Author: Dann Corbit
Date: 20:26:12 01/29/03
Go up one level in this thread
On January 29, 2003 at 18:59:28, Angrim wrote: >On January 28, 2003 at 15:18:43, Dann Corbit wrote: > >>On January 28, 2003 at 08:05:57, Omid David Tabibi wrote: >> >>>As someone mentioned after the game, it is hard to imagine Junior losing in just >>>27 moves, had it not used the opening book. Today, the top programs already play >>>in a super-Grandmaster level (well, that doesn't include Kasparov of course), so >>>why should they blindly play variations played by players weaker than them? >>> >>>Of course, turning the opening books off totally is not a viable option, as the >>>programs still don't have the needed strategic vision to find their way early in >>>the opening phase. But maybe a stricter limit (depending on type of opening, >>>games played, statistics, etc) should be imposed for choosing moves right out of >>>the opening book. >> >>The problem without book is worse than the problem with book. Steer a program >>into the Evans Gambit without a tuned book and a GM will tear it to shreds (for >>instance). >> >>Opening positions are quiet positions where it will take a while for action to >>develop. These are among the positions where computers perform the worst. >> >>Here is a fault which is easily corrected and I am astonished that it has not >>been performed. >> >>A book is a dense object with many, many lines of action. However, compared to >>the internal nodes, the exit points from the book are a very small fraction of >>the book size. Every commercial book should analyze every single exit position >>on a fast machine for ten minutes. Then, there will be no such thing as falling >>out of the book and into a bad surprise. >> >>If any bad positions are found, the engine should backtrack until the position >>is no longer bad. >> >>In other words, we need to check the perimeter of the book. There can still be >>internal problems where strategic or tactical moves are missed. But most of the >>problems are not like that. > >The perimiter(leaf nodes) of the book is actually a fairly large fraction >of the total nodes, since many of the internal nodes are shared >between many lines. This is a fairly minor point though. > >The idea to have computers verify their own books, insuring that they >do not play to a position that they feel is losing in book has been >brought up before. It has usually been rejected as taking too much >computer time, or else as having been tried in cray blitz and not >having worked well there. Neither of these points really bothers me.. > >I would take the very large book, estimate 1meg lines, and prune it with >the rule that a position is only important if it has been in multiple games, >likely giving roughly 1meg positions. Then backsolve the whole tree >at 10 minutes a move using a strong engine. I would not discard lines >based on their containing blunders, but would just mark the blunders as >being moves to avoid. It could be handy to have those lines in book >so that you have the refutation handy if the opponent makes that blunder. >This search would cost 10 meg minutes to compute. >10,000,000/(365days*24hours*60min)= 19 years. if you split the search up >between 38 computers it would only take 6 months. >Clearly you would not want to repeat this search very often. It would >likely be best to fix which engine was used for the search, and use >that for all positions, until a really major improvement to the engine's >strength was made, at which time you start a new search. >Also, the ability to maintain the resulting database would be quite >important, you should be able to add a new line without re-searching >everything. > >note: the difference between "backsolve the whole tree" and "search each >position in the whole tree" is vital. Without knowing the searched value >of the leaf nodes, the computers ability to evaluate the earlier opening >moves is much weaker. With the CAP data (assuming that 100 million node crafty searches are good enough) you would only have to solve the missing ones. There are many millions of positions at long time control. Chances are good that 90% or better of the interesting positions are in there. We also have hundreds of millions at fast (10 second) time control. If you minimax the whole thing, it might be better.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.