Handicap ignored in rated games' rank calculation?

If I won/lost a 9x9 auto handicap game, that seems to be counted as I played the game against the opponent without handicap.

E.g. check my last 2 games: https://online-go.com/user/view/277037/rarandomra and there effect on the rating graph.

@zellerfa, Can you explain more about what you expect to see instead of your current graph?

A developer can correct me if I’m wrong, but my understanding is that auto handicap stones are designed to make games essentially even. For example, a 14k playing against a 20k on a 9x9 with 2 handicap stones is treated similar to an 14k vs a 14k.

It probably gets slightly more complex, since on 9x9s multiple ranks get the same number of handicap stones. Ex: the 20k opponent and 19k opponent each got 2 handicap stones, but their skill level is different. A loss to the 20k should hurt your ranking more, and a win help you less. This might explain some of the variation you see.

@Swabby8 In graph: won a 14k vs 20k auto handicap game, received 1 point for the even game, then lost a 14k vs 20k auto handicap game, lost 16 point for the even game. I would expected to gain and lose the same amount.

I found this:

despite the title the discussion is mainly about handicap not being default on ogs… Anyway, it seems to be a common complaint, that the rating system cant handle handicap games (only) on 9x9 boards.
I cant say why people drew that conclusion, nor do i know anything about proper handicap for smaller boardsizes. hope this helps.

I think I understand what happened now.

You’ll notice that in both the game against the 19k and the 20k, you had the same number of handicap stones and komi. However, it should be easier to win against the 20k, so they can’t both be considered even games.

If the system considers that, with your handicap settings, a 14k vs an 18k would be a truly even game, then your first game would be equivalent to a win vs. a 16k, and your second a loss against a 15k. Wins against weaker players don’t give many points since they simply confirm your rank is accurate, while losses to weaker players hurt your rank significantly since they suggest it is too high, explaining the point difference.

Hard to explain, hopefully that makes sense… my advice is just to avoid handicap on 9x9s. I don’t think the small size lends itself well to the handicap system, which creates all these balancing issues

1 Like

There are two separate issues here: whether the handicap system itself works on small boards and whether the automatic handicap selection selects reasonable handicaps. In my experience, handicaps work ok on 9x9, but it takes some trial and error to find the right settings for a given pair of players. As for automatic handicap selection, well, the ranking system tries to do this, but one has to have realistic expectations about how successful it can be. :slight_smile: We have a bunch of ad hoc heuristics on top of Elo system adjusted to emulate Japanese dan/kyu system, which in turn was always too ambitious, trying to govern everything from handicap selection to castle politics. And this is applied to a wide pool of amateur players who are not required to participate in any qualifying games. How accurate this can be? :slight_smile:

1 Like