Here's a survey I made using the Cerebellum demo at zipproth.com.
Hopefully it's self-explanatory. All scores are from white's perspective, rounded to the nearest centipawn. Parenthetical remarks to the left of the scores name the openings, and those to the right describe subsequent best play.
This is just a survey -- an incomplete look at some main lines.
I'm fascinated by the possibility of solving chess. Based on estimates from Regan and others, it seems top engines are within 400 Elo of perfect play (and probably closer to 200). And cloud computing has gotten really cheap.
And the game appears not only to be drawn, but to contain tremendous drawing chances. It's possible that all positions scored from -0.5 to 0.5 are drawn, for example. That sort of thing makes a weak solution easier.
While it looks to me that Cerebellum (at least the demo) isn't there yet, some kind of weak solution may be in reach of a similar project with a bigger budget. Zipproth has stated Cerebellum represents about $2K of electricity. So perhaps $1M would do the trick. That's the prize fund for a single top tournament.
* Cerebellum favors simple, classical play. It doesn't like tricky-looking moves.
* It likes to exchange pawns in the center sooner than humans. It favors the Tarrasch in the French and the exchange Slav, for example.
* I'm a bit doubtful best play for both sides results in the French defense. The position out of the Tarrasch may not have been analyzed deeply enough in Cerebellum.
* Zipproth has stated (on another forum) that he considers the 0.17 score for the Marshall an error, and thinks the Spanish may not be any better than the Italian game.
* Fischer's "bust" to the King's gambit works. The Falkbeer counter-gambit produces equality.
* Cerebellum tries to transpose to Taimanov main lines several times before getting into the Sveshnikov. Top humans have played the Sveshnikov a lot for an opening that can only be reached by making several inferior moves in a row.
* It looks like black can avoid some disadvantageous lines in the Pirc by playing ...g6 before ...Nf6.
* The Nimzo-Indian works. (No surprise.) The Catalan doesn't. (I'm a bit surprised.)
* If he can't get a Nimzo-Indian, black should shoot for a semi-Slav. Many lines want to transpose to it.
* The Grunfeld may not be as good as Larry Kaufman thinks.
* The Budapest gambit is better than the Benko gambit?
* After 1.d4 d5, many lines want to be an exchange Slav. But the best line is the Marshall gambit in the semi-Slav. Here again we have central pawns exchanged and I wonder if the position has been analyzed deeply enough.
syzygy wrote:Cerebellum and current top engines have little to do with solving the game. Solving the game is about obtaining mathematical certainty about the outcome of the game.
Incorrect. Such an approach can be used to *weakly* solve a game.
* Bb5 lines (Moscow and Rossolimo) come out significantly worse than main lines.
* In the Najdorf, Cerebellum likes Be2 better than the English attack, where black will respond ...e5 after either Be3 or f3. White is not at all afraid of ...Ng4 and is quite happy to bring his bishop around to g3.
You have forgotten that the values of Cerebellum based on Stockfish only and supposing that the perfect play leads to draw is merely a speculation.
Moreover the idea of "perfect play" has not a well established definition.
clumma wrote: Zipproth has stated Cerebellum represents about $2K of electricity. So perhaps $1M would do the trick. That's the prize fund for a single top tournament.
-Carl
I have 2 very constructive and forward-looking suggestions:
- move the project headquarters from Germany to the US, where electricity costs will be much lower;
- start a donation subscription on this forum, with a minimal contribution requirement of 1000 bucks;
clumma wrote:Here's a survey I made using the Cerebellum demo at zipproth.com.
Hopefully it's self-explanatory. All scores are from white's perspective, rounded to the nearest centipawn. Parenthetical remarks to the left of the scores name the openings, and those to the right describe subsequent best play.
This is just a survey -- an incomplete look at some main lines.
Cerebellum is arguably the strongest opening book ever. Its strength lies in the memorized best value score of a given move. If it has not seen that move before then it behaves like a normal Stockfish chess engine/book.
With absolute certainty, the unchanging first move of Cerebellum as white is e4 and its favorite variation is the Giuoco Piano. With black against e4, it plays the constant e5.
So, it is possible to find ways to neutralize by thorough preparation and accumulation of the best move sequence against these. This is both a strength by having a large collection of store best move and at the same time a weakness as other book makers could assimilate that large best move collection and then use it against Cerebellum with the injection of more best moves that has not been memorized.
In theory, chess games should end a draw with correct play. Good opening books makes it easier to play correctly. The best opening book actively maintained thus far is Cerebellum but may not sustain its perceived strength when somebody else will have the focus to create an alternative book that will neutralize it.