[computer-go] Results of recent Computer Go events
Gian-Carlo Pascutto wrote: Apparently it was ruled a loss for Many Faces of Go. I am appealing it - there is no reason why the refree has to intervene when the players agree on the score. The result of the game could be very important for the tournament result. Hello, I looked into the tournament rules at http://www.icga.org/tournaments/olympiadgo.pdf Paragraphs 3.7.2 and 3.7.3 say that the result should be taken on which the programs agree. Greetings, Ingo. -- GMX Kostenlose Spiele: Einfach online spielen und Spaß haben mit Pastry Passion! http://games.entertainment.gmx.net/de/entertainment/games/free/puzzle/6169196 ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
[computer-go] Article By Zhou Junxun 9p on the MoGo Matches
http://post.weiqi.tom.com/BA000A6427.html in Chinese. A quick summary on the games -- He very rarely played 9x9 games. On the first 9x9 game after the 11th move, to his shock he figured he was already lost. But then he spent 5 minutes to design a Hamete to win the game back. The second game was easy. As for 19x19 game, he figured he had a sure win in about 30 moves. The following is the google translation of the excerpt on the two 9x9 games. The first game under the MOGO in the first 11 hands (black sixth-step) I have found that natural shock defeat (at the time of the heart is really very shocked ..) later took five minutes to calm feelings. How to start designing a computer can not be cheated easily see through and reversed. Article 20, I started out under the Council of the wins (also a Super cheat) (MOGO heard before the computer is in the Go Changhao Li. He is very good at only half-won Purpose of the final model) MOGO long after the test was the next out and I expect the same in the hand-shun. (MOGO if the first fight all the way to eat in the upper under accordance with the actual combat. This is the last game I lost half a head) I was very lucky The first set victory. Authority in the first place for me to experience the way the board has nine more. Black on the second set I was relatively easy victory. ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] symmetry optimizations vs statistics?
Perhaps this is also among the well-known material (or something
There have been discussions about handling symmetry in the past.
See for instance Heikki Levanto's group theoretic hashing paper.
For 'classic' (non MC) programs, board-symmetries were not important,
except for handling joseki/ fuseki collections: in a 'normal' game, the board
has no simmetries after a few moves, and there will
after that point *never* be a position that has one of its 15
mirror images present in the same gametree. So there is no need to check for
them,
because there is no possible gain.
MC is different, because it needs to re-evaluate the same
position many times, the even boardsize in Don's example
will make it even more pathological.
can be reached in multiple ways (a generalization of miai?), so counting
The miai-case is the topological variant of a symmetry.
Recognising these would need a whole different coding
('topological hashing' ?), and probably save a lot of
(wasted) effort in tsumego/semeai/endgame playing.
IMHO (I am not a statistician) not being aware of the symmetries
in MC (as in Don's case) just leads to wasted effort (and under-estimation of
the samples quality), but not to wrong results.
HTH,
AvK
___
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] symmetry optimizations vs statistics?
On Tue, Sep 30, 2008 at 04:29:12PM +0200, A van Kessel wrote: There have been discussions about handling symmetry in the past. See for instance Heikki Levanto's group theoretic hashing paper. I'm afraid you must have misattributed that - I don't know much about hashing, less about group theory, and not being in the academia, I am not publishing papers. I am just a programmer who likes to dabble with programming Go, when other interests don't claim all of my spare time. - Heikki -- Heikki Levanto In Murphy We Turst heikki (at) lsd (dot) dk ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
[computer-go] Presentation related to move prediction
Hi, sorry for the vagueness... but I was recently informed that David Stern would have kept a presentation including some (nice) results on move prediction earlier this year. I tried to email directly to him, but perhaps I didn't get his email right... anyway, does anyone know anything? (Hmm... lots of anys there...) -Esa ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] symmetry optimizations vs statistics?
Oops I confused you with Antti Huima. No offense... I meant: http://fragrieu.free.fr/zobrist.pdf Sorry, AvK ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Correct Komi for 6x6 is 2.0
I would be very interested in various opinions on this. Is the correct komi for 6x6 (under CGOS type rules) 2.0? Especially would I like to see some strong players review my (actually Leela's) analysis and weight in on this. Do you think we can say with relatively high confidence that 2.0 is the correct komi for 6x6 go? I may later crank up the level for Leela (for this study I left it at the default which was game in 5 minutes) on a loaded core 2 duo machine which typically had a load of 3-5 during the tests. - Don On Mon, 2008-09-29 at 15:10 -0400, Don Dailey wrote: On Wed, 2008-09-24 at 19:48 +0200, Erik van der Werf wrote: I'm quite confident that 4.0 is the correct komi for 6x6. I am quite confident that it is 2.0 :-) I admit there is some room for error on my part, but I have just done a fairly significant study of 6x6 Go using Leela. My primary room for doubt is if there is some kind of end of game issue (in programs like Leela) such as seki that causes a gross and systematic error. Here is why I think komi should be 2.0 and if you can prove me wrong, I think we will probably both have learned something interesting - I hope you can. Here is my analysis: I did an analysis of 784 6x6 leela games. These games were played at 2.5 komi and white tends to win most of the games. After converting all the game to a canonical representation, I discovered that Leela always plays the first 4 moves the equivalent of this: C3 D4 C4 D3 - all 784 games started like this or the equivalent After this, BLACK varied significantly: BLACK WINS/GAMES PERCENTAGE - - C3 D4 C4 D3 C20 out of6 = 0.000 percent C3 D4 C4 D3 C51 out of4 =25.000 percent C3 D4 C4 D3 D2 103 out of 428 =24.065 percent C3 D4 C4 D3 D5 82 out of 339 =24.189 percent C3 D4 C4 D3 E30 out of4 = 0.000 percent C3 D4 C4 D3 E40 out of2 = 0.000 percent C3 D4 C4 D3 E50 out of1 = 0.000 percent For the rest of this discussion, the Wins and Percentage are ALWAYS from BLACK'S point of view. In this sequence, the only statistically interesting moves are D2 and D5 for black, because these 2 choice constitute the vast majority of the games. (It might be slightly more interesting if other moves showed black doing well, but in the minor lines of play black is losing.) The 2 good moves appear to be approximately equal in value ... however, let's check that out. Let's consider D2 first. If black plays D2 white plays either E2 or C5. C5 appears to be a mistake. When white plays C5 black wins 70% of the games. However, Leela only played that move 20 times. 408 times it played E2 doing quite well - holding black to about 22% BLACK WINS/GAMES PERCENTAGE - - C3 D4 C4 D3 D2 E2 89 out of 408 =21.814 percent C3 D4 C4 D3 D2 C5 14 out of 20 =70.000 percent Let us assume that C5 is a blunder and with correct play black aways wins (all of this assume 2.5 komi.) AT this point black plays 3 different moves, all of them losing: C3 D4 C4 D3 D2 E2 C20 out of7 = 0.000 percent C3 D4 C4 D3 D2 E2 E41 out of 42 = 2.381 percent C3 D4 C4 D3 D2 E2 D5 88 out of 359 =24.513 percent Again, the only interesting move here is D5 for black. White responds with E5 or C2. C2 appears to be a blunder and Leela played it 15 times, losing every time as white. However, E5 keeps black down to 21.221 percent: C3 D4 C4 D3 D2 E2 D5 E5 73 out of 344 =21.221 percent C3 D4 C4 D3 D2 E2 D5 C2 15 out of 15 = 100.000 percent At this point we seem to be well into the 6x6 game and none of blacks responses seem to be game winning. Carrying this out 1 pair of moves farther (which I am not including here), I see more of the same - there is nothing that indicates that black has a surprise game winning move that white cannot avoid. Ok. So let's go back to our other early black choice, 5. D5 C3 D4 C4 D3 D5 82 out of 339 =24.189 percent Leela played 2 moves here, again, one of them may be a blunder because it allows black to win 81% of the games: C3 D4 C4 D3 D5 C2 13 out of 16 =81.250 percent C3 D4 C4 D3 D5 E5 69 out of 323 =21.362 percent So let's see if E5 is interesting. If it is, we might be able to furnish empirical proof that black can win a 2.5 komi: C3 D4 C4 D3 D5 E5 D2 69 out of 284 =24.296 percent C3 D4 C4 D3 D5 E5 F50 out of1 = 0.000 percent C3 D4 C4 D3 D5 E5 C50 out of4 = 0.000 percent C3 D4 C4 D3 D5 E5 E30 out of 34 = 0.000 percent Apparently, only 1 black choice here is reasonable, D2. After Black plays D2, Leela probably loses after white
Re: [computer-go] Article By Zhou Junxun 9p on the MoGo Matches
(Sorry, I meant to forward that to a Chinese-speaking colleague, not to re-send it to the list.) Peter Drake http://www.lclark.edu/~drake/ On Sep 30, 2008, at 8:47 AM, Peter Drake wrote: The article is in Chinese, so I have no idea if there's anything of interest. Peter Drake http://www.lclark.edu/~drake/ On Sep 30, 2008, at 7:12 AM, John Fan wrote: http://post.weiqi.tom.com/BA000A6427.html in Chinese. A quick summary on the games -- He very rarely played 9x9 games. On the first 9x9 game after the 11th move, to his shock he figured he was already lost. But then he spent 5 minutes to design a Hamete to win the game back. The second game was easy. As for 19x19 game, he figured he had a sure win in about 30 moves. The following is the google translation of the excerpt on the two 9x9 games. The first game under the MOGO in the first 11 hands (black sixth- step) I have found that natural shock defeat (at the time of the heart is really very shocked ..) later took five minutes to calm feelings. How to start designing a computer can not be cheated easily see through and reversed. Article 20, I started out under the Council of the wins (also a Super cheat) (MOGO heard before the computer is in the Go Changhao Li. He is very good at only half-won Purpose of the final model) MOGO long after the test was the next out and I expect the same in the hand-shun. (MOGO if the first fight all the way to eat in the upper under accordance with the actual combat. This is the last game I lost half a head) I was very lucky The first set victory. Authority in the first place for me to experience the way the board has nine more. Black on the second set I was relatively easy victory. ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] symmetry optimizations vs statistics?
Since statistics play such a vital role in modern Go engines, is there a danger that good old-fashioned low-level optimizations interfere with those statistics? Things like not re-evaluating symmetric positions (boar mirroring/rotation, move transposition, etc.) multiple times interacting with statistics based on number of evaluations. It seems that the keyword transpositions gives more useful search results in the list archives than just symmetries. There seem to be several issues, including: - if one uses hashing into a transposition table, one will simply be able to find and reuse results from earlier visits to similar positions via different paths; that might interfere with results that only exist as sums (eg, ownership maps/territory heuristic) stored higher up in the tree (boards resulting from playouts are added into a single higher-up array, then thrown away); the solution here is probably to keep more local data, but that could be expensive.. - if one uses transpositions to find all paths into the current position, one will be able to sample multiple simulations in one go; that might interfere with biases and confidence (giving more samples to positions that can be reached in multiple ways); the solution here is probably to be careful about the statistics, whatever that may mean;-) What I was wondering about was to what extent and with what results such interactions between statistical sampling and optimizing transpositions have been investigated. Is that any clearer? Claus ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Correct Komi for 6x6 is 2.0
I have been trying to see what Valkyria does. But it is a little unstable when it reads deep at 6x6. It should not be a problem for Valkyria but I have not had any time to search for the bug. Anyway the 2.5 komi black should lose if Don is right. So I have the following very cute complete game sequence: bC4 wD3 bC3 wD4 bD5 wE5 bD2 wE2 ... The beginning is very natural, and although I did not check carefully all what Don reported but I think there is agreement that this is a strong seqence for both colors. ... bE3 wE4 bE1 wF3!!!... Normally wF2 is played in the corner. But with wF3 white has the option to play aggresively with wF1 which usually is a bad idea because the ko fight risk to much. But on a small board things are not normal... ... bC1 wC5 bB5 wD6 bB6 wF1!... white starts a ko. If white first plays wC6 bF1 wins the game right away. ...bF2 wC6 bB4! if I understand this move it is a truly cool move. If black plays the larg ko at E3 as most people would instinctively do including myself white cuts at B4 and get a lot of kothreats and wins the game. With bB4 white has no ko threats. ...wF1 bE6 wF6 bF2 wE3 bD1 wF1 bF5 wF4 bF2 wPass bF1 B+3.5 So komi should be larger unless there are a mistakes by white in this sequence. Valkyria should have the credit for all strong moves here but they are not always found, so this analysis is also shaky. At short time controls I do not think any existing MC programs will get these things right. Even if it is wrong I found some of the moves really fascinating. The search behavior of Valkyria jumps up and down between 25-75% all the time as these moves are found. If it had been more stable maybe I could just let it run for a long time and find out stuff more or less for sure. But I hope this sequence adds some understanding at least. -Magnus Quoting Don Dailey [EMAIL PROTECTED]: I would be very interested in various opinions on this. Is the correct komi for 6x6 (under CGOS type rules) 2.0? Especially would I like to see some strong players review my (actually Leela's) analysis and weight in on this. Do you think we can say with relatively high confidence that 2.0 is the correct komi for 6x6 go? ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Correct Komi for 6x6 is 2.0
Magnus, It seems I didn't go deep enough into the game. I assumed that after quite a few moves we could reach some tentative conclusion that was likely to be correct, but it looks like there is a good chance it is not correct! What I should have done was to take the time to mini-max the statistics. What I'm doing is more tedious than just writing a program to do it for me. I found another line of play too. See my responses to your stuff below: On Tue, 2008-09-30 at 22:09 +0200, Magnus Persson wrote: I have been trying to see what Valkyria does. But it is a little unstable when it reads deep at 6x6. It should not be a problem for Valkyria but I have not had any time to search for the bug. Anyway the 2.5 komi black should lose if Don is right. So I have the following very cute complete game sequence: bC4 wD3 bC3 wD4 bD5 wE5 bD2 wE2 ... The beginning is very natural, and although I did not check carefully all what Don reported but I think there is agreement that this is a strong seqence for both colors. ... bE3 wE4 bE1 wF3!!!... In my data from Leela, F3 is looked at 100 times here and Leela considers it the best move too: b w b w b w b w b w b w C3 D4 C4 D3 D5 E5 D2 E2 E3 E4 E1 F3 24 out of 110 =21.818 percent C3 D4 C4 D3 D5 E5 D2 E2 E3 E4 E1 C56 out of 18 =33.333 percent Normally wF2 is played in the corner. But with wF3 white has the option to play aggresively with wF1 which usually is a bad idea because the ko fight risk to much. But on a small board things are not normal... ... bC1 wC5 bB5 wD6 bB6 wF1!... It's interesting that Leela considers C1 a losing move! However, C2 actually wins more than half the time in the games. (Just because Leela doesn't like bC1 doesn't mean it isn't correct however.) I have further analysis that shows that C2 works: b w b w b w b w b w b w After C3 D4 C4 D3 D5 E5 D2 E2 E3 E4 E1 F3 C10 out of 10 = 0.000 percent F20 out of5 = 0.000 percent B20 out of 18 = 0.000 percent C2 24 out of 44 =54.545 percent D10 out of 33 = 0.000 percent I had to follow this analysis pretty deep as the next several moves appear forced from Leela point of view. There were a few games thrown away where black played something stupid, so I'm not including them. I eventually get this which I will call Leela's principal variation: b w b w b w b w b w b w After C3 D4 C4 D3 D5 E5 D2 E2 E3 E4 E1 F3 b w b w b w b w C2 C5 B5 D6 B6 F1 F2 C6 24/35 = 68.571 percent I don't know about your line, but it appears that maybe black DOES win with this line (and perhaps your line too) unless we can find a white improvement. So it may very well be that we should test 3.5 komi! It appears that black had the choices - so unless we can find a white improvement I am inclined to believe that 3.5 komi is closer to the truth. The lesson here seems to be that even 6x6 is deep for very strong computer programs. Even in 6x6 after getting out 20 moves Leela is not sure this is a win for black. (I'm not sure either, I am just assuming that 24/35 = 68.571 percent represents a win for black.) - Don white starts a ko. If white first plays wC6 bF1 wins the game right away. ...bF2 wC6 bB4! if I understand this move it is a truly cool move. If black plays the larg ko at E3 as most people would instinctively do including myself white cuts at B4 and get a lot of kothreats and wins the game. With bB4 white has no ko threats. ...wF1 bE6 wF6 bF2 wE3 bD1 wF1 bF5 wF4 bF2 wPass bF1 B+3.5 So komi should be larger unless there are a mistakes by white in this sequence. Valkyria should have the credit for all strong moves here but they are not always found, so this analysis is also shaky. At short time controls I do not think any existing MC programs will get these things right. Even if it is wrong I found some of the moves really fascinating. The search behavior of Valkyria jumps up and down between 25-75% all the time as these moves are found. If it had been more stable maybe I could just let it run for a long time and find out stuff more or less for sure. But I hope this sequence adds some understanding at least. -Magnus Quoting Don Dailey [EMAIL PROTECTED]: I would be very interested in various opinions on this. Is the correct komi for 6x6 (under CGOS type rules) 2.0? Especially would I like to see some strong players review my (actually Leela's) analysis and weight in on this. Do you think we can say with relatively high confidence that 2.0 is the correct komi for 6x6 go? ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ signature.asc Description: This is a digitally signed message part
Re: [computer-go] Correct Komi for 6x6 is 2.0
On Tue, 2008-09-30 at 16:58 -0400, Don Dailey wrote: In my data from Leela, F3 is looked at 100 times here and Leela considers it the best move too: b w b w b w b w b w b w C3 D4 C4 D3 D5 E5 D2 E2 E3 E4 E1 F3 24 out of 110 =21.818 percent C3 D4 C4 D3 D5 E5 D2 E2 E3 E4 E1 C56 out of 18 =33.333 percent Of course I didn't mention this, but wC5 should be checked out too since it is a white move and Leela still sees it as a win. - Don signature.asc Description: This is a digitally signed message part ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
[computer-go] Correct Komi for 6x6 is 2.0
Hello Magnus, interesting and strange. I went through your constructed game (it is repeated here without the in-between text) bC4 wD3 bC3 wD4 bD5 wE5 bD2 wE2 ... ... bE3 wE4 bE1 wF3!!!... ... bC1 wC5 bB5 wD6 bB6 wF1!... ...bF2 wC6 bB4! ...wF1 bE6 wF6 bF2 wE3 bD1 wF1 bF5 wF4 bF2 wPass bF1 B+3.5 Here I do not understand your counting: Black has control over 21 sqaures, White over 15. So, the final score of your game would be B+6. But ... when White instead of passing continues wC2, the game should go on with bB2 wF1 bPass wF2, and now the score is B+2. Or did I miss something? By the way, in my experimental runs with Leela it was really strange to see the evaluation switching between rather extremes - on this little board. Ingo. -- Pt! Schon vom neuen GMX MultiMessenger gehört? Der kann`s mit allen: http://www.gmx.net/de/go/multimessenger ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Correct Komi for 6x6 is 2.0
Quoting Ingo Althöfer [EMAIL PROTECTED]: Hello Magnus, interesting and strange. I went through your constructed game (it is repeated here without the in-between text) bC4 wD3 bC3 wD4 bD5 wE5 bD2 wE2 ... ... bE3 wE4 bE1 wF3!!!... ... bC1 wC5 bB5 wD6 bB6 wF1!... ...bF2 wC6 bB4! ...wF1 bE6 wF6 bF2 wE3 bD1 wF1 bF5 wF4 bF2 wPass bF1 B+3.5 Here I do not understand your counting: Black has control over 21 sqaures, White over 15. So, the final score of your game would be B+6. I am assuming a komi of 2.5 that is 6-2.5=3.5. But ... when White instead of passing continues wC2, the game should go on with bB2 wF1 bPass wF2, and now the score is B+2. Black ignores w C2 and plays F1. I added the pass only because it gives us the shortest game with the correct score for this variation. Or did I miss something? By the way, in my experimental runs with Leela it was really strange to see the evaluation switching between rather extremes - on this little board. I think it is completely normal and expected. MCTS is so strong on small boards that most of the time it searches variations with a certain outcome. But when the search discovers overlooked moves the score quickly change. Much more so than on larger boards. -Magnus ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Correct Komi for 6x6 is 2.0
I stubbornly always use x.5 komi because I do not like Jigo. in this case it was 2.5. I do that simply because Valkyria is coded like that, it cannot play with integer komi. Quoting Christoph Birk [EMAIL PROTECTED]: On Tue, 30 Sep 2008, Magnus Persson wrote: ...wF1 bE6 wF6 bF2 wE3 bD1 wF1 bF5 wF4 bF2 wPass bF1 B+3.5 I guess you mean B+4. Couln't black win by just refusing to play the ko? I could B+4 in that case too. Christoph ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/ -- Magnus Persson Berlin, Germany ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Correct Komi for 6x6 is 2.0
Don Dailey wrote: 4. I believe Leela, at a higher level and with a correction book would play perfect or very close to perfect on 6x6. This may depend on seki issues however, it may not be possible for Leela (or other Go programs) to play perfectly without some minor adjustments to handle the weird corner cases. It probably makes sense to force a bit more exploration for this kind of analysis. I can make a version like that if you want. -- GCP ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
Re: [computer-go] Correct Komi for 6x6 is 2.0
On Wed, 2008-10-01 at 05:55 +0200, Gian-Carlo Pascutto wrote: Don Dailey wrote: 4. I believe Leela, at a higher level and with a correction book would play perfect or very close to perfect on 6x6. This may depend on seki issues however, it may not be possible for Leela (or other Go programs) to play perfectly without some minor adjustments to handle the weird corner cases. It probably makes sense to force a bit more exploration for this kind of analysis. I can make a version like that if you want. Yes, I would like that. I'll run it at a higher level too - and mini-max the game results. - Don signature.asc Description: This is a digitally signed message part ___ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
