Komi needs correction on 9 x 9


#1

I am tired of Komi issues on 9 x 9. Where is the rule that any board size besides regulation need to grant white 6.5 points for going second?

Let’s assume that 19 x 19 is the only board size which matters.

When Komi was introduced, it was deemed that white should get compensation for going second. On a regulation size board – 19 x 19 – a compensation of 6.5 points represents 1.8% (0.018005540166205) of the 361 available points.

With the brilliant minds who have studied the regulation size to no end, we should and can trust the idea that Komi will even things out.

Unfortunately, is simply does not work the same way on 9 x 9.

On 9 x 9, a 6.5 points Komi represents 8% of the 81 available points.

Do you realize how ridiculously insane this is?

This represents everything wrong with the small board: [http://online-go.com/game/915202][1]

If we follow the logic that only 19 x 19 matters, then a 9 x 9 Komi should scale down to a Komi of 1.5 points for white.

Since 1.8% is all that matters, it can be used all odd-numbered squared dimensions.

If OGS does this, the 9 x 9 games will be more meaningful. There won’t be situations like:

“Something of white must die or I need to be as aggressive as all hell. Short of that, guaranteed loss.”

Or…

“Awesome. I play as white. I can screw up 3 or 4 times and still win.”

You guys discarded traditional terms like “tsumego” in favor of clear terms like “puzzle.”

I hope you can do the same for smaller board sizes. If there is nothing written about the importance of the smaller size by great players in history and if there is nothing written about them in any kind of official documents about the game, then please innovate and fix it.

The other solution would be to make anything besides 19 x 19 as an unrated game but I don’t want that.

Just fix Komi. Thanks.


Komi on 9x9 game
Shouldn't komi for smaller board sizes be higher?
Nonstandard board size komi discussion
#2

Look at it this way:

How much more territory, in %, does Black’s first move secure/control/influence on a 9x9 board, compared to a 19x19 board?

I’ve seen discussions where folks said that the Komi on 9x9 should be even higher than on 19x19 for this reason. I couldn’t understand that at first, but as soon as I got a taste of “influence”, and esp. when I understood how much more the first move matters on 9x9, it was clear to me.

[quote]Do you realize how ridiculously insane this is?[/quote]Not insane at all :wink:

I for one prefer the term “Tsumego” :slight_smile:

Well, I rather hope they won’t fix what ain’t broken.

I often play on 9x9 and 13x13 as white, and yes, it hurts to play with the weight of Komi strapped to my feet … but 1) all it really hurts is my ego, and 2) the pain goes away with every move we play … after four moves it doesn’t hurt that much anymore, and after eight moves I can hardly feel it anymore :wink:

But don’t take my words for the bible … please cf. what people cleverer than me think about this:

Cordially, Tom


#3

They actually changed komi for smaller boards, it’s 5.5 according to http://senseis.xmp.net/?HandicapForSmallerBoardSizes – “Old Japanese Recommendation”

I won’t tell you some absolute truth about komi, however, your math assumes that playing the first stone is worth 1.8% of the points. It was my first line of thought, too, when the komi was 6.5 back then. But actually there are more differences between 19x19 and 9x9 than just the amount of points you can get.
On 9x9 the first stone black plays is involved in that local fight 9x9 is all about opposed to 19x19 where it’s a rather somewhere distributed stone.
Best would be statistically relevant data on win/lose ratios for different komis, but i can’t provide them.

If there is nothing written about the importance of the smaller size by great players in history and if there is nothing written about them in any kind of official documents about the game,

Well, can only quote what the formerly linked senseis article states on the komi which is used at the moment:

Around 1985 there was an article in the Japanese magazine Igo Kurabu by Ishikura Noboru (now 9p) on komi and handicaps for both 9x9 and 13x13 boards. His recommendations were based on the results of many pro-pro games on these boards.

As easy solution I can suggest playing unranked and setting your own komi, OGS supports that.


#4

I believe I read somewhere that for a time it was considered possible that komi for a 9x9 board could be up to 30 points! The idea being that the first move was worth much more on a smaller board.


#5

I was about to agree with OP, then I saw this comment. Thinking it that way, it makes a lot of sense.


#6

What is the correct Komi on a 5x5 board? What percentage of the available points is that?


#7

This is actually a hot area of contention. Yilun Yang 7p thought 2.5 was the right komi for 9x9. (See commented game #1.) This seems common-sense because the board is so small. But he’s in the minority. Virtually all professional 9x9 games have used 5.5 komi. That’s TV games, formal tournaments, everything. The RICOH Pair Go cup even used 6.5 for 9x9 decision games. Say what?

It may be helpful to discuss methods of deriving komi. Similar to 19x19, there’s basically two schools of thought: ideal komi and result-based komi. Ideal komi is the komi value that would result in a draw after perfect play from both. We don’t have that for 9x9, unfortunately. However, I’ve been conducting extensive research on 9x9 lately using a bot for an evaluation function. Having traced professional games backwards selecting moves with a high evaluation output score, my research currently suggests a value of 6 komi to be ideal. Therefore, the OGS 5.5 should be agreeable for black.

The second school is result-based komi. This method appeals to actual game results, looking to set komi at an equal win rate for black and white. This is probably the majority view, and my research also suggests 5.5 komi here. My collection of 267 pro games (linked above) contains a 47.3% win rate for black opening on tengen and 60% opening at 5-4 and 65.8% opening at 4-4. Over 200 of these games had 5.5 komi and 28 had 6.5. Therefore, OGS komi both conforms to the standard komi and statistical win rates. If anything, more is needed for white.

For these reasons, I believe the initial loss for black is kind of illusory and the advantage of going first will show its worth upon closer inspection.


#8

Komi is tricky for the alternate board sizes, due to the relative lack of data and consideration in comparison to the standard 19x19. It seems like 5.5 or 6.5 is pretty close, but I guess it will always be a matter of some debate.

@108: I doubt that 30 points of komi is reasonable for the 9x9 board. Virtually all of the suggestions (see the references that others have posted) seem to put it in the single-digits (usually 5.5 or 6.5). Maybe you’re thinking of the 5x5 board, where white can’t even live given perfect play and hence 25 is suggested as a fair komi.


#9

I’m pretty sure that if Black plays right, White cannot make a living group. Therefore komi on 5x5 should be 25 points.


#10

It was cool to be the heel in this debate. :slight_smile: Only one “almost” on my side :-).

@mark5000 had the most enlightening response. Thank you. Does results based mean that the Komi would change depending on black’s starting point? That would be a major
departure from the organic nature of the game.

Anyway, to everybody who thought I had a snowball’s chance in hell of persuading OGS to revise or eliminate 9 x 9 Komi, I appreciate the laugh. :slight_smile:

This was a case of “I’m probably going to lose so I’ll toss a stone deep into enemy territory and see what happens.” :slight_smile:


#11

No. The concept of komi is independent of where black chooses to play his first move. Komi points represent compensation given to white for having the disadvantage of moving second. Some first moves may be better than others for black, but I don’t think many have suggested to set komi based on black’s first move.

Don’t worry, I doubt that anyone was really concerned that your argument might actually affect change on OGS regarding 9x9 komi. I think the response was mainly due to people clamoring to point out the flaws in your arguments, which even seemed a bit like trolling given their tone.


#12

Black starting at 3-3 he has already secured ~8 points!


#13

I’m only 12 days old in this GO world - but I see the komi advantage as being a bit much. When you get a 25k (me) against a 22k and they have a 5.5 disadvantage. I’d imagine that these 22k folk are exactly setting the world on fire.

Look at this;

I’m 24k against 25k and have a 5.5 advantage. hmm.


#14

This is where you see the difference between komi and handicap.

The only role of komi is to counteract the advantage black has of going first in an otherwise even game.

If your desire is to play an opponent of a different rank but you want both players to have an equal chance of winning, this is where handicap stones come in (assuming both players’ ranks are accurate)


#15

I’m glad this thread got revived, as I would like to give moral support to @Professor_X. I agree that komi is too high for 9x9, at least for DDKs, but since you can choose to play white, or set a custom komi, it can be mitigated.Oddly enough, a compilation of more than 17,000 19x19 OGS games (Win rate by players ranks in OGS) showed that all but two ranks from 16k to 1d had a losing percentage as black. Doesn’t this suggest that komi is too high even for these people in 19x19?

In another thread (New way of deciding Komi), I pointed out that komi for 19x19 is, I think, based on pro games, so it is not logical to believe that it has equal validity for lower-level games. The weaker the skill, from pro to DDK, the less ability a person has to exploit the advantage of the first move. I suggested that the ideal komi probably exists on a sliding scale based on rank. This is similar to what @zedsdeadbaby argues above.

In a long interesting discussion in the chat a couple months ago, @thought argued, very cogently I believe, that the stats actually show 7 to be the ideal komi for 19 x 19 (and logically, komi should be an integer since there are no half points in the game). Consequently, the half-point tie-break distorts the komi from the start.

To come back to 9x9, let’s look at another avenue for analyzing this. Instead of considering komi as a proportion of total points on the board, consider it as a proportion of the winner’s score (not the difference). Let’s say, for argument’s sake, that 65 is the average winning score in 19x19 (and leaving aside issues of rank). Then the Japanese komi would be 10% of the winner’s score. For 9x9, my impression is that the average winning score is in the low 20s. If 22, a 6.5 komi would be roughly 30%. The 5.5 OGS standard would be 25%. That strikes me as very large. Even if you substitute different numbers, the point is that the practical limit on how many points can be scored is vastly lower in 9x9 and hence the significance of komi looms larger.


#16

But as @mark5000 pointed out, with less stones being played, while 6.5 points is a higher percentage in terms of score, the influence gained by getting the first play on such a small board is arguably larger.


#17

I’ve wrestled with that, and I’m not quite convinced. The size itself may be a limiting factor. Black has far more constraints of space on how he can take advantage of the first move. In other words, the small board limits the scope of what Black can do. Indeed, games do not always follow a linear logic. While we can probably safely conclude that the first move is still advantageous in 9x9, the limits on ts influence are really unknown. To illustrate what I mean, in poker your table position can have several consequences, good or bad, depending on the particular game being played. In 6-way Chinese checkers (traditional rules, no jumping in and out of home territories), I have analyzed hundreds of games against the computer and concluded that 4th position has an advantage (apparently because the networks of pieces first come to fruition in time for 4th position to take advantage of them).


#18

I would be carefull with statistics :smiley: they are tricky. AFAIK OGS gives black to the weaker player, that might partly explain why black is losing more often and maybe the komi is actually fair? There is many factors at play, you can’t just deduce komi is not fair from something so simple.
I dunno, I am just reluctant to start screwing with komi at our level of understanding the game. Also it would seem very silly when someone from OGS went elsewhere and was surprised by different komi.

Just dont change the rules :smiley: there is a standart set up by people who play better than us, let’s keep it simple. You know to create a far fetched analogy we are back to that one driver who thought driving on the other side of the road would actually be better. I for one think that at our level of play the 1 or 2 points make no real difference. It’s not like I would constantly by losing by 1 point, is it…


#19

Good point about the weaker player being assigned black. I hadn’t thought of that, and it may indeed offer part of the explanation for those numbers. However, I don’t believe that anything I said depends on go knowledge or playing strength. The issues are statistical, and they depend on such things as understanding validity, which, as you say, is certainly a very tricky business in statistics, For example the issue of the half point added to komi is not a go issue at all. It is a tie-breaker, a subject much discussed in philosophy of sports, particularly with reference to soccer (football to non-Americans), because many people think that tie-breakers are often ill-conceived and unnecessary.

My main beef is the continual trotting out of pro statistics to validate komi across the board, all the way down to DDK, as if the majesty of pros trumps all arguments. The argument is laughable from the standpoint of logic. It’s like polling some Wall Street bankers and then trying to apply the results to Blue Collar workers. I would love to see someone do a master’s thesis that compiled statistics from all levels to derive a komi for each level, Perhaps it would produce a curve that reflects the game’s learning curve.

I do not seek a change in komi. I do not have the activist personality needed to pursue such a program (especially not at my age). I just enjoy discussing various theoretical issues. I would perhaps like to see a reduction in what seems a misplaced reverence for komi as something writ in stone.

The solution for people who don’t like the komi in 9x9 is simple: play white.


#20

I haven’t actually looked at the records, but I feel like the percentage of DDK games that were won or lost by less than 5 points would be more or less statistically irrelevant. Even in 9x9 games I typically observe a 10+ point win one way or the other making komi in DDK games more or less a formality / novelty with little to no impact on the majority of games (due largely in part to DDK players not being able to efficiently take advantage of black’s starting sente)