< Earlier Kibitzing · PAGE 34 OF 34 ·
|Dec-11-14|| ||hoodrobin: Are you <truefriends> at <chesshere.com> too?
You must be a strong CC player to have reached the Final. I wish I were strong at CC or OTB... Well I suppose I like chess and that's all.|
|Dec-12-14|| ||hoodrobin: There's only one player from the U.S. and one from Australia. I guess you are the former.|
|Dec-12-14|| ||truefriends: I am neither... But my flag has the same colors as both ;-)|
|Dec-12-14|| ||hoodrobin: Are you from Netherlands? I was prepared to say that your English sounds very American. lol.
(or does it really sound American...?)|
|Dec-13-14|| ||truefriends: Nope... born and raised in Holland :-)|
|Dec-13-14|| ||hoodrobin: Hello, fellow European !
|Dec-13-14|| ||truefriends: Ciao :-)
But that's about all my Italian ;-)
|Dec-13-14|| ||hoodrobin: I do not even know a word of Dutch but I'm very fond of (at least) two Dutch <products>, namely V. van Gogh and Philips (electronics).|
|Dec-14-14|| ||truefriends: And I like almost everything out of your kitchen... even a simple pizza ;-)|
|Dec-14-14|| ||hoodrobin: Yes a pizza is a simple device, as simple as a board of 8x8 squares. But the possible combinations of flavors is almost infinite!
|Dec-15-14|| ||hoodrobin: BTW do you know this one?
|Dec-16-14|| ||cro777: Hi <truefriends>
Can White still win this? Probably not...
click for larger view
This position is a theoretical draw. Black has to play ...Kf4. (Of course, after ...Kd4 or ...Kf6 Black is lost).
The main line is:
1...Kf4 2.Kd3 Rd1+ 3.Kc4 Kg5
|Dec-17-14|| ||truefriends: I know... I already offered a draw in that game and my opponent agreed ;-)|
|Dec-17-14|| ||truefriends: If you have 2 simular engines, let's call them A and B.|
Engine A is sligthly stronger than engine B, but only supports 1 core.
Engine B supports multiple cores.
If you have multiple cores (let's say 2)... Which engine produces the most accurate analysis in the same amount of time?
|Dec-17-14|| ||MuzioFan: My guess would be the engine that supports multiple cores - running on 2 cores instead of 1 really does mean a speedup of almost 100%, which should outweigh the slight difference in strength. But important is that this is true only if the test of strength was done on a single core for both engines - if somebody else has tested the performance of engine A on one core against engine B on two cores and found A stronger then of course that answers your question.|
So: if the strength was tested on as many cores as possible then engine A, if it was tested on a single core then engine B.
P.S.: You asked for advice from spacebar masters, which I am far from. But I hope that they agree with what I have written above.
|Dec-17-14|| ||truefriends: < MuzioFan: My guess would be the engine that supports multiple cores - running on 2 cores instead of 1 really does mean a speedup of almost 100%, which should outweigh the slight difference in strength.>|
I did a little test an the speedup is about 50%...
|Dec-17-14|| ||AylerKupp: <truefriends> I am no spacebar master either but it is difficult to answer your question without some questions of my own.:|
1. What do you mean when you say that engine A is slightly stronger than engine B? That in an engine vs. engine competition engine A has a slightly higher winning percentage in a statistically significant number of games? And I would assume that if this is the case, this determination was done with engine A running on one core and engine B running on two cores, and that they are running on identical hardware.
2. What do you mean by the most accurate results? That one engine's evaluation of a statistically significant number of positions is more accurate than the other engine's? And how would you determine that the results are more accurate?
I elaborated in The World vs Naiditsch, 2014 that there is a difference between relative and absolute evaluations. There I indicated that to play a strong game it is only necessary that an engine produce accurate <relative> evaluations, but that for the type of analyses that we need in a game like the current game against GM Naiditsch we need an engine that produces accurate <absolute> evaluations in order to make correct voting decisions. But if engine A is considered to be stronger than engine B, that was probably the result of engine vs. engine competitions so engine A need not generate accurate <absolute> evaluations at all.
It is also difficult to determine how accurate an <absolute> evaluation is because the only judges we have are top-level human grandmasters (a different engine would not be of any use because we don't know how accurate its evaluations are, etc.). But human grandmasters don't think in terms of "equivalent centipawns" and, if they did, I doubt that their "resolution" would be better than 0.50 or so (half a pawn). So, if a an engine's evaluations <consistently> differ significantly from the evaluation of a top level grandmaster by a somewhat large amount (say more than one pawn), then that engine's evaluations can be considered to be somewhat inaccurate. But if the evaluations of 2 engines differ by less than 0.50 (assuming that my assumption about a grandmaster's effective evaluation resolution is correct), then a top-level human grandmaster could not determine which of the 2 engines produce the more accurate evaluation). At, at any rate, you would probably need a team of top-level grandmasters rather than just one to have confidence in the human evaluation.
The above is a very round about way of saying, as is usual for me, that I don't know how to answer your question. But I did want to point out that very often an engine running on one core is stronger than an engine running on 4 cores, as the engine ratings in, say, the CCRL engine tournaments indicate. See for example http://computerchess.org.uk/ccrl/40..., Complete list (Dec-14-14), where single-core 64-bit Komodo 8 and Stockfish 5 were ranked in 3rd place with ratings of 3221 and 3205 respectively while Gull 3.0 4CPU 64-bit was ranked in 4th place with a rating of 3189 along with Houdini 4 single-core 64-bit with a rating of 3187. 4CPU 64bit Komodo 8 was ranked #1 with a rating of 3303 and 4CPU 64-bit Stockfish 5 was ranked #2 with a rating of 3283, indicating that quadrupling the number of cores does not increase rating points all that much (about 80 Elo rating points), at least for these 2 engines, although this rating differential amounts to a P(Win) of about 61% which I consider significant.
|Dec-17-14|| ||truefriends: < AylerKupp: <truefriends> I am no spacebar master either but it is difficult to answer your question without some questions of my own.:|
1. What do you mean when you say that engine A is slightly stronger than engine B? That in an engine vs. engine competition engine A has a slightly higher winning percentage in a statistically significant number of games? And I would assume that if this is the case, this determination was done with engine A running on one core and engine B running on two cores, and that they are running on identical hardware.>
Just to keep it simple.
Let's asume engine A is a later version of engine B with just a few improvements.
So all other variables are exactly the same, including the hardware used.
|Dec-21-14|| ||cormier: usually the B is faster ...|
|Dec-22-14|| ||truefriends: < cormier: usually the B is faster ...>|
Yes, as I mentioned... B is about 50% faster. But the question was which one was stronger ;-)
|Dec-22-14|| ||truefriends: Just won (and sadly also lost) my first game in the Final!|
1. e4 c5 Nf3 e6 3. Nc3 Nc6 4. d4 cxd4 5. Nxd4 Nf6 6. Nxc6 <Ndb5 is played more often, but this move scores better> bxc6 e5 Nd5 8. Ne4 Qc7 f4 Qb6 10. c4 Ne3 11. Qd3 Bb4+ 12. Bd2 <All just book moves, now 0-0 is the most played move> Ba6 13. Bxb4 Qxb4+ 14. Kf2 Qxb2+ 15. Be2 Nxc4 16. Rhd1 Qb6+ 17. Qd4 <Still all known, but now my opponent plays a new move> Nb2 18. Bxa6 Nxd1+ 19. Rxd1 Qxd4+ 20. Rxd4 Rb8 <All forced moves. Now white has to think of a plan. I liked the next 3 moves, totally tying black's pieces down> 21. Bb7! Ke7 22. Nc5! Rhd8 23. Rd6! <The next step in whites plan is to block the Kingside pawns> h6 24. Ke3 Ke8 25. h4 Ke7 26. h5 f6 27. g4 fxe5 28. fxe5 <Now black can only wait with Ke8-e7 etc. for white to push his a-pawn to guard his Bb7> Ke8 29. Kd2 Ke7 30. Kc3 Ke8 31. a4 Ke7 32. a5 Ke8 33. a6 <Black can't play a6 or a5 because white gets a passed pawn or can take on d7> Ke7 Kd4 <The black King must be on e8 when white takes on d7> Ke8 35. Nxd7 Rxd7 36. Bxc6 1-0
|Dec-25-14|| ||hoodrobin: Best wishes! Hold on!|
|Dec-25-14|| ||wordfunph: <truefriends> Merry Christmas!|
|Dec-26-14|| ||cormier: Joyeux NoŽl ...|
|Jan-09-15|| ||truefriends: Still on 50% with 5/10 games finished.
After the 2 game I had 1 book draw (black) and 2 well played draws of about 40 moves (w&b).
Still 2 openings (w&b), 2 middlegames (both w) and 1 endgame (b) left.
Any interest in the games?!
< Earlier Kibitzing · PAGE 34 OF 34 ·