ELO on luck based games

[BarnardsStar]

And the award for greatest waste of space goes to … drumroll please …

Ok bud

Was it really necessary to quote 13 paragraphs to add zero meaningful content?

Ok bud.

[ChiefPointThief]

There is another point about ELO which isn't clear : Does it work ? Can a single number (or rather difference of numbers) give a good prediction of who will win (let's restrict to 2 player games for simplification). I've seen studies for chess-ELO which provide evidence that ELO works for chess (sample size was rather small for high ELO-gaps).

The question of whether Elo works would probably be answered:

1) "No", by ChiefPointThief

Huh? I said it isn't great for luck based games and gave you my solution to fix it which you agreed with. I've also said it was off for team games in another thread which you agreed w/ and gave my solution on that. I never said it doesn't work at all. It is just flawed in several aspects.

When it calculates I'm 90% vs a player in 2p caravan I agree with that analysis. Just another example how it can be flawed is that when I enter a 4p game it is still showing that I'm 90% vs that player but I can't have the same probability as in 2p. For one, caravan is a game the keeps the same amount of goods no matter the player count. Another one I can think like this is carcasoonne. So without even considering the factor of multiple player influence in a 4p game the actual game length is cut in half. Meaning the experienced player has half the time to outperform the weaker player which is a pretty large factor and most my elo came from 2p games. I've seen the argument on other games about if elo should be different for different player counts. This is game specific and I don't expect bga to go to this length.

My proposal for adjusting the probabilities isn't extreme and can be applied to all games within the same luck bracket (4 or higher).

https://boardgamearena.com/table?table=834900516

So take a look at this game. As you have agreed Frank in another thread even in games that have no luck factor 2 players who are only 64 elo a apart (you said over 100 in your example) are practically even. Now when you consider a 4 luck game these two players probabilities should be considered even closer than that of a non luck game. But as you can see Im projected to win 59% and only won 8.17elo (so I wouldve lost 11.83). If I'm projected to beat this player 18 more times out of 100 then I am a FAR superior player than them at this game. I can play 1000 games at the projected rate then lose 3 in a row and suddenly be ranked below this player. The system is pulling these two players back together rather quickly but they are in different tier brackets. This is why I end up being below players who only have a 50% winrate when mine is amongst the top players at 68%. My point is a better system wouldn't pull me below these players after just a bad 10 game streak.

[ChiefPointThief]

Also, let's not forget - BGA is a company. Companies want to make money. Maybe BGA's decision-makers believe the best way to do this is to appeal to the broadest audience, not the small number of players at the very high or very low end of things. Perhaps they believe that fixing every real or perceived problem for top players would turn off the vast majority of casual players.

I understand and agree with this comment. It isn't an accident that bga made the k factor 2x that of normal games for "competitive" arena mode. The leagues are unnecessary as only your final aes matters. It is designed to appeal to the mass and to get players to play more games which publishers get money for the time spent on their games by users. And after realizing that arena isnt neccasarily who performed the best but there are other factors ethical and non ethical that can enhance your score I became disinterested. I've accepted that. I've given examples in other similar threads how player a didn't have nearly as good as a season as player b but still placed above them. The issue I have is with users trying to tell me there's nothing wrong with the system, it works perfect, you're a sore loser, and a bunch of other nonsense.

[Fuchur]

The intention of ELO is quite clear : Predict the outcome of any match between players given their history of plays and has nothing to do with opinions.

It is easy to find exemples where something like that (a linear scale providing a good prediction) CAN'T exist (i've already given the exemple of the 3 non-transitive dices A, B and C : A > B, B > C and C > A with X > Y meaning X will win against Y in most cases).

The question remains open : Are there ANY games (except chess and GO) where ELO exists ?

There is another point about ELO which isn't clear : Does it work ? Can a single number (or rather difference of numbers) give a good prediction of who will win (let's restrict to 2 player games for simplification). I've seen studies for chess-ELO which provide evidence that ELO works for chess (sample size was rather small for high ELO-gaps).

The question of whether Elo works would probably be answered:

1) "No", by ChiefPointThief

2) "Yes, but with different k-factors for games with different amounts of luck in the games" by dschingis27

and

3) "That depends on what we are trying to accomplish with Elo", by me.

Elo (whether by intention or not) serves several functions. Maybe here on BGA, Elo serves purposes that make the site owners and developers happy, even if these purposes are different from the stated intentions of Elo (including any stated intentions for the purpose of Elo on this website.)

[FrankJones]

The intention of ELO is quite clear : Predict the outcome of any match between players given their history of plays and has nothing to do with opinions.

The intention of Elo ... where? when?

When it was first created? When it was first used for chess matches?

Or, here and now at BGA? The intention of Elo when it was first created and used may not be the same as its intended use and meaning here in the year 2026 at Board Game Arena.

[FrankJones]

Huh? I said it isn't great for luck based games and gave you my solution to fix it which you agreed with. I've also said it was off for team games in another thread which you agreed w/ and gave my solution on that. I never said it doesn't work at all. It is just flawed in several aspects.

Maybe I am getting people's replies mixed up. I thought you had been advocating for an entirely different ratings formula because you believed Elo does not work for a website like BGA.

And I do not recall agreeing with your solution. But, it has been a long thread. Perhaps you can redirect me toward the post(s) in which you proposed the solution you're saying I agreed with?

My recollection was that you wanted a different formula; dschingis27 wanted to keep Elo but use different k-factors, and that I was fine with things the way they are but that I would like additional information and leaderboards to be displayed (rolling average Elo and highest-ever Elo).

That recollection is why I answered Fuchur's question of "does Elo work" by referring to how I thought you, dschingis27, and I would answer.

[FrankJones]

The question remains open : Are there ANY games (except chess and GO) where ELO exists ?

What do you mean "exists"? Elo is used for all games here on BGA, many of which have zero luck. Why would you suggest or imply that Elo does not "exist" for those games?

[Jellby]

I'd suggest simply asking ChatGPT (or any other alternative) what the rank should be after each match

[Fuchur]

Whenever you have this kind of questions, WIKI is your friend :

The Elo[a] rating system is a method for calculating the relative skill levels of players, originally designed for rating chess players. It is a special case of the Bradley–Terry model. The Elo system was invented as an improved chess rating system over the previously used Harkness rating system[1] and has since been adapted for use in other zero-sum games and sports, including tennis, association football (soccer), American football, baseball, basketball, pool, various board games and esports. It is named after its creator Arpad Elo, a Hungarian-American chess master and physics professor.

(see https://en.wikipedia.org/wiki/Elo_rating_system)

The intention of ELO is quite clear : Predict the outcome of any match between players given their history of plays and has nothing to do with opinions.

The intention of Elo ... where? when?

When it was first created? When it was first used for chess matches?

Or, here and now at BGA? The intention of Elo when it was first created and used may not be the same as its intended use and meaning here in the year 2026 at Board Game Arena.

[Fuchur]

It is (always) possible to calculate a statistic and give it the name ELO -- the question is : Does this statistic have the predictive value for further outcomes (as the chess-ELO seems to have) ? There are situations where the answer is no --- in these cases ELO doesn't exist.

My question was (and is) : are there any studies about other games than chess (and GO) which show that the statistic has the claimed predictive value ??

The question remains open : Are there ANY games (except chess and GO) where ELO exists ?

What do you mean "exists"? Elo is used for all games here on BGA, many of which have zero luck. Why would you suggest or imply that Elo does not "exist" for those games?