9 stone handicap against a player of similar strenght

[Sennahoj]

Hey EdLee, I guess I'm sometimes too easily annoyed by what I consider to be overly-philosophical non-answers...

I read back and see that I wrote too sharply, and I apologise for that. No ill will intended.

Of course one can always say that the difference between any two players is unique, but I honestly don't think that's a helpful answer, it's just kind of tautological. It didn't seem like an answer to a generous reading of the question, i.e. trying to give as good meaning to the question as possible.

[EdLee]

Hi Sennahoj,

Apology accepted.
You have your frustrations; I have mine.
Betoacosta (BC) asks "Have any of you ever tried this?"
DrStraw mentioned the 140 number from pros (post 4);
Robert noted 130, from his own, actual games, at some level(s) (5d EGF?).
Everyone else's reply falls along the line: No, I haven't actually tried it; and cannot give a much better estimate than BC's "10 pt/stone".
( All estimates ranged from ~50% to ~150% of that number. )

I'm not a statistician, so I don't know what kind of sample size is needed for a good representation.
Suppose we have 10 people (of roughly equal level),
and each plays another 10 opponents (all roughly equal level),
10 games per opponent, for a total of 1,000 games.
Questions:
- Is this "enough" to find an answer to BC's question ? ( If the sample size is too small, we can increase it. )
- I imagine the higher the level (the closer to pro/AlphaGo), the more reliable. Is this true ? How true is it ?
- What would be the results for ~6dan ? For ~3dan ? ~1dan ? ~9k (~BC's level) ? ~15k ?
- What kind of distribution curves ?
- Suppose we find some number(s) for ~9k (~BC's level); how accurately would it relate to BC's own experience if he tries this experiment for himself ?
( This would depend on the quality of the 1,000 data points. )
- What level of an answer is BC looking for ? ⁽¹⁾

See also:

Of course one can always say that the difference between any two players is unique, but I honestly don't think that's a helpful answer, it's just kind of tautological.

You say 'of course' -- perhaps you have a different experience than mine.
I can only speak from my own experience: this particular understanding did not come naturally;
in fact, it took years (in another, non-Go field) and it was a profound discovery.
Maybe it's common sense to many or most people --
but I didn't make this presumption.

Based on the replies so far (aren't that many),
only Robert's (post 16) has an estimate from his own 9H games with his peers (at which level(s) ? EGF 5d? ).
We don't know how well Robert's experience correlates to BC's.
My original suggestion that BC finds out directly from actual experience doesn't seem too unreasonable. (Re: ⁽¹⁾ )

[Sennahoj]

EdLee, this is thing: it is not hard to take any question about a non-trivial subject and start generating lots and lots of related questions / sub-questions, etc. That is not providing an answer, it's a behaviour that we can call "pretending to be wise", because it can give the appearance of insight / depth, while not doing any actual work.

The most interesting of the questions you generated was "What level of an answer is BC looking for ?", and this is a good question that one should consider before giving a reply. In my judgement (and others), the right level here was "what is fair komi in 9 stone Go?".

[Bonobo]

[..] That is not providing an answer, it's a behaviour that we can call "pretending to be wise", because it can give the appearance of insight / depth, while not doing any actual work.

[..]

This I find to be a comment bordering on insult … at least “wiser than thou”.

Sometimes the wisdom of questioning the question won't be recognized because we insist on quick and easy answers.

Yes, there always exist quick, easy, and wrong solutions to any problem.

To me, the replies given by EdLee and others here seem to suggest that the question may be just too imprecise to give a precise answer (actually the setting for the question seems to be too foggy, IMO).

[Uberdude]

But the OP did not ask for a precise answer. In fact he didn't even ask for a truth, just an opinion "If you play against someone your own rank, how many points do you think you will need to make the game "even"? " Thus I think "About 120 points" is a perfectly good answer. I presume the OP is intelligent enough to interpret this sensibly, and not crucify me if he finds one pairing of players for whom the correct komi in a 9h is not 120 points. "120 points with a standard deviation of 30 points for 1kyus and 60 points for 10 kyus and 140 points with a standard deviation of 5 points for 9ps" is good too (though I'm just pulling numbers out of my backside but the principle of higher variation for lower ranks is probably true). It is possible to give helpful, imprecise answers with caveats (either explicitly stated or presumed in the intelligence of the reader).

[Uberdude]

1 stone is worth twice komi (easy to see if you consider black passes for his first move, it goes from black first and white get komi to white first and white gets komi; to balance this back to the original fair position but with colours reversed white would need to lose his komi, and then give another komi to black, a difference of twice komi; so a pass / 1 stone is worth twice komi) so that'd put it more like about 13 points per stone.

hm.. but playing with "one stone handicap" usually means black with no komi right?

Right, a traditional "one stone handicap" is only half a stone advantage, and a "two stone handicap" is only 1.5 stones. So if a 4.0 dan plays against a 2.0 dan (in say KGS, you can see fractional rating on the graph) in a 2 stone game the 4 dan has a better chance of winning, it's not actually a fair game (maybe 55% or so?). I believe the KGS rating algorithm takes this into account: if this 4.0d and 2.0d played many games and the wins were 50:50 the 2d's rating would go up and the 4d's down.

[Bill Spight]

As has been indicated, stronger players are better able to make use of handicap stones than weaker players. Uberdude vs. Uberdude should probably get around 120 points komi for giving 9 stones. For 10 kyu vs. 10 kyu, who knows? My guess is to start with 70 points and then adjust the komi up or down by 4 points per game, depending on who won or lost, until you get a better idea of the right komi.

[Sneegurd]

The most interesting of the questions you generated was "What level of an answer is BC looking for ?", and this is a good question that one should consider before giving a reply. In my judgement (and others), the right level here was "what is fair komi in 9 stone Go?".

Don't you think he also should have served you a wine? Did you really get what you deserve? -> Your order was not precise enough.
In my judgement (and others): you are kidding, right?

I was absolutely out of context here, I apologize.

[Sennahoj]

Not sure I understand your what you're getting at.. I wasn't complaining about the OP, if that's what you think! My last reply was w.r.t. EdLee's hidden, maybe that was confusing

[daal]

The most interesting of the questions you generated was "What level of an answer is BC looking for ?", and this is a good question that one should consider before giving a reply. In my judgement (and others), the right level here was "what is fair komi in 9 stone Go?".

If I am asked a question, it is up to me to decide how best to answer it. If in my opinion, the question has a false premise, I see nothing wrong with disregarding the intent of the question until the premise has been cleared up. In this case, the question was based on the premise that in a 9 handicap game, it is possible to determine a generally valid number for the value of komi. My impression is that you consider the premise unproblematic, whereas EdLee considers it to be an essential misconception. Just because other people agree with you (and not everyone does - Bill Spight for example also pointed out that the value can differ depending on the strength of the players) does not mean that other interpretations of the question are not valid. It seems that you consider EdLee's suggestion to have been impractical and thus not helpful, and yet Dr. Straw mentioned the use of a kadoban in earlier times to determine the strength difference between individuals. As much as you may prefer a straight answer to a roundabout one, the intent of the OP is not the only deciding factor as to which is best.

[Sennahoj]

If I am asked a question, it is up to me to decide how best to answer it. If in my opinion, the question has a false premise, I see nothing wrong with disregarding the intent of the question until the premise has been cleared up. In this case, the question was based on the premise that in a 9 handicap game, it is possible to determine a generally valid number for the value of komi. My impression is that you consider the premise unproblematic, whereas EdLee considers it to be an essential misconception. Just because other people agree with you (and not everyone does - Bill Spight for example also pointed out that the value can differ depending on the strength of the players) does not mean that other interpretations of the question are not valid. It seems that you consider EdLee's suggestion to have been impractical and thus not helpful, and yet Dr. Straw mentioned the use of a kadoban in earlier times to determine the strength difference between individuals. As much as you may prefer a straight answer to a roundabout one, the intent of the OP is not the only deciding factor as to which is best.

Ok, so there is a question whether the concept of komi in 9 handicap game is is valid. Yes, to me this is obvious (or at least: as obvious as the same question about komi in a normal game of Go) --- fair komi is a property of the game, not of the players.

The other thing: I think that EdLee has a general tendency to give "mysterious" answers to various questions --- not only in this thread, but generally. By "mysterious" I mean answers that amount to tautologies, or something like "this question cannot be answered, because Go is so complicated", or this expanding an original question into many many new subquestions (like in that hidden post). Not all the time of course, but not very seldom either, and I decided to "call him out" on this habit a little bit. In retrospect, this seems like a very bad idea, since it appears I came across as attacking him. I'm sorry for that and it was not my intention.

[Bonobo]

Sennahoj, you end our text with this:

[..] it appears I came across as attacking him. I'm sorry for that and it was not my intention.

But right in the beginning of that same paragraph you are doing it again, in my perception:

The other thing: I think that EdLee has a general tendency to give "mysterious" answers to various questions --- not only in this thread, but generally. By "mysterious" I mean answers that amount to tautologies, or something like "this question cannot be answered, because Go is so complicated", or this expanding an original question into many many new subquestions (like in that hidden post). Not all the time of course, but not very seldom either, and I decided to "call him out" on this habit a little bit. In retrospect, this seems like a very bad idea, since it appears I came across as attacking him. I'm sorry for that and it was not my intention.

TBH, to me this looks like intentionally slapping somebody in the face and then saying “oh, it appears that I came across as hitting you, I’m sorry.”

Reminds me of that religious leader who once insulted the followership of another religion. Later he said “I am sorry that I was misinterpreted”, which most media—idiotically, but actually biased—interpreted as an apology while with that comment he had actually expressed how stupid he thought his audience was for “misinterpreting” him.

[daal]

But right in the beginning of that same paragraph you are doing it again, in my perception

I personally don't see it that way. I think Sennahoj is just saying what it is about some of Ed's posts that raises his hackles.

The other thing: I think that EdLee has a general tendency to give "mysterious" answers to various questions --- not only in this thread, but generally.

I think you are wrong. I think Ed's tendency is to reject questions that request shortcuts, and instead point to the necessity of examining specifics, which he does time and time again in the countless game reviews and answers to specific questions that he offers (rather helpful, I might add).

Ok, so there is a question whether the concept of komi in 9 handicap game is is valid. Yes, to me this is obvious (or at least: as obvious as the same question about komi in a normal game of Go) --- fair komi is a property of the game, not of the players.

Not just if it is valid, but rather if a single number is adequate. While on the one hand, regarding komi in an even game, we all seem to accept that what is good enough for the pros is good enough for us - but when have we seen a pro play a 9 handicap game? Would every pro choose the same value? Would it not depend on their estimation of their opponent? What about amateurs - can't stronger players use handicap stones better than weaker players, and shouldn't this affect komi? BTW, I ask these questions neither to avoid doing the work nor to appear wise, but rather to point out that I see your answer as inadequate. You may be right that this is not a discussion that the OP was looking for (appears not) but also does not appear to be one that everyone else finds uninteresting.

[wineandgolover]

I decided to "call him out" on this habit a little bit. In retrospect, this seems like a very bad idea, since it appears I came across as attacking him. I'm sorry for that and it was not my intention.

Yeah, that is how it looked. Ed is a long-time contributor to this board. Why would you feel the need to call him out at all? I don't think you can really expect the rest of the world to adjust to your behavioral desires.

[Sennahoj]

Ok, so there is a question whether the concept of komi in 9 handicap game is is valid. Yes, to me this is obvious (or at least: as obvious as the same question about komi in a normal game of Go) --- fair komi is a property of the game, not of the players.

Not just if it is valid, but rather if a single number is adequate. While on the one hand, regarding komi in an even game, we all seem to accept that what is good enough for the pros is good enough for us - but when have we seen a pro play a 9 handicap game? Would every pro choose the same value? Would it not depend on their estimation of their opponent? What about amateurs - can't stronger players use handicap stones better than weaker players, and shouldn't this affect komi?

I think there is some confusion that comes from "overloading" the term komi a little bit, so let me try to clear this up.

By komi (K) I mean: the number of points given to white which gives a tie in a perfectly played game. This is a single integer number, and it is well defined because the game tree is finite (so perfect play exists). It is equally well defined for Go with or without any extra handicap stones.

Then there is another sense in which one could use "komi" --- what number K_AB should players A and B use in their game to give ~50% win rates, given that A and B have the same rank? The number K_AB can of course vary around K, because A and B can have the same rank (i.e. same average win rate against some other population of players), while at the same time A struggles when playing with B. Like Bill and Uberdude said, that variance will likely be larger for weaker players. Asking for the size of K_AB is like asking "what hadicap should A and B play with?" (because using any K_AB different from K is a de facto handicap), and of course here one must use trial and error.