Hey guys,

i just had a discussion on discord and wanted to know, if your rating is affected and how much it is affected, if you got an opponent who's playing some games until he loses one and then surrender his other games and leaves the seasonal ladder.

Let's say he's playing the first 4 games, wins 3 of them and loses the 4th to leave the ladder.

I always thought it would have a significant effect on my rating. Then i heard a long time ago, that it shouldn't be the case or if, than just to a very minor extend, that you can almost ignore it.

So what's the truth?

Edited 11/21/2019 21:36:46

His apparent rating is low since he doesn't get the game bonuses, but the rating used for calculating your bonus isn't affected by his leaving. There BayesElo specific effects due to the low number of games, but I don't know the specifics. They have been described many on the forum though, usually by Farah.

Yeah if someone wins his first 9 games then loses 1 to you, you will still benefit immensly despite his rating appearing to be low in the end of the season.

For every game you receive you earn 65 extra points. That measure is to prevent the situation like this:

https://www.warzone.com/LadderGames?ID=4007&LadderTeamID=2078

Hhh won the season despite only having played 17 games instead of all 20. If he played all 20, he might have lost 3 more games so adding 65 points for each game seems like a good incentive for players to play all 20 games.

The point where players leaving the ladder hurts your rating is, when the leaving player surrenders his ongoing games no matter if he was winning or losing. In that case his rating that is the base of calculating his opponents rating is much lower that his skill suggests.

Edit: in fact, someone leaving the ladder at a point where he is overrated, is actually an advantage. You get more points than you'd get if he played stronger opponents and lost a few more games

Edited 11/21/2019 22:59:56

To address one minor detail: the extra points you get for completing a game are not taken into account when calculating the ratings. All the ratings are calculated with your actual BayesELO rating and the points for completing a game are added afterwards on Fizzer's side. Given the matchups people are receiving at the top end of the ladder (or at least the matchup cases i've seen), I'd say it's very likely that these extra points are also not considered when giving people new matchups.

i just had a discussion on discord and wanted to know, if your rating is affected and how much it is affected, if you got an opponent who's playing some games until he loses one and then surrender his other games and leaves the seasonal ladder.

This has no definite answer. Since the player gets his initial matchups assigned randomly (i think), they can be against anyone. If he got matched against what is going to be the top 5 of the season and he beats them all, then leaves after losing to someone, he will have an extremely high rating. If he wins against the bottom 5 of the season and then loses a game, he'll have a way lower rating. Since his rating is what influences yours, the more significant rating he had, the more significant impact it's going to have on you.

I always thought it would have a significant effect on my rating. Then i heard a long time ago, that it shouldn't be the case or if, than just to a very minor extend, that you can almost ignore it.

So what's the truth?

So in conclusion, both are the truth, given specific cases.

Beren, mod and Farah explain it correctly.

BayesELO calculated the ratings solely based on the finished games and does NOT include the rating modifier of +65/game that Fizzer adds when displaying the rating on the website (note: in the Heyheuhei season, this already existed, but was only 40, it was changed to 65 after that season as a reaction to it).

I can also confirm that the matchmaking algorithm does not take this 65/game bonus into account either.

So, in short: the value of a win/loss against a player who leaves early is relative to the results and underlying BayesELO rating of that player's games, no matter how many or few those were, as summarized above.


In addition to the answers above, do note that BayesElo also calculates standard deviations (a measure of how precise the rating estimation is).

If I remember correctly, the "weight" that a given game has on your rating depends on this - winning or losing against a player that has a low standard devation (very precise rating estimation) impacts your rating more as this information is seen as more informative. Meanwhile, a game against a player who is 3-1, and hence doesn't have a very precise estimate of his rating yet, has less impact on your rating.

This is done through a complicated iterative procedure, so it's difficult to say how much less impact a game against such a player has compared to your regular games. But rest assured, it will by definition be all but guaranteed to impact your rating less than your games against the players who play all 20 games.