How well does Handicap apply to 9x9?

Not that long ago I was challenged to a 9x9 game by a new person on the site and surprised to find myself facing a handicap. (Obviously, I should have read the challenge better!).

At the time I thought “do people really play 9x9 with handicap?”.

This topic has come up again in a recent thread about handicaps and TPK play, where I guess 9x9 is the most common size.

Do people really use handicap much in 9x9 play?

Do we expect 1 stone of handicap on a 9x9 board to equal one stone of rank?

2 Likes

Yes and no. I agree with your implied point. I doubt that handicapping makes much sense in 9x9, generally speaking, and we certainly have no reliable way to gauge its effect (and as I have stated many times, I believe that komi and handicap stones are both really sliding scales in their significance, because a strong player can make more use of them than a weak player).

On the other hand, handicap stones may be useful for absolute beginners. When I began playing again with my local group—after 40 years away from the game—I was given 5 stones in my first game (of 9x9, which I had never even heard of 40+ years ago). I lost, making all the possible basic mistakes in that one game. We have introduced a couple of kids (early elementary school, maybe 7 or 8 years old) to go at the club, and similarly used 5 stones. Nothing scientific in this. Just a guess.

2 Likes

You just gave us a great idea for a tournament here.

As far as I’m aware no one is playing handicap on 9x9 except for the handicap of 1 because some feel komi should be less.

I played some auto handi games in the past and just now. Naturally, 9x9 handicap should be less in actual stones on the board [senseis]. On OGS I’d say it’s not unreasonable. I’m not sure how handicap calculation works but against 25k it gave them only 6 stones. But I feel like if your opponent is ~10k and you have to give them 2-3 stones it’s too hard because at that point players become too smart.

Some bragging

https://online-go.com/game/12766879
https://online-go.com/game/12802307

I have been in a few 9x9 handicap tournaments, and what I noticed was that whenever my opponent starts with 2 stones down, I lose. I usually win in the games I start with 3.5 komi and my opponent doesn’t have any stones. It is definitely something that will happen in beginners tournaments and I now have to look out for that before I join so I don’t have to play matches like that again. Of course most of them are multi rounds, and I went up a few ranks recently, so if I have to give 3 stones to someone for the first time😶not looking forward to it.

1 Like

I play 9x9 with handicap a lot. Works perfectly fine. And why wouldn’t it?

How the handicaps stones on 9x9 should relate to the ratings and to 19x19 handicap stones is a different question. But so far (admittedly only few games that led to reduced komi, no real handicap) what I saw here on OGS was in the right range.

4 Likes

I guess it depends what you mean by “works”.

Clearly placing additional stones for weaker players will give them assistance.

I’ve realised that my question is specifically whether 1 stone in 9x9 is worth 1 rank.

I don’t see how it can be, because 1 stone is worth 1 rank in 19x19, and yet 1 stone seems to me to be much more powerful in 9x9…

Or is this already taken into account in the matching and handicap calculation - do you get less stones of handicap per rank in 9x9?

2 Likes

Of course! Yes, here on OGS you get less than 1 stone per rank.

I (around 8k) have given total beginners up to 6 stones on 9x9 in their first game. For opponents around 25k-30k 3 stones seem to work well enough to get close games for me.

2 Likes

One stone in 9x9 is obviously far more than one rank, if you’re talking about actual practical play. Think about the reasons for why players would be different ranks. :slight_smile:

For a well-calibrated ranking system, when a correctly-ranked 9k plays a correctly-ranked 10k on a 19x19 board the reason why the 9k can have an even game when playing one stone down (e.g. reverse komi, or a 2H game but where white gets 7.5 komi - both of these are 1 stone total of disadvantage) is not because there is an intrinsic magic “one stone betterness” about the 9k compared to the 10k.

Rather, it’s because on average, across many games and many different situations, the 9k makes slightly fewer or slightly less bad mistakes. Since Go is a pretty long game and even pros make dozens of mistakes per game, there’s lots of time and lots of separate mistakes to add up, so it turns out actually that a not too bad model is to say that players randomly may lose a certain amount of advantage each time they have to make a decision - i.e. each move, which adds up to a certain average amount per move.

Then, what determines the 1 stone difference is simply that the average amount that a 9k loses per move, summed across the whole game, is about 1 stone less than the average amount that a 10k loses per move, summed across the whole game.

A 9x9 game lasts for only about a quarter of the length, so that’s only a quarter of the moves to add up, so the average might only be a quarter of the 19x19 average. Therefore on 9x9, with this super crude baseline we might predict 1 stone to be about 4 to 5 ranks difference. And on 13x13, we might predict 1 stone to be about 2 ranks difference.

This model is horribly wrong in many ways. But it’s still far more accurate than saying that 1 rank is 1 stone regardless of board size. And it’s not entirely off the mark. If you play with 4H on 13x13, it does feels pretty close to a 8H or 9H game on 19x19. On 9x9 it breaks down a bit more due to the very different nature of that size and different styles of play and so on, but depending on the players, in practice 1 stone per 4 to 6 ranks is about right.

5 Likes

Thank you for posting this. I’m pleasantly surprised that it matches fairly well with the handicap tables that I use in my children’s club: http://goratings.eu/content/voorgift-tabel.v6.pdf

It has 3 stones handicap between 8k and 26k-21k and it has 6 stones handicap between 8k and total novice (42k in my table).

It is based on one handicap stone per 6 ranks difference.

In more detail (without komi and substracting half a stone correction to compensate for that): handicap = round(1 + rankdiff / 6 - 0.5).

1 Like