How Automatic Handicap Works on OGS

Hey guys! We are making a handicap tournament and are using OGS automatic handicap tool. I would like to clarify how handi works. I’ve run tests before making tournament rules and it looked like the formula for handi stones on the board is difference in ranks minus one.

Then actual tournament games started, and weird system behavior started to appear. For example, why in game between 2k and 1d it’s just no komi, though it is expected to be 1 handi stone on the boad and no komi, I am confused and would like to figure it out.

Game for reference: Beginning of Magic Tournament Round #1

1 Like

I’ve always noticed something is strange in handicap tournaments. One thing I suspected is that maybe the rankings at the time of signing up for the tournament are used instead of the current ones. But no idea if this is true.

Edit: Ah, it was not an OGS tournament game, which means my theory might be correct or not, but is definitely unrelated to what happened in that game.

1 Like

I’m not expert but there are rounding effects to take into account. E.g. 4 and a bit k versus not quite 2k might get no Komi instead of 2 stones or something.

And then I wonder if there is an issue across the kyu/Dan divide. Maybe the 0k/0d point messes with the handicap calculation somehow?

But as I said I’m not the expert on how this works.

This was my first thought too, but in this case even with rounding players are closer to their stated ranks and nothing else: 2k (1.7k) and 1d (1.2d) :slightly_smiling_face: :upside_down_face: :slightly_smiling_face:

But actually thinking about this no Komi is the same as one handicap stone on the board and no Komi.


Difference is 1.9 ranks. Maybe floor() instead of round() is used? I could even see how one could argue for that being right.

It’s based on the full decimal ratings at the beginning of the match. Your 1.7k and 1.2d probably have a difference of only 1.9 that gets rounded down to 1.

This page is my understanding of how it works:

Hm, this might be correct. Then it would not care about what rank is shown in the profile, only consider a digits before “.”! Then it would see players as 1k and 1d and indeed autohandi them to just 0,5…

Oh, this is slightly unfair then :smiling_face_with_tear:

Define ‘fairness’! :slight_smile:

I don’t have a strong opinion myself, but you could say getting a handicap stone if the rank difference is less than 1 would be unfair. It seems the code follows this logic.

I think the rounding logics should be followed, and this game should be considered a game of players with 2 ranks difference. It might be not critical on kuy level, but the higher ranks are, the more sensitive handi flows like that would be.

With rounding of handicap to full stones (as is the tradition), there is bound to be some unfairness for either black or white.

This unfairness could be reduced by using intermediate handicaps by adding a number of points komi to make the expected winning probility as close to 50% as possible, with a higher resolution than you get with only full handicap stones.

Or it can be compensated for by having the rating system account for any unfairness in the handicap by adjusting the expected winning probability. For example in this case it could compute that black has only 40% expected winning probability, so if black loses the game he loses fewer rating points than if he lost a game where has 50% expected winning probability, and if black wins the game he gains more rating points. This may even already be the case on OGS.

1 Like

“1 handicap” = no komi.

Those are the same.

0 handicap is an even game, with regular komi.
1 handicap is a game with no komi.
2 handicap is a game with two stones on the board and no komi (or a 1.5 komi in Chinese rules).


My question was why in the game with 2 rank difference there were just no komi. According to your description, it should have been 2 handicap.

But it seems like guys figured it out above. Thank you for your input!

Ah! Well, you answered that question yourself:

Between a 2k and a 1d the difference is 2. Thus the difference minus one is 1.

1 Like

You can’t tell just from the rounded ranks. If it had instead been 1.7k vs. 1.7d, which is also “2k and a 1d”, that would have been a difference of 2.4, for a 2-stone handicap.

Yet this pages states the handicap is “equal to rank difference”, not “rank difference minus one”.

Also, the komi table is false in the case of Chinese rules. There should be one point of komi per handicap stone after the first one.

Unless it has changed recently, I think that OGS lists the points for handicap stones, under Chinese rules, as a line item separate from the komi.

1 Like

On that page, “rank difference” means specifically the rounded-down rank difference computed as described in the first section. Do you think I should make that more clear?

Maybe someone should add a note about handicap compensation. As @Conrad_Melville said above, I believe that OGS does not adjust the displayed komi for Chinese handicap games, but instead considers that as an extra item to be added during scoring. That approach seems especially appropriate for AGA scoring, since otherwise which komi would we display? You are supposed to be able to use either method for counting.

1 Like

This is very good news. My rank is only 10k, so I have been among the lowest of the players in two recent automatic 9x9 tournaments. Each time I lost fairly quickly, in one case to a 3d player who kindly allowed one of my groups to live.

The lack of handicap in a tournament is quite frustrating. I know the goal is to pick one grand winner, and that is a worthy goal. Until you consider that it means everyone else has to lose. I’ve commented in the tournament chat that a handicap would be appreciated, so everyone has a chance of winning. I got scorn as my reply. But what is wrong with a tournament whose goal is to allow everyone a chance? The winner, ideally, is just the person who did the best that day, regardless of their rank.

One way of determining handicaps is to initially set them to some reasonable value, then refine them as needed to make sure that everyone’s chance of winning is about equal, from 25k to 7d. While that is for sure possible in 19x19 games, it might require some rather special handicap and/or komi modifications to achieve that goal for 9x9 or 13x13 tournaments, since fractional handicaps cannot be set on a board.

1 Like