Life In 19x19
http://lifein19x19.com/

Ideal komi and winning odds, based on pro games.
http://lifein19x19.com/viewtopic.php?f=45&t=10517		 Page 2 of 2
Author:  Ootakamoku [ Thu Aug 21, 2014 4:10 am ]
Post subject:  Re: Ideal komi and winning odds, based on pro games.
Latest discovery. With pass stone and white always getting last move, area scoring and territory scoring produces equal results for all practical purposes. However, since chinese rules dont use pass stone, this ends up favoring black by half a point, as roughly 50% of the time, black gets the last move instead of white. This means that chinese 7.5 komi is in fact 7.0 komi when compared to territory scoring. Accounting for this I redid the calculation, getting theory and practice to match better.

Not only did exponential curve match by far better than either logistic or normal, which was my expectation, especially for handicap. But the eventual total error was much smaller too.

Ideal komi : 6.516
Multiplier per point : 0.1737

There by if two equal strength pros played against each other, with 0.5 komi and 2 handicap, white would have 5% chance of winning, which matches intuitive expectations much better.

0.5^( 1 + ( 1.5 * 2 * 6.156 - 0.5 ) * 0.1737 ) = 0.0504

The 1.5 in this formula comes from the fact that black gets 2 handicap stones but white gets first move (which is worth half a stone).
Author:  HermanHiddema [ Thu Aug 21, 2014 5:05 am ]
Post subject:  Re: Ideal komi and winning odds, based on pro games.
Pass stones are actually irrelevant to all that. Pass stones do not impact area scoring at all, since they are prisoners and prisoners are not counted under area scoring.
Author:  Ootakamoku [ Thu Aug 21, 2014 7:00 am ]
Post subject:  Re: Ideal komi and winning odds, based on pro games.
HermanHiddema wrote:
Pass stones are actually irrelevant to all that. Pass stones do not impact area scoring at all, since they are prisoners and prisoners are not counted under area scoring.


http://senseis.xmp.net/?EquivalenceScoringExample

http://senseis.xmp.net/?LogicalProofOfT ... reaScoring

In area scoring, black has by default half point advantage. If white fills last dame, the result is equal to territory scoring. If black fills last dame, the result equals to territory scoring plus 1 point for black. Since (and I checked) its equally likely either player gets the last dame, the result is that chinese rules favor black by half a point more than komi would imply, hence chinese rules 7.5 komi is in practice more like 7.0 komi.
Author:  HermanHiddema [ Thu Aug 21, 2014 9:11 am ]
Post subject:  Re: Ideal komi and winning odds, based on pro games.
Ootakamoku wrote:
HermanHiddema wrote:
Pass stones are actually irrelevant to all that. Pass stones do not impact area scoring at all, since they are prisoners and prisoners are not counted under area scoring.


http://senseis.xmp.net/?EquivalenceScoringExample

http://senseis.xmp.net/?LogicalProofOfT ... reaScoring

In area scoring, black has by default half point advantage. If white fills last dame, the result is equal to territory scoring. If black fills last dame, the result equals to territory scoring plus 1 point for black. Since (and I checked) its equally likely either player gets the last dame, the result is that chinese rules favor black by half a point more than komi would imply, hence chinese rules 7.5 komi is in practice more like 7.0 komi.


Yes, exactly. This is a property of area scoring, and is not related to pass stones.
Author:  Ootakamoku [ Thu Aug 21, 2014 11:33 pm ]
Post subject:  Re: Ideal komi and winning odds, based on pro games.
HermanHiddema wrote:
Yes, exactly. This is a property of area scoring, and is not related to pass stones.


Yes its property of area scoring that it favors black by half a point on average, compared to territory scoring. I was merely trying to convey how the use of pass stone + white moves last (which affects territory scoring), is one way to ensure area scoring and territory scoring gives same result.

And yes, it was incorrect of me to say how chinese rules dont use pass stone, should be how japanese rules dont use pass stone.
Page 2 of 2 All times are UTC - 8 hours [ DST ]
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group
http://www.phpbb.com/