Win-Ratio 29.10. - 21.11.2014 + Other Fun Stats

3vs3 and 4vs4 will never be balanced, in those games you are to much relied upon the general IQ or depths of stupidity that is called your "teammate".

their is a big different beiging good in a rts game and and have a high IQ.

So, overall, 1v1 is fairly balanced. Good. That's a start.

The next steps should be to fix the remaining issues with game variety and enjoyability, such as:
- Map Balance: most if not all the maps favor either Axis or Allies due to their varying competences at long- and short-range combat.
- Call-in Meta: reduces strategic options and invalidates many commanders, especially for the Soviets and Ostheer.

With OZ making that kind of a comment, there is no need for him to even look at the spread sheets, Statistics is clearly above him

Then please enlighten me how this kind of statistics represents the game balance and how exactly it has been measured?
Was the skill of each player measured and compared? What was the size of initial sample? What's the data distribution in term of player skill? Their ELO? Was the number of games even for each faction? In team games, was the skill of each player (or ELO) even for each team for each game?

i can only say this : LOLLLLLLLLLLLLLLLLLLLLLLLL. 1 vs 1 is balanced. in fact soviets and usf have far less things to complain about then the ostheer who are only being propped up by the tiger. nearly every single faction is close to the 25% marker. a few weeks a ago the soviets had 28 % btw it simply fluctuates slightly but overall the balance is good in 1vs 1.

A few weeks ago soviets only had a marginal greater win ratio in 1v1s only. My statement still stands about Axis being favored heavily in 2v2's, 3v3's, and 4v4's.

Then please enlighten me how this kind of statistics represents the game balance and how exactly it has been measured?
Was the skill of each player measured and compared? What was the size of initial sample? What's the data distribution in term of player skill? Their ELO? Was the number of games even for each faction? In team games, was the skill of each player (or ELO) even for each team for each game?

This statistical experiment shows there is an un even win loss distribution between the factions. In an RTS balance being achieved can be measured in how close the win loss ratios are to each other or number of wins in a given sample. In this case the numbers show a correlation between axis and higher wins/ win loss ratio in higher player count game modes. The numbers are also pointing that in a 1v1 situation the game appears to be balanced. A statistical representation in this case has no way of proving why something is the case, that would revolve around unit statistics and end game stats, but this is a good way of proving there is a problem.

The sample is games from the top 200 players of each faction, they have a good representation of how the game is supposed to work with out as many learn to play issues as lower ranked players. This means the statistics can be trusted as a reasonable example of the state of the game. He could of done a pure random sample, but then you could have very skewed results from the majority of players playing casual and suffering from player issues rather than balance. Not that balance isn't affecting them, but lower level players run into issues of loosing units to multi tasking rather than something being over powered.

A few weeks ago soviets only had a marginal greater win ratio in 1v1s only. My statement still stands about Axis being favored heavily in 2v2's, 3v3's, and 4v4's.

Allies, according to these statistics, won 50.2% of the 2v2 matches. Does not compute how that's supposed to be a massive Axis Advantage.

If anything, the problem I see with 2v2 is that OKW is so much better than Ostheer. Still, compared to 3v3 or 4v4 it's all pretty close.

Nice work on the statistics- any chance of arranged team match scores though? random teamgames doesn't seem like it would reflect actual balance very well.

Allies, according to these statistics, won 50.2% of the 2v2 matches. Does not compute how that's supposed to be a massive Axis Advantage.

If anything, the problem I see with 2v2 is that OKW is so much better than Ostheer. Still, compared to 3v3 or 4v4 it's all pretty close.

go to the original document on the drive and you'll see the whole picture. These pie charts are not the only information

i beleive that the okw should only receive 2 fuel and not 3. It would fix a lot of issues.

I think it got fixed already. Last time my wehr mate built a fuel cache I only got +2 fuel as OKW.

Great work, thanks. Can you make same but not with top200?

This statistical experiment shows there is an un even win loss distribution between the factions. In an RTS balance being achieved can be measured in how close the win loss ratios are to each other or number of wins in a given sample. In this case the numbers show a correlation between axis and higher wins/ win loss ratio in higher player count game modes. The numbers are also pointing that in a 1v1 situation the game appears to be balanced. A statistical representation in this case has no way of proving why something is the case, that would revolve around unit statistics and end game stats, but this is a good way of proving there is a problem.

The sample is games from the top 200 players of each faction, they have a good representation of how the game is supposed to work with out as many learn to play issues as lower ranked players. This means the statistics can be trusted as a reasonable example of the state of the game. He could of done a pure random sample, but then you could have very skewed results from the majority of players playing casual and suffering from player issues rather than balance. Not that balance isn't affecting them, but lower level players run into issues of loosing units to multi tasking rather than something being over powered.

Sorry but no.

All this statistics does is show the win ratio for all factions and that's it. You can't read game balance from these statistics due to the fact that although the size of the samples is equal, you can't guarantee that quality of these samples is equal as well. Top 200 of OKW players maybe be slightly more skilled than first 200 of USF players. What's more you can't even provide that each match will be against equally skilled opponents all the time because auto match will expand its brackets when no suitable opponent is found within specific period of time. I fought against JellyDOnut once and I don't have to tell you how this game ended due to skill level difference between us. It is quite common to get lower skilled opponent in a match up and it's quite common that your opponent will be much more skilled. The number of equal games I had is quite low. I am pretty sure it's the same when you're top 200. Maybe you get less skilled opponents more often because 200 players that's not that high number and they won't be able to play any games at all if match making only paired them against other top 200 player every time.
Also you can't guarantee that none of this data wasn't affected by balance differences, for examples maybe OKW win ratio would have been much lower if it wasn't for over performing Obersoldaten or Kubel Wagon, etc.


What we have here is a classic example of interpretation of the data to suit your need. First thing they are going to warn you against on every Statistics lecture.


Edit:
One more thing. Test sample is not even big enough to safely draw any conclusions.

Sorry but no.

All this statistics does is show the win ratio for all factions and that's it. You can't read game balance from these statistics due to the fact that although the size of the samples is equal, you can't guarantee that quality of these samples is equal as well. Top 200 of OKW players maybe be slightly more skilled than first 200 of USF players. What's more you can't even provide that each match will be against equally skilled opponents all the time because auto match will expand its brackets when no suitable opponent is found within specific period of time. I fought against JellyDOnut once and I don't have to tell you how this game ended due to skill level difference between us. It is quite common to get lower skilled opponent in a match up and it's quite common that your opponent will be much more skilled. The number of equal games I had is quite low. I am pretty sure it's the same when you're top 200. Maybe you get less skilled opponents more often because 200 players that's not that high number and they won't be able to play any games at all if match making only paired them against other top 200 player every time.
Also you can't guarantee that none of this data wasn't affected by balance differences, for examples maybe OKW win ratio would have been much lower if it wasn't for over performing Obersoldaten or Kubel Wagon, etc.


What we have here is a classic example of interpretation of the data to suit your need. First thing they are going to warn you against on every Statistics lecture.


Edit:
One more thing. Test sample is not even big enough to safely draw any conclusions.

Except it is, there is a reason statistic samples do not need to have thousands upon thousands of numbers to be accurate. As long as you have a sample that is correct enough (which this is). If the top 200 are not accurate enough for you, then your samples will never be good enough unless you take the whole population. It is highly unlikely that the majority of the OKW players are just all strait up better than all the allies players, the same could be said of when the Soviet win loss ratios were out of control

The sample quality is fine here, you could take a larger sample size and the results would not change very much (because this sample was chosen adequately).

It seems you want the sample to be absolutely perfect, which is absolutely absurd

I think these stats are all fake

I'll say there no much info you can draw from a 5-10%> difference. But it's nice to see trends and quantify how big the gap of axis play on 3v3+ is, specially on random.

Great work, thanks. Can you make same but not with top200?

Theres no way to draw information from those who are 200+ on the rank. Only Relic has that data.

Except it is, there is a reason statistic samples do not need to have thousands upon thousands of numbers to be accurate. As long as you have a sample that is correct enough (which this is). If the top 200 are not accurate enough for you, then your samples will never be good enough unless you take the whole population. It is highly unlikely that the majority of the OKW players are just all strait up better than all the allies players, the same could be said of when the Soviet win loss ratios were out of control

The sample quality is fine here, you could take a larger sample size and the results would not change very much (because this sample was chosen adequately).

It seems you want the sample to be absolutely perfect, which is absolutely absurd

It doesn't matter. There is no way to prove it is otherwise neither.
For this to be true you have to assume there is an equal number of games for each of the sides + all of these games were against opponents of equal skill. We all know that's not possible hence this is all one big assumption and can't be treated as a research. You bend the data to suit your ideology.
All this indicates is how it's the win ratio among 200 players for all of the factions, but that's not enough to draw any conclusion from this as there are too many variables.

yuou know if their was a rifleman with a bar for each volks the result would be a horrific slaughter of all those volks.

rifleman?! nope. i play soviet side. Its only 1 man. Another spam g43 rifles and obers so u cant do anithing. In late game u isu cant penetrate tiger or tiger 2 and F8ed by shreks spam. so soviet sucked every time

The problem with these statistics is that you do not know the skill level of the different top 200 faction groups.

For example, if it is considerably harder to play faction X (requiring higher skill/effort), the players in top 200 of faction X could still have the same win percentage as top 200 players of faction Y.

Another problem: These stats does not reveal whether individual units or stages of the game (early/late) are balanced. The win percentages might be equal, but it might surely still be a problem if faction X wins 100% of games shorter than 20 minutes but faction Y wins games longer than 20 minutes.

The more obvious difference in win percentage that can be seen from 3v3 and 4v4 actually reveals that the above might be the case, as large team matches often end in the late game. If a 1v1 match goes to the late game, it might suffer from the exact same balance problems most often seen in 3v3 and 4v4.

These statistics does provide some insight, but one should be careful when interpreting them. Concluding that everything is fine based on these statistics is a stretch. In other words, well-argued balance suggestions still have merit.

Except it is, there is a reason statistic samples do not need to have thousands upon thousands of numbers to be accurate. As long as you have a sample that is correct enough (which this is). If the top 200 are not accurate enough for you, then your samples will never be good enough unless you take the whole population. It is highly unlikely that the majority of the OKW players are just all strait up better than all the allies players, the same could be said of when the Soviet win loss ratios were out of control

The sample quality is fine here, you could take a larger sample size and the results would not change very much (because this sample was chosen adequately).

It seems you want the sample to be absolutely perfect, which is absolutely absurd

You are right and wrong at the same time.

The test sample, in this case, is in fact, too small. A Test sample needs be somewhere ~1.000 to be accurate.

If you have a sample of ~1k then every other sample higher than that, 10k, 100k will change or deviate the 1k statistic by roughly x<2% (at most, it's actually closer to 0 < x < 1 ).

I think these stats are all fake

please elaborate

Sorry but no.

All this statistics does is show the win ratio for all factions and that's it. You can't read game balance from these statistics due to the fact that although the size of the samples is equal, you can't guarantee that quality of these samples is equal as well. Top 200 of OKW players maybe be slightly more skilled than first 200 of USF players. What's more you can't even provide that each match will be against equally skilled opponents all the time because auto match will expand its brackets when no suitable opponent is found within specific period of time. I fought against JellyDOnut once and I don't have to tell you how this game ended due to skill level difference between us. It is quite common to get lower skilled opponent in a match up and it's quite common that your opponent will be much more skilled. The number of equal games I had is quite low. I am pretty sure it's the same when you're top 200. Maybe you get less skilled opponents more often because 200 players that's not that high number and they won't be able to play any games at all if match making only paired them against other top 200 player every time.
Also you can't guarantee that none of this data wasn't affected by balance differences, for examples maybe OKW win ratio would have been much lower if it wasn't for over performing Obersoldaten or Kubel Wagon, etc.


What we have here is a classic example of interpretation of the data to suit your need. First thing they are going to warn you against on every Statistics lecture.


Edit:
One more thing. Test sample is not even big enough to safely draw any conclusions.

Oz, we do not have all the data that you wish to see, you can ask Relic o please hand them over nicely.

But i think straight up ignoring all this data is unreasonable unless we are sure that it has been fabricated (which i doubt.. who would take all this time just to fake some numbers and charts).

On the subject sampling.. I think having the top 200 players of each faction is a good sample group and size. True, we do not know the average ELO of these pools, but we can assume that top 200 players are fairly skilled and not newbies. I would argue that using random sampling for research in drug testing, public safety, etc., are needed, but not so much in this game. Competitive games usually balance to top-tier play, so having a healthy sample of the top 200 should be sufficient. Having random sampling, or a larger sample size would mess up some of the results and it would be more difficult to assume why the numbers are so. Top 200 players are ranked as such because they have adequately mastered the mechanics and units of their respective factions. People rank 200> may have L2P difficulties; We do not know if these players have trouble because they have not practiced enough, or if it really is a balance issue.

SO in short, we do not have all the numbers. That is up to Relic to release. We do not have the whole picture, but i believe we have a nice blueprint for discussion and debate.



Things i would like to see:

1. Average ELO for each top 200 bloc in their respective factions
2. Like and said (great idea!), it would be nci to know the winrates of games lasting in 15 min, 20min, 25, min etc..
3. win rates on specific maps, which would be extremely interesting

You are right and wrong at the same time.

The test sample, in this case, is in fact, too small. A Test sample needs be somewhere ~1.000 to be accurate.

If you have a sample of ~1k then every other sample higher than that, 10k, 100k will change or deviate the 1k statistic by roughly x<2% (at most, it's actually closer to 0 < x < 1 ).

I am not saying that having 1000+ sample is bad, i actually would love if we had data for many things. Having a large sample size is necessary for seeing if specific health plans work, if a certain medication works, if cameras decrease accidents (more of a time lapse, but eh), etc. I would like to coment that having a huge sample size would be nice, but it would probably be difficult to draw some conclusions from it. WE do not know if these players are newbies, have L2P issues, or what not. The players may be losing game because they have not utilized the strengths of their armies well enough.

I think looking a high-level players and their experiences with the game is good for balancing such as adjusting units here and there. Balancing using a large sample size of players ranging from the pro to the fresh newbie is good too though. These could lead to general "quality of life" changes that could make it easier for players.

in conclusion, having a smaller sample size of top level players is helpful when it comes to balancing, but having a large sample size of many players is also helpful for general "ease of use" mechanics. But for now, in the context of "balance", i think it is more helpful to stick with the top 200.

why i think thats its not balance yet? look at this. germans may win without OKW and soviet don`t do anything without USF. This factions problem. soviet still sucked.
http://www.coh2.org/file/6595/germans-op-without-okw.rec