On Tue, Jul 12, 2011 at 12:01 PM, Brian Sheppard <[email protected]> wrote:
> >I think 9x9 go, even though compared to chess in complexity, is still > more complex than chess and that the book will have a little less impact, > although still a lot.**** > > ** ** > > ** ** > > My projection is the opposite: I think that 9x9 will be "played out" within > 5 years. Not weakly solved, exactly, but close to it. Zen and CrazyStone > have the ability to start on that project already. > I'm not in a good position to know as I'm just trying to interpolate from what I know about chess. So maybe on 9x9 Go is indeed much simpler than chess and is much closer to being solved. In Go, the branching factor starts at 81 but in general tends to be reduced after every move. This is true of chess but not as much so. In fact it goes up for several moves before it starts to slowly reduce and does not necessarily reduce that much. But it's not just about branching factor either as shown by the opening book situation - what is probably most important is the "relevant branching factor" which considers how easy or difficult it is to identify plausible moves and how many there are. So in order for a book to be really effective it has to have responses to the majority of plausible moves to a depth which takes you past the most dangerous parts of the game. Which gives us another clue. In Go, the game tends to clarify with each passing move (as a generalization), in chess that is not so much the case - unless one side obtains a winning advantage or a draw becomes relatively simple. Also, in 9x9 go the game is extremely unlikely to go past 5 or 60 ply without the result being obvious. Not true of chess. So maybe it's really 11x11 go that should be compared in complexity to chess. This could be tested on CGOS by putting two identical programs that are also provisioned identically, the only difference being that one has a high quality opening book. You cite Valkyria but is this a comprehensive study with strictly defined controls or just anecdotal evidence? Don > **** > > ** ** > > My impression is that the opening books are routinely worth a few hundred > rating points in 9x9 CGOS.**** > > ** ** > > I would cite Valkyria, which has a version that is playing near the top of > the CGOS ladder most of the time. A comparable version was playing ~200 > rating points within the last year, and I suspect that the opening book > knowledge that comes from its long-term memory is the dominant contributor. > **** > > ** ** > > I also cite the Little Golem server, which is dominated by programs that > have opening books.**** > > ** ** > > Based on the work of Mogo and Valkyria, I suspect that if you take a pretty > good player and create a feedback system then you get a great opening book. > With an effective branching factor of maybe 2 to 3, you can get pretty far > into the game.**** > > ** ** > > Brian**** > > ** ** > > ** ** > > *From:* [email protected] [mailto: > [email protected]] *On Behalf Of *Don Dailey > *Sent:* Tuesday, July 12, 2011 12:39 AM > > *To:* Aja; [email protected] > *Subject:* Re: [Computer-go] 19x19 opening books**** > > ** ** > > ** ** > > On Tue, Jul 12, 2011 at 12:17 AM, Aja <[email protected]> wrote:**** > > Hi Erik, > > Thanks for the clarification. > > In the last Olympiad in Kanazawa, David (Fotland) and me, we both thought > that 9x9 Go now becomes a race of book.**** > > ** ** > > You guys know more about go than I do, but I have a feeling that you are > just discovering that the opening book can make a big difference. In > computer chess a decade or two ago it was know how important opening books > were and people were even saying, "it's now all about the book", but that > has been proved to be quite wrong. Given 2 equal programs however a > superior book will give a nice edge. **** > > ** ** > > I expect that is what will happen with 9x9 go too, but of course I could be > wrong. I think 9x9 go, even though compared to chess in complexity, is > still more complex than chess and that the book will have a little less > impact, although still a lot. On the other hand, I might be mistaken and > perhaps it's even more important than in chess. **** > > ** ** > > Don**** > > **** > > ** ** > > **** > > There is an interesting "opening book game" between Erica and Mogo, > > http://www.grappa.univ-lille3.fr/icga/round.php?tournament=215&round=4&id=11 > If I remember correctly, they were both in the book until the 36th moves. I > added that line just before the tournament (from Mogo's games). 36 moves in > the book might be a record. :) > > Mogo forgot to remove that line, so Erica won. But in the second game Mogo > was again in the book for many moves and secured a solid win. > > Aja > > ----- Original Message ----- From: "Erik van der Werf" < > [email protected]> > To: "Aja" <[email protected]>; <[email protected]> > Sent: Tuesday, July 12, 2011 1:33 AM > Subject: Re: [Computer-go] 19x19 opening books**** > > > > > Hi Aja, > > On Mon, Jul 11, 2011 at 4:47 PM, Aja <[email protected]> wrote:**** > > Hi Erik,**** > > Without a book Steenvreter usually still plays decent opening moves > because it is pretty good at predicting pro moves (predictions are > used to set priors in the tree). However, with longer thinking times > there is sometimes a tendency for its style to become more cosmic (not > sure if that is good or bad).**** > > > On 9x9 or 19x19 do you mean?**** > > > I used to have a different predictor for small boards, but now I just > use the one trained on 19x19; it also works quite well on 9x9. > Actually I was quite surprised, some years ago, when I found that > something trained on 19x19 made the 9x9 program stronger... > > Of course for 9x9 the move predictor alone is not good enough. Sadly, > to win tournaments it seems having a good book is now a must. > > Erik > _______________________________________________ > Computer-go mailing list > [email protected] > http://dvandva.org/cgi-bin/mailman/listinfo/computer-go**** > > ** ** > > _______________________________________________ > Computer-go mailing list > [email protected] > http://dvandva.org/cgi-bin/mailman/listinfo/computer-go >
_______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
