Life In 19x19

Shawn Ligocki wrote:

skydyr wrote:Komi bidding, however, does seem rather pointless when I suspect the majority of the games played at the tournament are won by margins greater than people would be bidding over.

Isn't that the whole point of komi bidding? If you think that having komi < 10 won't affect the outcome of the game, but you would rather play black, then you would want to bid up to at least komi of 10. If your opponent thought that that komi 10 would affect the outcome more than going first, then you both are happy with the result. Seems sort of like a new Joseki for Nigiri.

Of course, a lot of this depends upon how it's done. I think komi bidding would be fun to do before a game, but if people were being really pedantic about the bidding rules or trying to game them in some way, that could be really annoying. The same is true if someone tries to be really annoying about normal Go rules (like agreeing on which groups are dead, etc.). It seems like most of this comes down to playing with the nice people

Well, if you want to do it on your own, go ahead. I encourage you to play however you think is fun.

However, in a tournament, you have to be able to maintain the integrity of the results. How do you take into account different komi when it comes to calculating a rating, or a result? Is white's win with 10 komi worth less than white's win with 5 komi when it comes to determining winners?

What if one player took black, or white, in every game? How is this compared to someone who took an even mix?

Or if a player won all their games, but their opponents won the auction with abnormally high or low komi values?

What happens if both players want one colour at the same specific komi, but neither want it any higher than that?

What if players disagree on what komi was set to at the end of the game?

What if players collude to manipulate komi so that one player gets a better result in the tournament? How could you stop this, and would it be different from players colluding to fix a game on the board?

What if they pick a whole number komi and get jigo?

And perhaps most important of all, what is the gain to the tournament by having komi set this way instead of by fiat?

With komi bidding (or pie rule) games should be considered totally even for rating purposes. That's the whole point of bidding/split rules

RBerenguel wrote:With komi bidding (or pie rule) games should be considered totally even for rating purposes. That's the whole point of bidding/split rules

But for rating purposes, you are now comparing some games where the result reflects PGS (player go skill) with games where the result reflects PGS + PKBS(player komi bidding skill). I think it's fairly clear from a logical perspective that PGS != PGS + PKBS unless PKBS is meaningless. If it's meaningless, why bother? And if it isn't, you're comparing apples to oranges.

skydyr wrote:

RBerenguel wrote:With komi bidding (or pie rule) games should be considered totally even for rating purposes. That's the whole point of bidding/split rules

But for rating purposes, you are now comparing some games where the result reflects PGS (player go skill) with games where the result reflects PGS + PKBS(player komi bidding skill). I think it's fairly clear from a logical perspective that PGS != PGS + PKBS unless PKBS is meaningless. If it's meaningless, why bother? And if it isn't, you're comparing apples to oranges.

Well, the bidding skill is supposedly related to playing strength. High correlation even if low impact. And it's expected in the ideal case to make for an even game, thus no need to worry about who plays first. If it doesn't, it's as harmful (shouldn't be worse, at least) than just being white when slightly weaker

John was probably talking about the Maidenhead Tournament which has komi bidding (though it never took half an hour to explain it when I attended). The results are entered into the European Go Database as normal games I believe (picking a random year ):

http://www.europeangodatabase.eu/EGD/To ... n=14337147

I don't recall myself having such a close game the komi bidding made a difference, but there are usually a couple in the tournament.

RBerenguel wrote:Well, the bidding skill is supposedly related to playing strength. High correlation even if low impact. And it's expected in the ideal case to make for an even game, thus no need to worry about who plays first. If it doesn't, it's as harmful (shouldn't be worse, at least) than just being white when slightly weaker

How can the correlation be high? Excepting some ddk players, don't we all know the proper Komi?

skydyr wrote:However, in a tournament, you have to be able to maintain the integrity of the results. How do you take into account different komi when it comes to calculating a rating, or a result? Is white's win with 10 komi worth less than white's win with 5 komi when it comes to determining winners?

What if one player took black, or white, in every game? How is this compared to someone who took an even mix?

Or if a player won all their games, but their opponents won the auction with abnormally high or low komi values?

Tie breakers are always confusing, I suppose that they could be done in many ways, but I'd expect a win to be a win regardless of what komi was. (For example, tie breakers in normal tournaments do not consider point difference.)

What happens if both players want one colour at the same specific komi, but neither want it any higher than that?

Then you can do normal nigiri. (Don't know if that's how tournaments do it, but seems like how I'd do it).

What if players disagree on what komi was set to at the end of the game?

What if you disagree on moves or number of captures at the end of the game? Go depends upon some level of agreement between players. If you think you will forget, write it down somewhere. If you think your opponent will debate it, get them to sign that paper. If this is a really serious game, there should be a record keeper, etc.

What if players collude to manipulate komi so that one player gets a better result in the tournament? How could you stop this, and would it be different from players colluding to fix a game on the board?

Seems the same as players colluding in a normal tournament by intentionally losing a game. If you are worried about this, I think your problem isn't the rules format, it's the rotten players you're playing with

What if they pick a whole number komi and get jigo?

I'd expect you require all komi values to be X.5

And perhaps most important of all, what is the gain to the tournament by having komi set this way instead of by fiat?

I've never done it and have no idea if it improves the experience, but it makes sense to me. If both players want to play a normal game, they agree to 6.5/7.5 komi and nigiri. But if someone thinks that the first move is worth more to them, they can take randomness out of the equation and bid until they really feel equal playing as either Black or White.

Shawn Ligocki wrote:

skydyr wrote:And perhaps most important of all, what is the gain to the tournament by having komi set this way instead of by fiat?

I've never done it and have no idea if it improves the experience, but it makes sense to me. If both players want to play a normal game, they agree to 6.5/7.5 komi and nigiri. But if someone thinks that the first move is worth more to them, they can take randomness out of the equation and bid until they really feel equal playing as either Black or White.

For example, consider this passage from Kageyama from "Lessons in the Fundamentals of Go":

I could not match the Meijin in either skill, stamina, or spirit. The only place where I could compete with him on equal terms was in guessing even or odd correctly to see who took black and who white. I decided therefore, to gamble everything on this guess, and spent my days and nights feverishly working out black openings. This was my one chance. I was resigned to losing if I drew white.

Obviously, this is a bit of an extreme example, but it's worth way more for Kageyama to play Black than to get 6.5 komi in this situation. If the Meijin had also especially wanted to play Black, I think it makes more sense for them to do komi bidding than to both hope to win nigiri.

I think this is more common for us kyu players who generally don't end up with close games and thus would be willing to give a few points for the chance to go first and play our prefered opening. (Or in my case, I think I actually prefer to play White, so I would get to play White in most games when my opponent outbid me.)

hyperpape wrote:

RBerenguel wrote:Well, the bidding skill is supposedly related to playing strength. High correlation even if low impact. And it's expected in the ideal case to make for an even game, thus no need to worry about who plays first. If it doesn't, it's as harmful (shouldn't be worse, at least) than just being white when slightly weaker

How can the correlation be high? Excepting some ddk players, don't we all know the proper Komi?

I was thinking more of pie rule when writing that, actually

hyperpape wrote:

RBerenguel wrote:Well, the bidding skill is supposedly related to playing strength. High correlation even if low impact. And it's expected in the ideal case to make for an even game, thus no need to worry about who plays first. If it doesn't, it's as harmful (shouldn't be worse, at least) than just being white when slightly weaker

How can the correlation be high? Excepting some ddk players, don't we all know the proper Komi?

Proper komi in what sense? With perfect play, it seems that proper komi in the sense of komi that would lead to a tie, is 7 or maybe 9. (Maybe 11?) But proper komi in the sense of komi that makes the win rate for Black the closest to 50% is another matter. It also depends upon the skill of the players. Proper komi in that sense for DDK vs. DDK may well be 3.5.

People say that proper komi may vary a great deal, but as far as I've seen, there's never been much evidence put forth for it. Someone showed that nearly random engines did have different komi values, but that's a far lower level of play than even DDKs exhibit.

And yes, I mean proper komi for making the win rate close to 50%.

RBerenguel wrote:With komi bidding (or pie rule) games should be considered totally even for rating purposes. That's the whole point of bidding/split rules

komi is discrete, hence it wouldnt be "totally" fair. say with 6.5 komi black has slight advantage, with 7.5 komi white has slight advantage. however, if you allow fraction komis when bidding, then it should be totally fair

fractions of course affecting odds of either player being declared winner, in case of a tie on the board after integer komi.

Ootakamoku wrote:

RBerenguel wrote:With komi bidding (or pie rule) games should be considered totally even for rating purposes. That's the whole point of bidding/split rules

komi is discrete, hence it wouldnt be "totally" fair. say with 6.5 komi black has slight advantage, with 7.5 komi white has slight advantage. however, if you allow fraction komis when bidding, then it should be totally fair

fractions of course affecting odds of either player being declared winner, in case of a tie on the board after integer komi.

Hmm, interesting idea

Sensei's Library defines correct komi and fair komi. Any need to introduce perfect and proper komi?

For example, consider this passage from Kageyama from "Lessons in the Fundamentals of Go":

Quote:
I could not match the Meijin in either skill, stamina, or spirit. The only place where I could compete with him on equal terms was in guessing even or odd correctly to see who took black and who white. I decided therefore, to gamble everything on this guess, and spent my days and nights feverishly working out black openings. This was my one chance. I was resigned to losing if I drew white.

I'm not sure what date this quoted game might be from, since the concept of komi in Japan only started post-WWII.

I'm sure John Fairbairn knows exactly the details, but I don't think it was common until the 1960s, so Kageyama may be reflecting on the fact that he needed the extra benefit of being Black to beat the Meijin.

Personally, I think that few amateur players play good enough yose and have close enough games that make komi matter.