Forum for Drago
Counting bug (chineese rules) [#1322]
I have got a bug while playing against MoGo (Chineese rules, handicap 3, komi 0.5)
White wins below, while Drago shows black as winning:
SGF is below:
(;GM[1]FF[4]AP[Drago:3.03]SZ[9]HA[3]ABcg?gc?cc?KM0.5RE[];W[df]KM0.5PB[MoGo:level NoLevel]PWUser?DT2008-03-12;Bcf?;Wed?;Bec?;Wcd?;Bdc?;Wce?;Bdg?;Wef?;Beg?;Wfg? ;Bff?;Wfe?;Bgf?;Wfc?;Bde;Wdd?;Bee?;Wfd?;Bgd?;Wge? ;Bhe?;Whf?;Bhg;Whd?;Bif?;Whc?;Bbe?;Wbd?;Bbf?;Wid? ;Bae?;Wad?;Bfh?;Web?;Bie?;Wgb?;Bch? ;TBaf?df?ef?hf?ag?bg?fg?gg?ig?ah?bh?dh?eh?gh?hh?ih?aibi?ci?di?ei?fi?gi?hi?ii?TWaa?ba?ca?da?ea?fa?ga?ha?ia?ab?bb?cb?db?fb?hb?ib?ac?bc?cc?dc?ec?gc?ic?gd?)
Numbers of intersections and stones seem correct. How do you calculate the score? --Gilles
It appears the komi of .5 is set as a reverse komi-- white gets 3 points of compensation for the unavailable locations on the board due to the 3 handicap stones, less .5 points as reverse komi. This explains why 2.5 points are being added to white's score.
Dave: See the handicap section on Territory Scoring Versus Area Scoring where it quotes the AGA rules "If the players have agreed to use area counting to score the game (Rule 12), White receives an additional point of compensation for each Black handicap stone after the first." (emphasis added) So the compensation that White should receive for a handicap of 3 is only 2 points and the above score looks correct.
What IS strange that I have checked this setup on other programs - Leela, SmartGO and MGS – they all show white as wining – Chinese rules, handicap 3, komi 0.5.
According to ![[ext]](images/extlink.gif)
http://www.britgo.org/rules/compare.html there is a difference between AGA rules and Chinese rules on how to compensate for handicap stones. With N handicap (in this case 3), AGA uses compensation N-1 (in this case 2), while chinese uses compensation N (in this case 3).
So it seems that according to AGA rules, black won, while according to Chinese rules, white won.
This can be checked with cgoban which score tool enables AGA counting and Chinese counting.
Part of the problem comes from the fact that MoGo does not support the final_score GTP command but rather the final_status_list command giving the prisoners. As a consequence, the client program has to calculate the score with the risk of error (or misunderstanding). However, the calculation used by Drago has been confirmed to me by MoGo author. I will check again.
The way of counting should definitively been explained somewhere. I will add this in the game engine page from Drago website.
Here is the sgf in a more convenient form:
(;GM[1]FF[4]SZ[9]HA[3]KM[0.50]PW[User]PB[MoGo:level NoLevel?]DT[2008-03-12]AB[cc][gc][cg] ;W[df];B[cf];W[ed];B[ec];W[cd];B[dc];W[ce];B[dg];W[ef];B[eg] ;W[fg];B[ff];W[fe];B[gf];W[fc];B[de];W[dd];B[ee];W[fd];B[gd] ;W[ge];B[he];W[hf];B[hg];W[hd];B[if];W[hc];B[be];W[bd];B[bf] ;W[id];B[ae];W[ad];B[fh];W[eb];B[ie];W[gb];B[ch])
I have checked it in CGoban, which gives:
- Chinese rules: white wins by 0.5
- AGA rules: black wins by 0.5
- New Zealand rules: black wins by 2.5
Which confirms the britgo rules comparison.
After clarification, it appears that MoGo uses H+0.5 and not AGA H-0.5. This will be fixed in next version.
Anton, actually you won that game! Sorry for inconvenience.
![[Welcome to Sensei's Library!]](images/stone-hello.png "Sensei's Library")