New way of deciding Komi

The rules (japanese rules to be precise)

Honestly I do not really see any added value to the game by doing this. The komi is designed to create an equal challenge for both players. I think deciding players’ color by random is much more fair, simpler and quicker than this. But hey, if you like it, maybe others will as well.

4 Likes

The komi is designed to create an equal challenge for both players.

exactly. how can one determine if 6.5 is equal for both players? Just like in economic, the only “fair” price is when both party are free to negotiate, the only komi that create an “equal” challenge for both is when they can freely bid.

Now I’m just a newbie so I may well be wrong, but isn’t this problem exactly what the handicap system with extra stones are there to address? I see what you mean that the Komi might not be right for everyone, but as there is already a compensating system in place to correct for different skills, why would you need an extra system? Wouldn’t that make it unnecessarily complicated, where 6,5 Komi and no handicap stones may be equal to 2,5 Komi and 4 stones, or 0,5 Komi and 6 stones, and so on… It seems to me that the usefulness of handicap stones is partly because it’s so simple and straightforward, but it rests on the foundation everything else is equal.

Or am I missing something fundamental?

2 Likes

i think the handicap system and komi is 2 quite different concept.
the handicap system is used for bringing together players with DIFFERENT level of skills.
the komi is used for compensating players with SAME level of skill for playing white.

Yes, that is my understanding as well. And I know that different rule systems and associations sets the komi at different levels, which can change over time. But isn’t the point with komi that it is the general consensus of how large the handicap is for playing white? You can’t really claim that a different komi is required to compensate for playing white when the players have the same skill level, because if that is needed they simply aren’t at the same level.

I think I’ll use an example to see if I understand what you are saying:
Player A, 10 kuy, needs a komi of 4,5 when playing white to have an average chance of beating player B, 10 kuy, with no handicap stones.

When A plays white against C, also 10 kuy, he needs a komi of 6,5 to have an average chance of beating C without handicap stones.

My understanding of the system is that in the above example the players are at the wrong ranks, rather than that the komi would need to be changed. If you make komi changeable you also throws out the handicap system, as it rests on the komi at a fundamental level.

As I’ve said I’m a newbie so I may be completely off the reservation here, but so far this is how I understand the situation.

2 Likes

DGS implements something like that, although I never used this feature myself.

Non-integer komi serves as a tiebreaker. White gets favored.

Handicap is used to roughly level the playing field. The handicap = 1 means you play black and get to move first, and white wins if it’s a tie. Depending on rules the point value of first move is 6 or 7 points.

As I understand, the value of komi is essentially a statistically determined value that would result in about a 50:50 balance of wins:losses in games between evenly matched opponents. I don’t see how that is not viewed as fine-tuning the handicap.

This leaves me thinking a good metaphor for handicap and komi is getting the wheels on your car balanced.

Article about Komi http://senseis.xmp.net/?Komi

1 Like

why do people keep trying to add absurd complexities to the beautiful simplicity of go…

3 Likes

Playing games vs. playing the game?

1 Like

Exactly! I just had to use lots more words to try and say the same.

oh i am sure loads of people ask the EXACT same question 80 years ago when they first introduce the Komi rule.

1 Like

you have stated my concern: a statistically determined value. But not everyone is “average”. why shouldn’t one player be able to determine the komi depending on his own style of play?

Your interesting proposal addresses an issue that no one in this thread seems to have expressed with any clarity (or perhaps I have overlooked it–if so I beg forgiveness). I believe komi was determined on the basis of pro games. I seriously doubt that it has much validity for DDK games–and less and less the closer one approaches absolute beginner level. I know for sure that the first move is NOT worth 6.5 points to me, because I’m not good enough to exploit the advantage. Ideally, there is probably a sliding scale of komi that corresponds to one’s rank. It would be interesting to learn just where the dividing line is.

1 Like

If he is willing to play only with himself, none at all.

I you think about it, the system of rank, handicaps and komi is a measurement system. When the original standard meter bar was made, it was of platinum not mercury.

The komi gambling system is something that exists within a standardized system; otherwise you’d have little basis for well-informed betting.

Would you have an objection to using a statistically determined value of “normal” body temperature if your child is younger than 3 months and has a “fever” (defined as oral temperature of >= 100 F) Mayo Clinic says seek medical help. Would you engage in a fever auction? No one’s kid is just “average”.

Actually, nobody is average. What that has to do with komi, though?
Komi is one of the ways to compensate for the first side advantage. The advantage is measurable: there must exist an optimal play in go, so we just look by how much points black wins when both sides are playing optimally. As of yet, we don’t know what optimal play on 19x19 is, and so we don’t know how to measure black’s advantage precisely. But statistics suggest that advantage corresponds to either 6 or 7 points of komi (for territorial scoring). IIRC, the largest board with known optimal play is 7x7 and the correct komi for it is found to be 9.

If we have two random players playing each other, it is fair to use that semi-theoretical value for komi. If one or, more likely, both players happen to play below optimal, well, it’s their problem, the game is still fair.

Now, if we want to take their skills into account and give each one an equal chance of winning, we’re talking a handicap. There is a lot of ways of determining one, altering komi by holding an auction is one of them, although it doesn’t strike me as precise (what one’s bids are supposed to be based on? one’s arrogance? ).

4 Likes

The komi negotiation would add more clutter and more complexity to the site with only debatable benefit. If there is a need to game-settings negotiation, I suggest all parameters to be negotiable (time settings, board size, maybe even color and game name).

In the meantime, you can open challenges with custom komi.

3 Likes

Komi betting has been used in Batoo and Go variants because the exact value of equal handicap for the first move has not been properly researched yet for those sizes. For the standard 19x19 however, it has been researched.

Why must it be 6.5? From trial and error from many decades (at least) of games and researching that B had a lower win percentage than W with 5.5. While one might argue that pros not making perfect play might have made the data inaccurate, it’s about as accurate as you can get using humanly possible means. Even your betting is not objectively fairer, it’s fair to the players of the game but it’s not objectively fair universally.

The best AI play now, even if not perfect (not like we mere humans would know), is godly perfect to us as we can possibly comprehend and could possibly change the current Komi or prove what we are using is right all along.

Your suggestion is intriguing and could add a little twist for Go just like One Colour Go but ultimately does not aid in making it any fairer.

“Advantage” is an abstraction, which is not measurable. What is measured is the exploitation of the advantage. This is not hair-splitting; it is the basis of my previous post. That exploitation reflects the quality of play at the level of the population used for the measurement. Hence, the level or quality of play is baked into the number. The idea that one number is statistically valid and reliable for all levels is not logical.

Now don’t get me wrong. I am comfortable with things as they are. I don’t support the proposal, though I think it is interesting. As others have said, I like the simplicity of the status quo.

1 Like

I you like simplicity, you ought to appreciate the simplicity of one number that works so well across skill levels for so many for so long without a worldwide revolt.

1 Like

Good idea, but Auction Komi has been around for a while (and some actually use it!) I think it’s a bit simpler than a system of calling and folding. There have actually been several similar proposals. See Sensei’s Library.

(By the way, you can always check out any new idea in Sensei’s Library. If someone has already thought of it, you’ll probably find it there!)

2 Likes