Game #2 (where B287 captured 29 white stones) had the most white stones captured by black (51), but white won the game. It also had the most prisoners total (80), resulting in the fact that over 25% of the stones played in that game were captured by the end.
Game #10 (where W244 captured 22 black stones) mirrored #2 with an equivalent difference in magnitude between black and white prisoners but favoring white, yet black won the game.
When I was starting out, I noticed that the number of captured stones was a pretty good predictor of the winner of a pro game. The player who captured more stones usually lost. (That's captured stones, not dead stones. It may matter.) Not that I made too much of that, and was not sure at all if it was generally true, but I certainly had no fear of sacrificing stones.
The Adkins Principle: At some point, doesn't thinking have to go on?
— Winona Adkins
Game #2 (where B287 captured 29 white stones) had the most white stones captured by black (51), but white won the game. It also had the most prisoners total (80), resulting in the fact that over 25% of the stones played in that game were captured by the end.
Game #10 (where W244 captured 22 black stones) mirrored #2 with an equivalent difference in magnitude between black and white prisoners but favoring white, yet black won the game.
When I was starting out, I noticed that the number of captured stones was a pretty good predictor of the winner of a pro game. The player who captured more stones usually lost. (That's captured stones, not dead stones. It may matter.) Not that I made too much of that, and was not sure at all if it was generally true, but I certainly had no fear of sacrificing stones.
I suppose it takes a little bit of the sting out of losing ko fights as well (=
Until/unless we have a better description of what is meant by AlphaGo vs AlphaGo I don't think we should draw conclusions. The same program, yes, in the sense of program vs program. But the same values in the neural nets the programs were emulating? I thought these self play games were being used to train the neural net (result in the altering of cell values) so I would expect the sets of those values to be slightly different << current "best" set of values vs set that MIGHT be better >> If that is the case, we can't speak of these games indicating "white favored at this komi" unless we know that the "current" and "trial" nets were randomly assigned colors.
Mike Novack wrote:Until/unless we have a better description of what is meant by AlphaGo vs AlphaGo I don't think we should draw conclusions. The same program, yes, in the sense of program vs program. But the same values in the neural nets the programs were emulating? I thought these self play games were being used to train the neural net (result in the altering of cell values) so I would expect the sets of those values to be slightly different << current "best" set of values vs set that MIGHT be better >> If that is the case, we can't speak of these games indicating "white favored at this komi" unless we know that the "current" and "trial" nets were randomly assigned colors.
Hmm. My impression was that the same version of AlphaGo played itself, with suitable randomization of close choices. And then those games were used to train the neural networks, both for guessing moves and evaluating positions.
The Adkins Principle: At some point, doesn't thinking have to go on?
— Winona Adkins
Game #2 (where B287 captured 29 white stones) had the most white stones captured by black (51), but white won the game. It also had the most prisoners total (80), resulting in the fact that over 25% of the stones played in that game were captured by the end.
Game #10 (where W244 captured 22 black stones) mirrored #2 with an equivalent difference in magnitude between black and white prisoners but favoring white, yet black won the game.
When I was starting out, I noticed that the number of captured stones was a pretty good predictor of the winner of a pro game. The player who captured more stones usually lost. (That's captured stones, not dead stones. It may matter.) Not that I made too much of that, and was not sure at all if it was generally true, but I certainly had no fear of sacrificing stones.
I suppose it takes a little bit of the sting out of losing ko fights as well (=
Yeah, I usually lost, errr, sacrificed two or three groups per game.
Fairly early on I heard the proverb, If you don't know ko, you don't know go, and welcomed ko fights.
The Adkins Principle: At some point, doesn't thinking have to go on?
— Winona Adkins
Mef wrote: I suppose it takes a little bit of the sting out of losing ko fights as well (=
More often than not, the compensation you get for losing the ko is much bigger anyway..
I am looking forward to seeing pros react to these Alphago games. Especially these unprovoked attachments all over. Maybe Alphago would be even stronger if not doing these. I don't think so. Extremely interested in what the comments will say.
If something sank it might be a treasure. And 2kyu advice is not necessarily Dan repertoire..
Kirby wrote:Bill, I think 7.5 komi may very well be correct komi - just want to point out that testing with current version of AlphaGo is not necessarily rigorous.
For example, maybe AlphaGo trained with 10.5 komi plays very aggressively as black to overcome the point difference. You come up with a different program, so it remains possible that black wins more often against itself with this aggressive strategy.
Can't you apply that argument to players of yore who made overplays as White to overcome the lack of komi?
Seems unlikely to me, but I just feel a single version of AlphaGo may not be qualified to generally prove correct komi.
the alphago games are incredible beyond words. the level is so high, with so many brilliant new ideas. I can tell you the sharp attachments look like very clever probes, moving towards perfect play, and most of the games have sequences with inhuman reading power, so that no longer seems to be a weakness. it will be difficult to find its mistakes.
I found the games inspiring on the one hand - I just won the challengers tournament and will play uberdude in the final of the British championship. I played some very fun games influenced by ideas from alphago. I was bold enough andrelaxed enough to be creative with it. and it worked very well especially with my background of familiarity with its style. my level probably will jump at least a stone from understanding it's ideas.
though on the other, individual ambition into high level go loses some of its meaning. it will be a collaborative effort to understand alphago's moves. It would be nice to see a movement to react against some of its style though, so there is some counterpoint.
I will work on the usual reviews of my games and then alphago self play in the summer as exams are next week. while I could be satisfied with master game reviews after about a week (I stopped because of the quality of Redmonds reviews), the self play reviews will take months or years with the depth and complexity.
It is more or less known for ages that the correct komi is 7.0, so there is no surprise in that department. As for AlphaGo, what I don't understand is why retire before playing the obvious last step: handi games against top pros.
moha wrote:... As for AlphaGo, what I don't understand is why retire before playing the obvious last step: handi games against top pros.
What will be the benefit for professionals to play handicap games with regard to their "usual" job (that consists on even games) ?
Knowing and understanding more about the game maybe? Such as when playing in areas where the opponent has more stones, or the reverse? Also, since AlphaGo strongly positioned itself as a research project, it would be logical even if pros would not be that interested (but I'm sure they would).
moha wrote:It is more or less known for ages that the correct komi is 7.0, so there is no surprise in that department.
It has been more or less known by a relatively few players that practical komi for modern pros is around 7, for about 40 years. Even in dog years, that's hardly ages. And the Nihon Kiin only changed to a 6.5 komi in this century. There was a distinct possibility that a komi of 7.5 would favor Black in AlphaGo vs. AlphaGo games.
The Adkins Principle: At some point, doesn't thinking have to go on?
— Winona Adkins
Bill Spight wrote:There was a distinct possibility that a komi of 7.5 would favor Black in AlphaGo vs. AlphaGo games.
Sure, the possibility was there, esp since MC games are quite different from what we know so far, maybe komi-wise as well. Still the actual results agree with our current knowledge.
Bill Spight wrote:There was a distinct possibility that a komi of 7.5 would favor Black in AlphaGo vs. AlphaGo games.
Sure, the possibility was there, esp since MC games are quite different from what we know so far, maybe komi-wise as well. Still the actual results agree with our current knowledge.
If the possibility was there, then our current knowledge was not knowledge of correct komi.
The Adkins Principle: At some point, doesn't thinking have to go on?
— Winona Adkins
