How important is an opening database to a computer?

[MikeB]

Kraminik back in 2006 was only rate around 2750 FIDE and Deep Fritz was rated 3110 by CCRL and t Kramnik almost drew the match, therefore there is a difference of 300 rating discrepancy that are very close to Humans FIDE rating and CCRL. I strongly believe that if Magnus ever decide to play vs a computer DF would be a perfect match that would be very close.

That is not quite accurate. Fritz was almost 2850 then.

From wayback:

Screenshot 2018-10-15 00.01.58.png

[jdart]

GM Kramnik will more likely beat Deept Fritz since there has been several opening lines and variations improvement in the last 12 years

I disagree. Opening prep is very important for top-level humans playing other humans. But it is not so helpful when a human is playing a computer. Once in a while someone will find a "super-novelty" that gives a practically winning position, but that is quite rare. More commonly, even good opening prep only gets you to a position that is even or where you have a slight advantage. Then, against a computer, you have to play the entire rest of the game without making any serious error. Empirically, that is very hard for even GMs to do. Sometimes they can: but even high level games frequently contain errors and some of these are serious; a human opponent might overlook them, but computers will find and exploit these.

--Jon

[Chessqueen]

Kraminik back in 2006 was only rate around 2750 FIDE and Deep Fritz was rated 3110 by CCRL and t Kramnik almost drew the match, therefore there is a difference of 300 rating discrepancy that are very close to Humans FIDE rating and CCRL. I strongly believe that if Magnus ever decide to play vs a computer DF would be a perfect match that would be very close.

That is not quite accurate. Fritz9 was almost 2850 then.

From wayback:

Screenshot 2018-10-15 00.01.58.png

I see that Uri posted a CCRL 40/40 tested with a X2 4600+ which is rated 2831 with that hardware and Mike Posted a CCRL 40/40 dated back to July 8, 2006 for Fritz9 NOT Deep Fritz 10 for Multi processors, but both are using different hardware than what was used Versus Gm Kramnik match Deep Fritz used an intel Xeoan CPUs DC5160 with dual core 3 GHz capable of evaluating 8 millions moves per second,which is way faster than both systems used by CCRL that both of you posted that could be a difference in rating point of at least 50 points for those system that you posted Deep Fritz 10: it is a question of hardware in direct proportion to speed and rating https://en.chessbase.com/post/deep-frit ... n-of-speed. And GM Kramnik almost came out with a score of 3.0 to 3.0 if it wasn't for overlooking a simple mate in one. it is true that GM Kramnik receive a copy of the beta version of Deep Fritz but without the actual opening that DF team develop for the match which included an updated opening book that cover most of GM previous opening ever played by him an analyzed in detail by Deep Fritz.

PS: Back when Alekhine was champion there were no computer, and we can not compare champions from the past to current champions, but now that we have computer that played better than any champion at least we can go back to a computer like Deep Fritz 10 that played so close to GM Kramnik and compare it with future champions 20, 40 or 100 years from now. To bad that GM Carlsen does not like to play versus computer, since Deep Fritz 10 thast is very close to his rating around 2850 to 2875 using the same hardware used against Kramnik NOT the one used to test CCRL could be used in the future when Carlsen is dead 75 years or over to compare it versus other future champions https://en.chessbase.com/post/deep-frit ... n-of-speed

PS: Note I made an error to estimate DF at around 3110 is is more closer to (2850 to 2875) using that same hardware used against GM Kramnik.

[Chessqueen]

GM Kramnik will more likely beat Deept Fritz since there has been several opening lines and variations improvement in the last 12 years

I disagree. Opening prep is very important for top-level humans playing other humans. But it is not so helpful when a human is playing a computer. Once in a while someone will find a "super-novelty" that gives a practically winning position, but that is quite rare. More commonly, even good opening prep only gets you to a position that is even or where you have a slight advantage. Then, against a computer, you have to play the entire rest of the game without making any serious error. Empirically, that is very hard for even GMs to do. Sometimes they can: but even high level games frequently contain errors and some of these are serious; a human opponent might overlook them, but computers will find and exploit these.

--Jon

It still matter, imagine Deep Fritz 10 playing against GM Kramnik in 2006 without an opening book at all, probably it would had ended 4 to 2 in favor of GM Kramnik.https://en.chessbase.com/post/man-vs-ma ... ate-in-one Do you ever wonder why the team of Deep Frit programmers went thru all the hard work of collecting all the games of GM Kramnik and analyzed in detail with DF and developed a special book which was not given to GM Kramnik two months prior with the Beta version of Deep Fritz10.

[yurikvelo]

I think opening books (those readily available) have inferior quality and slightly benefit in "Machine vs Machine" match only for bullet-timecontrol due to time-savings, not due to good moves. Book is "no blunders at zero time cost", that's why it benefit STC.

For LTC, I believe book-less engine (latest SF, Komodo or Houdini) will beat itself, provided by user-supplied inferior opening book.

For weak engines (e.g. Deep Fritz 10 from 2006), opening books are definitely good both is LTC and STC

[yurikvelo]

[d]rn2kb1r/pp2pppp/2p2nb1/q7/2BP2P1/2N2N1P/PPP2P2/R1BQK2R w KQkq - 1 9


SF-Dev

Code: Select all

D=44, 61 BN
+1,55	9.h4 Nxg4 10.Bd2 e6 11.h5 Bf5 12.Qe2 Na6
+1,37	9.Bd2 e6 10.h4 Nxg4 11.h5 Bf5 12.Nh4 Nf6
+1,28	9.Qe2 e6 10.Bd2 Bb4 11.0-0-0 Nd5 12.Nxd5 Bxd2+
+0,84	9.Nh4 e6 10.Bd2 Qc7 11.Nxg6 hxg6 12.Qf3 Nbd7
+0,64	9.0-0 e6 10.Re1 Be7 11.Ne2 Nbd7 12.Nf4 Nd5
+0,53	9.Ne5 Nbd7 10.Nxd7 Nxd7 11.f4 Nb6 12.Bb3 e6	-------- opening book suggestion
+0,45	9.Bb3 e6 10.Bd2 Qa6 11.Ne5 Nbd7 12.h4 Nxe5
+0,44	9.a3 Ne4 10.Bd2 Nxd2 11.Qxd2 Nd7 12.0-0-0 e6