Score estimator, final game score disparity (the phantom point)


#21

My understanding is that in the recent past, both Chinese and Japanese rules used 5.5 for komi. People generally came to the consensus that 5.5 was a bit too small leaving white at a disadvantage, which was observed in pro game statistics. Hence, a logical response was to increase komi by a point.

However, with area counting (Chinese rules), most games end with an odd score difference (on the board before komi is applied), since the exception requires a seki. Hence, a 6.5 komi would not be much different than 5.5 komi under Chinese rules (B+5 still loses, B+7 still wins, and B+6 is unlikely). Thus, for Chinese rules, the komi was increased to 7.5 to make a more substantial difference.

To rephrase your last sentence: playing a Chinese game with 7.5 komi is (almost) like playing a Chinese game with 6.5 komi.


#22

In such a scenario, both black and white are idiots, as they both passed when there was at least one dame point on the board. This meaningless point in Japanese scoring that counts as an extra point in Chinese is precisely why Chinese use 7.5 komi and Japanese use 6.5 and the result doesn’t change.
Edit. Sorry @yebellz I replied before seeing your post I think we basically say the same thing but you probably said it better :slight_smile: good job on also remembering seki


#23

ICYMI, I think @smurph was subtly implying that counting is better than using a computer’s score estimator :wink:


#24

Oops… :slight_smile:


#25

Ugh… no? What dame? Last time I checked, Go board had an odd number of intersections.


#26

Exactly! Which means they can’t both make the same number of moves, as that is an even number.


#27

It’s amazing to observe confirmation bias in the making. Are you honestly suggesting that it’s impossible to have a game in which, let’s say Black and White made 100 moves each and out of remaining 161 points of territory 77 is white’s and 84 is black’s ?


#28

No… just pretty sure this is how I’ve always heard it taught…? I can’t fault your logic, the example provided certainly seems like it should be a win for white under Chinese rules and a win for black under Japanese… but it’s very unnerving because I’m sure I’ve always heard it taught that the ruleset almost never affects the result of the game (apart from a few fringe examples)


#29

I think what you remember is that using territory or area scoring does not affect the result of the game, usually.

The intrinsic rules of Japanese and Chinese do not differ on that game that was mentioned, it’s because of the height of the komi alone. Assuming that 7 points should be the “correct” value to ensure a tie, the Japanese rules choose to give the win to black with 6.5 komi, while the Chinese rules choose to give it to white with 7.5 komi.


#30

It is possible for two players to make the same number of moves and for black to have a 7 point advantage in territory:


#31

Quote from sensei, where the “myth” (or is it?) probably originated (https://senseis.xmp.net/?Komi)

Area scoring
The usual komi in China was formerly equal to the 5.5 used in Japan, but was changed to 7.5. The jump is two points, because under area scoring the score is almost always odd.[3] The Ing rules also have a komi of 7.5, specified as 8 points with Black winning jigo. The American Go Association have also changed komi from 5.5 to 7.5 in August 2004, effective 2005. The New Zealand rules specify a komi of 7.
(…)
When area scoring is used (as in Chinese, Ing, AGA and New Zealand rules), the winning margin without komi is always odd, unless there are an odd number of points in seki. Since seki is fairly rare, and since a komi of 5.5 points has proven insufficient in professional play, and since a komi of 9 is generally considered far too much, it seems likely that the perfect komi is 7.

Discuss. :wink:


#32

Not the perfect komi again. Moderator and troublemaker in one is a dangerous combination :smile:


#33

I suppose 7.5 Chinese vs. 7.5 Japanese is free of the ‘phantom point’ problem (same as 5.5 komi) for the reasons @Vsotvep explained so clearly.


#34

Given that most bots agree that 7.5 is favourable for white, I feel like if komi were ever to change it would go down. Personally, I am an advocate of both rulesets using 7 komi and allowing jigo, but I can already feel the death glares from tournament directors on the back of my neck, so I imagine komi values won’t change any time soon.


#35

So here’s one of those fringe examples where (assuming 7.5 komi used regardless of rule set), the rule set used determines the winner: https://www.youtube.com/watch?v=3d00lk-U-ss

The final position involves a seki. Under Japanese rules, white wins by 0.5 points, whereas under Chinese rules, black wins by 0.5 points (or rather 0.25 stones, which is the equivalent of that). Michael’s explanation starts at 1:12:17, which is the point at which white resigned the game. The final board position if the game had been played to the end can be seen starting at 1:17:52.

Now, Michael’s explanation is a bit confusing to me. It seems to me that his explanation mirrors BHydden’s remarks: Michael says that if black wins by 7 points on the board under Japanese rules (as is the case in this game), that means that white has played the last dame point - how come? Michael argues that if black is up by 7 points, that means that the total amount of territory must be odd - so far, so logical. But he says that as a consequence of that, the number of stones on the board must be even. Now I don’t see why that has to follow under either rule set. Does anyone else feel qualified to elaborate on this?

Interestingly, if this game was scored under Japanese rules with 6.5 komi, then the winner would be the same as determined by Chinese/7.5. :wink: (I’m not saying that this means that there are no fringe cases under which different rule sets with different komi could also lead to different winners.)


#36

Odd plus even = odd
odd plus odd = even

If territory is odd, and total points is odd, stones played must be even, ergo white played last
If territory is even, obviously points are still odd, stones played must be odd, ergo black played last

did that help clear it all up?


#37

Oh duh. I confused territory with area. I kept thinking “but the territory already includes all the stones on the board”. Nope, the area does :slight_smile:

But hang on - under territory counting, doesnt it also matter whether the difference between prisoners captured is odd or even if youre trying to figure out who played the last move?


#38

To put Michael’s explanation in simple terms, if W was able to play the last move, the difference of areas would have been 7 points (usual odd result), but because of the seki, missing last move from W made the area difference 8 points, while the territory stays the same.

There are other cases where rule subtleties (like bent-four, etc.) can affect the result in more spectacular way, but in this case the outcome is simply result of the difference in counting.

One thing I don’t really get is that after I uploaded the .sgf file, the score estimate shows B+2.5 as a result. Why would that be the case? :smiley: https://online-go.com/game/13569426 I’m a little too lazy right now to verify the score manually.

EDIT: I did the count and it adds up to 184 points, so it’s B+0.5 indeed. I’m just wondering if Chinese players do the ad-hoc counting in Chinese or do they just score territory… It takes forever to count it that way :smiley: