This one fails becouse small number of players induces bias.

There is no way around that. The more data you will gather and the smaller the population of players tested, the bigger bias you will get. So the trend in the data will not mean anything on its own. What can mean something is a difference in results between two big enough datasets based on same players, but I highly doubt there will ever be a single tournament based dataset like that big enough to support any conclusions.

There are two ways to gather datasets for faction ballance problem.

One is automatch. It fails due to low awerness of game mechanics.

The other is tourneys. This one fails becouse small number of players induces bias.

There is no way around that. The more data you will gather and the smaller the population of players tested, the bigger bias you will get. So the trend in the data will not mean anything on its own. What can mean something is a difference in results between two big enough datasets based on same players, but I highly doubt there will ever be a single tournament based dataset like that big enough to support any conclusions.

What on earth are you on about, this isn't just 'some' players these are the best players in the world playing the game as efficiently as they can on the best maps we have....

you now have 151 games in your data set.

It satisfies any quality or quantity measurement you need.

You will never get a better opportunity to judge faction strength in 1v1 than this.

This patch shows that "Axis" isn't overpowered. Rather OKW is good but Ost is weaker than SU and USF. Let's not make the mistake that happened a year ago to buff allies across the board because OKW is in the lead and leave Ost in the dust. I don't think I remember a patch where Ost was actually better than OKW.

What on earth are you on about, this isn't just 'some' players these are the best players in the world playing the game as efficiently as they can on the best maps we have....

you now have 151 games in your data set.

It satisfies any quality or quantity measurement you need.

You will never get a better opportunity to judge faction strength in 1v1 than this.

I agree that there will never be better data set than that. Sadly, it still doesn't get any close to being enough to conclude anything. You might have 151 games, which is very little for the number of parameters involved, but even that is divided between 6 types of matchups. You also have as little as around 32 players, which is a very small population. A population way too small to remove individual bias. Especially as results and strategies of these 32 players are highly correlated.

All in all data gathered this way can only show current state of the meta, which is what people think the game is and which we by definition know a lot of, but it is very far from showing anything about true power of the factions. And even further from giving hints for informed ballance decisions.

I'm not saying that statistics has no way to work in coh2. For example the automatch veto stats were very interesting and the dataset was big enough, even if the results were highly misinterpreted. But the size and quality (correlations) of this particualar set is truly laughable if you would like to prove anything with it.

EDIT: Forgot one thing. The map pool is obviously even smaller than the player population. The good thing is that it has much less correlation. The bad one is that its tiny size induces additional bias, that highly reduces the confidence level of any conclusions drawn for maps that were not included in the tournament map set. And we can't really ballance the game for tournament maps only, can we?

Ost is suffering because they’re basically pigeonholed into getting a puma to counter the t70, otherwise the t70 runs rampant because 3 AT gun shots to kill it is never happening, and the vet 2/3 sight is essentially map hacks. This means Ost usually gets stuck on t3, which is completely dominated by t34/85s, su85s, Mark, and kv1s.

OKW can LUL in a command panther even after going luchs/puma (if needed). Plus OKW early game is a lot stronger.

UKF lacks basic tools, no way around it. They need a snare, they need a way to contest houses and some mobile indirect fire wouldn’t hurt.

USF is mainly good because of the Pershing in my opinion. Sov late game is much stronger in general but being able to LUL out a Pershing is so punishing to Ost because zooks/Pershing totally dominate Ost t3.

That’s the way I see things at least.

This patch shows that "Axis" isn't overpowered. Rather OKW is good but Ost is weaker than SU and USF. Let's not make the mistake that happened a year ago to buff allies across the board because OKW is in the lead and leave Ost in the dust. I don't think I remember a patch where Ost was actually better than OKW.

Ostheer is better than the usf, look at the stats again.

It's the qualifier, there is a wide range of player skill in this right now and for the top brackets it's a noob stomp.

18/32 of the matchups were 2:0 wins. These threads are interesting but people need to wait for more data before jumping to conclusions.

the player switch faction after a match. a 2:0 noob stomp basically balance itself out since it will give a win to both the allies and axis team.

the player switch faction after a match. a 2:0 noob stomp basically balance itself out since it will give a win to both the allies and axis team.

There are 5 factions, not just 2 sides. Looking at balance as axis vs allies is a recipe for disaster.

This patch shows that "Axis" isn't overpowered. Rather OKW is good but Ost is weaker than SU and USF.

How does it show this? I don't think the tourney data means everything, but Ost is doing just fine against US so far.

Sure, but I don’t think we need statistically significant evidence to conclude that UKF is weak. I mean... does even a single player in the top 50 think Brits are the strongest faction right now?

As for why Sov got better, my take is that people have gotten better at mitigating the Puma, thus they’re better at making it to the later stages of the game where Sov t4 starts to dominate Ost - especially Ost t3, especially especially when mark and t34/85s are involved.

Ost is suffering because they’re basically pigeonholed into getting a puma to counter the t70, otherwise the t70 runs rampant because 3 AT gun shots to kill it is never happening, and the vet 2/3 sight is essentially map hacks. This means Ost usually gets stuck on t3, which is completely dominated by t34/85s, su85s, Mark, and kv1s.

OKW can LUL in a command panther even after going luchs/puma (if needed). Plus OKW early game is a lot stronger.

UKF lacks basic tools, no way around it. They need a snare, they need a way to contest houses and some mobile indirect fire wouldn’t hurt.

USF is mainly good because of the Pershing in my opinion. Sov late game is much stronger in general but being able to LUL out a Pershing is so punishing to Ost because zooks/Pershing totally dominate Ost t3.

That’s the way I see things at least.

I agree 100% with basically every single point in these posts. Nailed it.

Ost is suffering because they’re basically pigeonholed into getting a puma to counter the t70, otherwise the t70 runs rampant because 3 AT gun shots to kill it is never happening, and the vet 2/3 sight is essentially map hacks. This means Ost usually gets stuck on t3, which is completely dominated by t34/85s, su85s, Mark, and kv1s.

Hasn't it been the case for years now? T-70 has been nerfed several times, it was much stronger before. Now with the incredibly cheap 222 you can build two of them and do the same rampage as a T-70 before it comes to the field. And two 222 counter a T-70.

USF is mainly good because of the Pershing in my opinion. Sov late game is much stronger in general but being able to LUL out a Pershing is so punishing to Ost because zooks/Pershing totally dominate Ost t3.

That’s the way I see things at least.

IMO it is the other way around, You pick a Pershing because Ostheer players prefers to go Mobile defense + T3 spam.
Mobile defense hard counter both T1 and T2 and can compete with USF T3 with the addition of some Ostheer T3 units.
At the end, this is the same logic as for Soviet going KV1 or IS2 commanders to counter Mobile Defense.

I think we really need to get ride of Mobile defense to see what Ostheer is capable of. Today Matchmaking or tourney the result is the same, Mobile Defense everywhere that invalidate any analyse other than Mobile Defense has to be nerf.

And for OKW, 4 volks into Luch is too strong vs USF. The moment you usually start getting some air with a Stuart, the Puma hit the field. If you survive it and start thinking about expending, T3 is already there denying 1/3 of the map or a command panther that will invalidate you first sherman...

What on earth are you on about, this isn't just 'some' players these are the best players in the world playing the game as efficiently as they can on the best maps we have....

you now have 151 games in your data set.

It satisfies any quality or quantity measurement you need.

You will never get a better opportunity to judge faction strength in 1v1 than this.

No, scientifically speaking this is incorrect. It's basically an unfair comparison. Tournament players usually have only two factions they're very good at, and they will perform less with the others. This means that data is basically corrupted or biased. Like how in the quali rounds none of the top players picked Brits, leaving the UKF statistics to be filled in by the less good players. This likely artificially brings the loss rate upwards (not denying UKF in 1v1 appears weak though).
The only fair comparison with such a low sample size would be when all results are from the same player with the same level of skill for each faction, which is obviously impossible.

There's also still the fact that in CoH2 there are so many factors (RNG engagements, map balance, positioning, faction vs faction balance, player skill and endurance, etc, etc) that a sample this low is basically insignificant.

IMO the only way to reliably measure faction balance is to take any top percentage (like top 10% or 20%) of automatch matches so the sample size will be big enough (in the thousands) while also using players who will generally understand the game and play the mechanics to their full potential.

IMO the only way to reliably measure faction balance is to take any top percentage (like top 10% or 20%) of automatch players so the sample size will be big enough (in the thousands) while also using players who will generally understand the game and play the mechanics to their full potential.

What I'm saying is for this game this is the only way to reliably measure faction balance. You're talking about thousands? There's only a few hundred people that play this game efficiently, I myself am capable of getting into the top 200 and I am a very casual 28yo 1v1 automatcher that works a 40 hour week job...

You're honestly looking at 50 "tournament players" that are the best 1v1 automatch players, they're one and the same. There's just a few people that can't travel for gcs purposes like Luciano, Paula etc. but other than that there's pretty much no one else.

This is reliable data.

Don't forget the wildcard brackets too, 40 additional games there our sample size is 180 games in tourney conditions from the best players, it will rise to 320 by the end of July..........

Even when we hit 360 tourney games in ideal test conditions between top players, I bet you people are still talking about phantom 1000s of automatch players and what is "significant".

What I'm saying is for this game this is the only way to reliably measure faction balance. [...]

This is reliable data.

Well, just because this is one of the few if not the only way to gauge faction balance doesn't make it reliable.

Not saying that the numbers are meaningless, far from that. They are probably the best thing we have; automatch doesn't work, really, because the number of games between top players is really small (I know, I looked into this and discarded the idea); and then people would argue anyways that the players were potentially not playing to win but trying something out.

That said, as others pointed out there are tons of biases here (player preference, map pool, fixed map positions) and if you breakdown the numbers further to smaller subsets to get rid of some of these, you again look at obviously small sample sizes.

So, yes, the numbers can certainly indicate some trends, but they are no hard proof (just compare OST/SOV and OKW/SOV winrates for both tourneys).

What I'm saying is for this game this is the only way to reliably measure faction balance. You're talking about thousands? There's only a few hundred people that play this game efficiently, I myself am capable of getting into the top 200 and I am a very casual 28yo 1v1 automatcher that works a 40 hour week job...

Woopsie, I meant matches instead of players when I described the adequate sample size.
So thousands of matches by a top percentage of players would IMO be the only way to gather reliable data. Although the game's population by now is probably too small even for that to work as Siphon described above.


I get what you're trying to say, and the GCS2 results are surely a good indicator, but it's not statistically reliable data. As Siphon already mentioned, the big difference in win rate between OST/SOV in the first and second quali rounds is proof of that.