106% success !??

[FastPlayPlz]

HI,

I like playing spades and was curious to see statistics on highly ranked players. I picked one and discovered he played only 53 games, won 38, and have a 106% success .
-> how can you have more than 100% success, especially when not winning all games ???

There is obviously a flaw somewhere. I am curious if someone know about that.

[Volker_B]

I wondered about this, too, and tried to reverse engineer the numbers for me.
BGA takes into account that winning a 2-player game is easier than winning a 3-player or 4-player game. Therefore, victories are weighted depending on the number of players. You get 1 victory for a two player game, 1.5 victories for a 3-player game, and 2 victories for a 4-player game. I can't tell for more than 4 players or ties.
It explains my numbers very well.

[tarongot]

HI,

I like playing spades and was curious to see statistics on highly ranked players. I picked one and discovered he played only 53 games, won 38, and have a 106% success .
-> how can you have more than 100% success, especially when not winning all games ???

There is obviously a flaw somewhere. I am curious if someone know about that.

Check this topic. We wonder same some time ago:
https://boardgamearena.com/forum/viewtopic.php?t=48891

[BarnardsStar]

Simply put, 50% of players in 2p games are winners but only 25% of players in 4p games are winners. So it’s twice as hard to win 4p, and the algorithm here counts it as twice as much. So for each game win you get n/2 wins, and then divide by the total games played. 50% is average no matter how many players there are.

[tbhp]

The numbers are adjusted for multiplayer games.

While you may or may not agree with the concept, it has some justification, because when looking at the win percentage of a player for a particular game, the number displayed wouldn't mean anything if that method wasn't applied. You'd see a 25% winrate and may assume they lose most of their games, when in reality they only play 4 players games and so their win percentage is exactly what is expected.

The methodology to combine games with different player counts into a single number may be a bit wonky, but it serves to give a rough idea.

[tarongot]

The numbers are adjusted for multiplayer games.

While you may or may not agree with the concept, it has some justification, because when looking at the win percentage of a player for a particular game, the number displayed wouldn't mean anything if that method wasn't applied. You'd see a 25% winrate and may assume they lose most of their games, when in reality they only play 4 players games and so their win percentage is exactly what is expected.

The methodology to combine games with different player counts into a single number may be a bit wonky, but it serves to give a rough idea.

But this methodology gives false and hard to understand values, like mentioned 106% of wins. Not say that games are different. In one there is always 1 winner rest lose (for example 4 player 1 winner, 3 losers) in other there will be gradation of victories (first player, second player, third etc...). Do this methodology reflects that?
Not to say that for coop games, result is divided by half just because it is coop game. So hardly this methodology is consistent with itself.

After all ,winrate has value only by comparison to other players winrate score. So for some games even a 15% can be highest value in players pool.

[tbhp]

The numbers are adjusted for multiplayer games.

While you may or may not agree with the concept, it has some justification, because when looking at the win percentage of a player for a particular game, the number displayed wouldn't mean anything if that method wasn't applied. You'd see a 25% winrate and may assume they lose most of their games, when in reality they only play 4 players games and so their win percentage is exactly what is expected.

The methodology to combine games with different player counts into a single number may be a bit wonky, but it serves to give a rough idea.

But this methodology gives false and hard to understand values, like mentioned 106% of wins. Not say that games are different. In one there is always 1 winner rest lose (for example 4 player 1 winner, 3 losers) in other there will be gradation of victories (first player, second player, third etc...). Do this methodology reflects that?
Not to say that for coop games, result is divided by half just because it is coop game. So hardly this methodology is consistent with itself.

After all ,winrate has value only by comparison to other players winrate score. So for some games even a 15% can be highest value in players pool.

It seems to be the best way to reflect it. How else would you convey that someone wins 50% of the time in a game where they are expected to win 25% of the time?

Another alternative would be to give a separate winrate for each player count. But if you are gonna combine different player counts together, it wouldn't be very informative to just display the raw win percentage. It would be factual in telling you how many games they have technically won, but that's not really what we want to know when we look at someone's winrate.

[tarongot]

It would be factual in telling you how many games they have technically won, but that's not really what we want to know when we look at someone's winrate.

I think that exactly what we want to know. For rest ELO point/tittle is a way to reflect actual skill in game.

Current one only creates more questions about methodology and implementation. Like:
What is reason that coop has that rate divided by half?
What with games that has only 1 winner and rest in lose-draw?
How are games in friendly mode included in that percentage winrate?

And probably more. Since that is second topic about same in short time, current one is not clear nor understandable.

[tbhp]

I think that exactly what we want to know. For rest ELO point/tittle is a way to reflect actual skill in game.

Then what exactly do you want to know the win percentage for?

If you see that, say, a player won 25% of the time at a certain game, it doesn't tell you anything useful on its own. Because if they only played 4 player games, then that means they got the expected result. If they only played 2 player games, then it means they did much worse than their opponents. And if they only played 100 player games, then it means they did extremely well.

So the meaning of the number is totally different depending on player count. So what does the number mean to you personally? Why do you find it more useful than the adjusted number?

Current one only creates more questions about methodology and implementation. Like:
What is reason that coop has that rate divided by half?
What with games that has only 1 winner and rest in lose-draw?
How are games in friendly mode included in that percentage winrate?

The friendly thing would be an interrogation regardless of which method they use.

Same goes for ties, you'd have to see the specifics of what the site counts as a win or not.

For coop games, I suppose the rationale is that if you are two players, for example, then you are only "half" responsible for the victory. Admittedly, that doesn't make much sense to me in the context of integrating that data with the non-coop data. I suppose the problem is that those two modes are just fundamentally incompatible when trying to mesh them into a single number, because in coop you aren't on your own. It's impossible to know how big a part you played in a victory, but also, you can't hold it against a player to not have 100% been the reason for the win, because coop games are designed to be won together. So it just doesn't make sense to try to mix the winrates.

[tarongot]

That's why I wrote up that percentage gives you only knowledge compared with other players percentage.
Current one doesn't give you info either. Unless you know methodology behind it and how game result are counted. And we don't, we are only guessing based on observation. So what the point of that percentage anyway?

And yet coop have same percentage shown, just, by my observation lowered by half to what would be calculated if that wasn't coop. So yet another factor you need to know to properly read that winrate.

For me it is pointless. You have ELO, you have wins/all games. If that work properly (and for coop games it doesn't but that is different and much bigger problem) then what you need that percentage win anyway, especially without info how to read it?

EDIT: Now when I think of it I said it wrong. For coop it looks like it is half of REAL percentage win. So not inlcunding 2/3/4/x players games. So even more different. So how you read that percentage win for coop games anyway?