Is the max handicap still only +9? That is not enough for a DDK
And sounds like for 9x9 it might be capped at 2 (one actual stone) now?
Having spent a lot of time reverse-engineering the previous algorithm, Iād love to hear more details about this bug!
I played a 4 handicap stone -2 reverse komi 9x9 game yesterday (OGS said it was equivalent to 19 stones
Are you talking about
That is a new display I think, and if I had to guess the first number is the handicap as you said, 4 stones, and the second number is the rank difference between the players 22kyu vs 3 kyu approx.
The komi is +5 normal komi (which seems like what would happen if the table kind of cycled in the OP),
As in If we imagine the table went (I donāt know for sure) as follows
Rank Diff | Stones | New Zealand |
---|---|---|
0 | 0 | 7 |
1 | 1 | 5 |
2 | 1 | 3 |
3 | 1 | 1 |
4 | 1 | -1 |
5 | 1 | -3 |
6 | 2 | 7 |
7 | 2 | 5 |
8 | 2 | 3 |
9 | 2 | 1 |
10 | 2? | -1? |
11 | 2? | -3? |
12 | 3? | 7? |
13 | 3? | 5? |
14 | 3? | 3? |
15 | 3? | 1? |
16 | 3? | -1? |
17 | 3? | -3? |
18 | 4? | 7? |
19 | 4? | 5? |
As in without the question marks itās the OP 9x9 table, and then guessing the pattern afterward
In Samrakuās game the opponent did get to place 4 stones.
Draws supported with integer komi! Great news!
As a small token of my gratitude Iāve now subscribed as a āsupporterā
But it seems strange to give White the advantage, when you set color to automatic the stronger player gets whiteā¦
Regards
Iād agree that in even games, colors should be assigned randomly instead of assigning white to the stronger player.
There is actually a ārandomā setting which does this when you search for a gameā¦ I just think that for the default setting (which is āautomaticā) the stronger player shouldnt get an additional automatic advantage
Yep. I mentioned my game as a counterexample to @Feijoa saying that the max might be 2 stones. I say -2 reverse komi because the 7.0 komi is still there, you add the -2 reverse komi to get the 5 that is displayed
Correct.
Incorrect.
A maximum handicap is (still) applied, just like it always was. Itās only rank restrictions for challenges that have been removed (for now).
The result is that ranked challenges now have the same behaviour as handicap tournaments:
- For 19x19, number of stones is capped at 9
- For 13x13 and 9x9, no limit on the number of stones
Why?
- For 19x19, we didnāt remove the cap (yet) because the advantage of
>9
handicaps isnāt well-understood, and removing the cap could hurt the integrity of the ratings system. - For small boards, the cat was already out of the bag.
Part of what weāre working on in the rating system is adding Ā± error bars for modelling the advantage provided by handicaps (larger error bars for less-understood handicaps, such as >9
on 19x19, or reverse komi, or really any handicap on small boards). Once thatās in place, itāll be safe to relax the maximum handicap for 19x19ā¦ I imagine we can move it to a configuration on the tournament/automatch/challenge instead of being hardcoded to 9 on the server side.
It was two bugs: komi was calculated based on number of stones (instead of rank difference), and the ratings system adjusted after the game result based on number of stones (instead of rank difference). A few more details in the proposal thread.
Correct. Since colours are assigned randomly in even games (unless the challenge manually specifies who is white), then the tie-breaker advantage is also assigned randomly. Either white or black could be the stronger player.
Instead of more stones past 9, what about adding a reverse komi for each rank past it?
Seems like an interesting option to consider once we have the ratings system update in place (wonāt do it before then for the same reasons).
I accepted a 3k game. The settings was no handicap Komi automatic. I expected to be white with no Komi but I was black
āHandicap=Noneā means āeven gameā. This what it meant before the recent changes as well. In even games, the colours are chosen at random.
Or how about making the reverse komi a proper feature alongside the traditional handicaps?
A handicap stone is worth about 2 x komi so I think your proposal would be quite advantageous for white.
I would have some hesitation about encouraging large reverse komi games to be too common too quickly for a rating system. While one can approximate a handicap stone as 2x komi, because the style of the resulting game is very different compared to traditional handicap, it could be that for large values the practical difficulty for the weaker player to preserve their lead could scale a little differently. Itās not immediately obvious that the same number of āobjectiveā starting points of advantage will always compensate the same rank difference if those points are provided in a very different form, and the playing community overall doesnāt have nearly as much experience with such games.
If there is any difference, itās probably nowhere as extreme as a factor of two though - a reasonable guess probably would be 2x komi per stone, and it would be still be an interesting experiment.