EP Win Rate Statistics

[Lukas_L99]

Would be interesting to see each EP players‘ win rate with each civ. But that might be too much work I guess?

[BrookG]

Would be interesting to see each EP players‘ win rate with each civ. But that might be too much work I guess?

probably, unless we extract the data from esoc elo ladders

[_H2O]

You could do a filter on ELO or PR difference between players. You could also show average PR of players on each civ to help with context.

[kami_ryu]

Wtf Brit players?

that is exactly what I thought when I saw this

[bwinner]

Would be interesting to see each EP players‘ win rate with each civ. But that might be too much work I guess?

probably, unless we extract the data from esoc elo ladders

I am trying to do it now, doesn't seem to be hard, although I only have the datas from tourneys that you have provided so far, maybe @Development Team could help me to have more ? who should I contact ?
Edit : missread that's not what I am doing missread^^
but I think I still would like to get this, I am trying to do statistic on civ winrate that include players elo

[bwinner]

Ok I did it and here are the result I have obtained :

This was my methode
- civ strength = 0
- for each game of EPL : civ strength change = elo player change for the player playing this (exept I did as if both player were playing with their max elo)

I also join a full PDF with explanation and code if you are interested in exploring more about it. I think we could creat a good methode to evaluate the civ power with this kind of stuff @zoom

PS : Note that the results here are not significant at all since the number of games in EPL is quite small. Although I think this method could be extended to a large database of games played on EP.

[wardyb1]

Ok I did it and here are the result I have obtained :

This was my methode
- civ strength = 0
- for each game of EPL : civ strength change = elo player change for the player playing this (exept I did as if both player were playing with their max elo)

civStrength.JPG

I also join a full PDF with explanation and code if you are interested in exploring more about it. I think we could creat a good methode to evaluate the civ power with this kind of stuff @zoom

PS : Note that the results here are not significant at all since the number of games in EPL is quite small. Although I think this method could be extended to a large database of games played on EP.

I'm confused. Is the more postive a number is the stronger it is? Or the more negative?

[BrookG]

Ok I did it and here are the result I have obtained :

This was my methode
- civ strength = 0
- for each game of EPL : civ strength change = elo player change for the player playing this (exept I did as if both player were playing with their max elo)

civStrength.JPG

I also join a full PDF with explanation and code if you are interested in exploring more about it. I think we could creat a good methode to evaluate the civ power with this kind of stuff @zoom

PS : Note that the results here are not significant at all since the number of games in EPL is quite small. Although I think this method could be extended to a large database of games played on EP.

I will check it tomorrow. NWC was also played in EP6 afaik. So we can include those data too

[bwinner]

Ok I did it and here are the result I have obtained :

This was my methode
- civ strength = 0
- for each game of EPL : civ strength change = elo player change for the player playing this (exept I did as if both player were playing with their max elo)

civStrength.JPG

I also join a full PDF with explanation and code if you are interested in exploring more about it. I think we could creat a good methode to evaluate the civ power with this kind of stuff @zoom

PS : Note that the results here are not significant at all since the number of games in EPL is quite small. Although I think this method could be extended to a large database of games played on EP.

I'm confused. Is the more postive a number is the stronger it is? Or the more negative?

Positiv ist stronger. As said though it's just on how they performed during EPL+it's biased by people with no games on EP like wicked Cossack (he has max 1600 elo...). Although it's more to show how we could do because if I was using all games for pr30+ ep for instance, I think we would get something much more relevant.
Also the method was a bit arbitrary here, but we can easily use any other formula that would mix elo/winrate with this method.

[Goodspeed]

Ok I did it and here are the result I have obtained :

This was my methode
- civ strength = 0
- for each game of EPL : civ strength change = elo player change for the player playing this (exept I did as if both player were playing with their max elo)

civStrength.JPG

I also join a full PDF with explanation and code if you are interested in exploring more about it. I think we could creat a good methode to evaluate the civ power with this kind of stuff @zoom

PS : Note that the results here are not significant at all since the number of games in EPL is quite small. Although I think this method could be extended to a large database of games played on EP.

You mean civ strength is the average of the ELO difference between the player playing the civ and their opponent? I think that's what we're looking for here.
So if you have someone playing Brit and they're -100 ELO from their opponent, and someone else playing brit who is +50 ELO from their opponent, civ strength would be -25?
"Civ strength" is a rather misleading name in that case

[bwinner]

Ok I did it and here are the result I have obtained :

This was my methode
- civ strength = 0
- for each game of EPL : civ strength change = elo player change for the player playing this (exept I did as if both player were playing with their max elo)

civStrength.JPG

I also join a full PDF with explanation and code if you are interested in exploring more about it. I think we could creat a good methode to evaluate the civ power with this kind of stuff @zoom

PS : Note that the results here are not significant at all since the number of games in EPL is quite small. Although I think this method could be extended to a large database of games played on EP.

You mean civ strength is the average of the ELO difference between the player playing the civ and their opponent? I think that's what we're looking for here.
So if you have someone playing Brit and they're -100 ELO from their opponent, and someone else playing brit who is +50 ELO from their opponent, civ strength would be -25?
"Civ strength" is a rather misleading name in that case

No, that's not what I did. What I did is:
-Initiate all civ strength to 0
-Iterate on all match played during EPL ;
-let's say we have turk 1800 elo playing India vs sheckler 1600elo playing dutch.
-if turk wins, India strength is updated to something like 0.1 and dutch to -0.1 (fake number but small number is what is important)
-if sheckler wins though, India strength is updated to like -2 and dutch strength to 2.
-then move to next match and do the same.
I think it's better than raw winrate because it reduce the impact of games where one player would win with any civ.
Although now that the tools are here, I can do what you asked very easily if you want

[Goodspeed]

Ah yeah I guess that works but it does make it a bit confusing for people because the numbers seem arbitrary (since there's a magic formula hidden behind them). If you can easily get the average ELO difference per civ, that would be helpful

[bwinner]

Ah yeah I guess that works but it does make it a bit confusing for people because the numbers seem arbitrary (since there's a magic formula hidden behind them). If you can easily get the average ELO difference per civ, that would be helpful

There you go. I did it only for EPL here once again, and it's player max ep elo (I could do current ep elo if you think it's more relevant).
EDIT : it was the average elo of the player who have played a civ during EPL @BrookG
Now I did what goodspeed asked as well so average elo difference of players for a civ compared to opponents elo

For instance, spain players were on average 180elo higher than their opponents during EPL and had an average elo of 1826.5.

EDIT2 :with more in the table @Goodspeed

[BrookG]

What is this last table? I can't understand @bwinner

[Goodspeed]

Nice. If you put those numbers next to the win rates, it becomes much more meaningful. Like if Spain won 50% of their games but the average ELO was 180 higher than their opponent, the civ is probably UP.
But only EPL is too small of a sample I guess.
Also I would take JP elo if possible. EP elo ladders aren't (even close to) settled

[BrookG]

@bwinner As I said you can take the data from nwc too. Great job, I am interested. I am working today, tomorrow I can give you access to edit the sheets. If you are interested we can talk a bit, how we can extract more info

[Kaiserklein]

No offense but maybe we should try to fix EP Elo before we even think about getting stats out of it

[Hazza54321]

Ep stats will never be accurate with such a low player base.

[bwinner]

Yeah I agree that it will probably be more accurate with jp elo, especially for EPL. I will look at it when I have time but there shouldn't be problem using it @Goodspeed
Although it doesn't entirely solve the problem of people using smurf with sometimes very inacurate elo, but I guess it will still be a lot better.

I will also try with NWC when I have time @BrookG , I mainly use EPL because I am just trying to developpe the idea so far, but this is next step for sure.

[HeadKilla]

The spain pain train has reached each destination without any delays or derailing.

Much unlike a EU train, so I hear.

[Riotcoke]

No offense but maybe we should try to fix EP Elo before we even think about getting stats out of it

I always like looking through ep elo and finding a 15 have the same elo as two majors

[[Armag] diarouga]

No offense but maybe we should try to fix EP Elo before we even think about getting stats out of it

+1

[bwinner]

No offense but maybe we should try to fix EP Elo before we even think about getting stats out of it

About that, I know that in chess for instance, change rate decrease after you've played a certain amount of games. I think this is a good feature that could participate in improving ep elo accuracy because it means you reach quickly a representative level that correspond to your starting level and then your elo doesn't move too fast which correspond to how your real level moves.

[Kaiserklein]

Well yeah, it's also how pr works. The 10-20 first games or whatever should make you gain/lose much more elo, so that you can quickly reach your "real" elo.

[[Armag] diarouga]

Glicko system