< Earlier Kibitzing · PAGE 2 OF 2 ·
|Feb-28-08|| ||daniellovesbeer: If 14... Bxc4 15 Qc2 g6 16 Qd2 Kh7 17 Bf6 Nb3 18 Ng5# Kg8 19 Qf4 Nxa1 20 Qh4 white wins|
|May-27-09|| ||Crocomule: Of course Keres would have seen 16... Qe8, and rejected it because of 17. Nh5.|
|May-27-09|| ||DrGridlock: Rybka scores it:
16 ... Qe8 (.41)
17 Nh5 Qe7 18 Bf2 Cxc4 19 h4 Nb3
16 ... gxh4 (.48)
17 Nxg6 Re8 18 Nh8 Re7 19 Qh7+ Kf8 (Rybka agrees with the game continuation).
Rybka has a slight preference for Qe8, but gxh4 was not a decisive mistake.
|Apr-26-12|| ||zydeco: Tim Krabbe puts 16.Nf4 at #46 on a list of the most fantastic chess moves ever played. It seems like Kotov had it in mind when he played 12.Bg5; otherwise white's just getting pushed back. The year before, Geller, with a similar piece alignment (...g5 hadn't been played) played Nf4, inviting the ...g5 pawn fork, with the idea that the knight would jump to h5 and f6. Maybe that's where Kotov got the idea. |
For some crazy reason, I've been playing over all the games from this tournament, and I'm finding that I'm having the hardest time predicting Kotov's moves: he has these strange elaborate maneuvers out of the opening and then goes in for fantastically sharp variations in the middlegame.
Maybe it's part of his calculating style that he avoids natural moves and instead hunts for a move that will really give him something tangible. (Bg5 instead of castling is a good example). He can obviously beat anyone, but every so often he'll get crushed in 20 or 30 moves - usually, I think, because, in inferior postions, he doesn't see an immediate loss and plays a little too aggressively without realizing how much trouble he's in.
|Dec-21-12|| ||An Englishman: Good Evening: I believe that Soltis pointed out that 13...d5 was much worse than 13...d6. Unfortunately, I can't find the reference at this time.|
|Dec-21-12|| ||sofouuk: seeking alternatives to the game, rybka notes after 15 Qc2:|
click for larger view
15...f5! 16.exf6 Nf5 17.Bg3 Qxf6 18.Be5 Qe7 19.cxd5 exd5 20.Kf2 ~=
click for larger view
instead after 15...Ng6 16.Nf4 Qe8 17.Nh5 Qe7 18.Bf2 Bxc4 19.h4 Nb3 20.Ra2 Na5 21.Nf6+ Kg7 22.hxg5 Bxa2 23.gxh6+ Kh8 24.Bxa2 Qxa3 25.f4 Nc4 26.f5
click for larger view
white has invested the exchange for a strong attack. not clearly winning tho
|Dec-21-12|| ||Naniwazu: Seems like 20. Bg6 would work just as fine..fxg6 21. Nxg6+ Ke8 22. Nxe7 Qxe7 23. Qg8+ Qf8 24. Qxe6+ Qe7 25. Qg8+ Qf8 26. Qxd5 .|
|Dec-21-12|| ||morfishine: <Naniwazu> Your line looks very convincing|
|Dec-21-12|| ||Kikoman: h6 and g5 pawn moves weakened Black kingside.|
|Dec-21-12|| ||Garech: Wow, Keres played this very poorly - remarkable as he was such a talent.|
|Dec-21-12|| ||DrGridlock: Interesting to see that I had posted in 2009 a computer analysis of the position at black's move 16. I have switched search engines in the past 3 years, and decided to look again at the position with Komodo (instead of Rybka). Komodo finds a different analysis of the position:|
click for larger view
Analysis by Komodo32 3 32bit:
1. ± (0.79): 16...Qe8 17.Nh5 Qe7 18.Bf2 Bxc4 19.h4 Nb3 20.Ra2 Na5 21.Nf6+ Kg7 22.hxg5 Bxa2 23.gxh6+ Kh8 24.Bxa2 c5 25.dxc5 Rfc8 26.cxb6 axb6 27.0-0 Nc6 28.f4 Rxa3 29.Bb1 Nb4 30.Qd2 Rcxc3 31.f5 exf5 32.Bxf5
2. ± (1.26): 16...Qd7 17.Nxg6 fxg6 18.Qxg6+ Qg7 19.Qxe6+ Qf7 20.Qxf7+ Rxf7 21.Bg3 Nb3 22.Ra2 Bxc4 23.Rb2 Nc1 24.Kd2 Ne2 25.Re1 Nxg3 26.hxg3 Kg7 27.Bd3 Bxd3 28.Kxd3 Re7 29.g4 Re6 30.a4 Rd8 31.Rbb1 a6 32.Re2
3. ± (1.28): 16...gxh4 17.Nxg6 Re8 18.Nh8 Kf8 19.Qh7 Re7 20.f4 Bxc4 21.f5 Ke8 22.f6 Kd7 23.Nxf7 Qg8 24.Qxg8 Rxg8 25.fxe7 Kxe7 26.Nxh6 Rxg2 27.Kd1 c5 28.dxc5 bxc5 29.a4 Nc6 30.Re1 Rxh2 31.Bc2 Rh3 32.Rb1
Komodo finds a greater game value to white (.79 vs .48) and much prefers Qe8 (.79) to gxh4 (1.28).
I re-ran the Rybka analysis, and got pretty much what I had found in 2009. The difference between the two lines is Rybka's difficulty in finding 20 f4 for White (Rybka preferred Bg6 instead at White's move 20).
Komodo is the stronger of the two engines, and it took Komodo to find:
(i) the strength of White's position at move 14
(ii) White's continuation 20 f4.
A lesson in search engine position searches.
|Dec-21-12|| ||Eduardo Bermudez: Probably Kotov was one of the most powerful players at early fifties with white pieces !|
|Dec-21-12|| ||kevin86: Black is down space,material,and position. Little wonder that he would resign.|
|Dec-21-12|| ||keypusher: <kevin86: Black is down space,material,and position. Little wonder that he would resign.>|
More to the point, White threatens Qb5#, and if 33....Nxa3, then 34.Rxb6+ Kxb6 35.Qc6#.
|Dec-21-12|| ||AylerKupp: <<DrGridlock> A lesson in search engine position searches.>|
It is very rare for two engines, or for that matter the same engine running an analysis on the same position at a different time, to come up with the same evaluation and the same lines. Not only does each engine have a different evaluation function, but they prune their search tree differently, likely using different heuristics. And, even when using the same engine, the non-deterministic nature of chess engines, particularly using multi-core computers, will likely lead to a different evaluation when run at different times, even if run to the same depth, particularly if the hash table is cleared between runs.
That's why, if I have the luxury of time, I use different engines, at least 3, to analyze the same position. Then I average each of their evals (I typically run with MultiPV=3 or =5) to try to eliminate some of each engine's evaluation biases and try to get a better absolute eval of the position.
I'm not sure that Komodo is a stronger engine than Rybka although I think that they're very close. In the latest CCRL 40/40 engine tournament (Dec-08-12) Rybka 4 64-bit, 4 CPU is ranked #3 with a rating of 3203 (Rybka 4.1 is rated marginally lower at 3201) and Komodo 5 64-bit is ranked #7 with a rating of 3158. In the latest CEGT 40/120 engine tournament (Dec-17-12) their rankings are reversed, Komodo 5 64-bit is ranked #2 with a rating of 2975 and Rybka 4 64-bit (but with only 1 CPU) is ranked #8 with a rating of 2943. That's not surprising since when only single-CPU engines are used Komodo is consistently ranked #2 and rated higher than Rybka.
Of course, these rankings and ratings are for game-type situations with time management and other factors included. Analysis situations are different and an engine that is stronger in game-type situations may not be stronger in analysis-type situations; we just don't know.
I don't know about you but I'm eagerly waiting (and waiting, and waiting) for the multi-CPU version of Komodo 5 (or 6, or 7), which is supposedly coming out "soon". Of course, by that time Rybka 5 (or 6, or 7) will probably be released and we'll have to do the comparison all over.
|Dec-21-12|| ||gars: What is the pun about?|
|Dec-21-12|| ||AylerKupp: <gars> Try "Another Coat of Paint".|
|Dec-21-12|| ||DrGridlock: <AylerKupp> The Rybka I used in 2009 was Rybka 2.2. The Komodo I currently use is version 3. Neither is the most recent version, but the Komodo is more recent which makes it stronger than the older Rybka.|
What interests me is the type of positions that pose problems to search engines - and this game is apparently one of them.
|Dec-21-12|| ||Rama: 33. ... Qe8?|
|Dec-21-12|| ||Octal: Yes, but how does "another coat of paint" relate to the game?|
|Dec-21-12|| ||AylerKupp: <Octal> It doesn't. chessgames.com doesn't let applicability to the game get in the way of a good (or bad) pun.|
|Dec-21-12|| ||AylerKupp: <DrGridlock> Oh, I didn't think of that. Duh. I'm sort of an engine junkie so I've been keeping track of CCRL and CEGT engine tournaments for about 2 years. In the earliest CCRL list I have (Oct-10-11) Rybka 2.1 64-bit, 4 CPU was rated 3085 and Rybka 1.2 64-bit (and presumably 1 CPU) was rated 2998. So if you used Rybka 2.0 it was probably rated somewhat between these 2 versions.|
At that time Komodo 1.2, 64-bit was rated 3056, closer to Rybka 2.1 than to Rybka 1.2. The earliest reference to Komodo 3, 64-bit was in the Sep-09-11 where it was rated 3229 (Rybka 4.1 was rated 3271). But, like rating players from different eras, it probably doesn't make much sense to compare engine ratings from different tournaments since the ratings are relative to the other engines in the tournament, and there are different versions of each engine entered in the various tournaments.
Out of curiosity I tried analyzing the position after 15.Qc2 with Critter 1.6a, the latest version, and at d=25 it gave the following evals for its top 3 lines:
1. [-0.04]: 15...f5 16.exf6 Nf5 17.g4 Qxf6 18.Rg1 Kh8 19.gxf5 exf5 20.Nxg5 Rae8+ 21.Kd1 Nxc4 22.Qf2 Ne3+ 23.Kd2 Nf1+ 24.Rxf1 Re2+ 25.Qxe2 Bxe2 26.Kxe2 hxg5 27.Bf2 Qd6 28.Bd3 Kg7 29.Rg1 Re8+ 30.Kf1 Qf4 31.Rg3 Rh8 32.Kg2 Kf6 33.Rd1 Qh4 34.Bg1 f4 35.Rg4
2. [+0.61]: 15...Ng6 16.Nf4 Qe8 17.Nh5 Qe7 18.Bf2 Bxc4 19.h4 Nb3 20.Ra2 Na5 21.hxg5 Bb3 22.Nf6+ Kg7 23.Rxh6 Rh8 24.Rxg6+ fxg6 25.Qxg6+ Kf8 26.Ra1 Rh1+ 27.Ke2 Bd1+ 28.Ke3 Qg7 29.Qxg7+ Kxg7 30.Bd3 Rah8 31.f4 c6 32.f5 exf5 33.e6 Nc4+ 34.Bxc4 dxc4 35.Bg3 b5 36.e7 Kf7 37.e8Q+ Rxe8+ 38.Nxe8 Kxe8 39.g6 Kf8 40.Be5
3. [+3.56]: 15...Nf5 16.g4 Qd7 17.gxf5 exf5 18.Rg1 f6 19.cxd5 Bc4 20.Bxg5 fxg5 21.Nxg5 Kh8 22.Ne6 Bxd5 23.Nxf8 Rxf8 24.Qd2 Qc6 25.Ba2 Nc4 26.Bxc4 Bxc4 27.Kf2 a5 28.Qf4 Kh7 29.a4 Bd5 30.Rg2 Rf7 31.Qg3 Rd7 32.Rag1 Bb3 33.Qg6+ Qxg6 34.Rxg6 Bxa4 35.Rf6 Bb3 36.Rgg6 a4 37.Rxh6+ Kg8 38.e6 Re7 39.Rfg6+ Rg7 40.Rxg7+ Kxg7
So Critter 1.6a agrees with <sofouuk>'s Rybka's analysis of 15...f5 and it does not consider 15...Ng6 as played by Keres totally losing, although it is certainly to White's advantage.
I also had Critter 1.6a analyze the position after 16.Nf4 and these were its evals for its top 3 lines, also at d=25:
1. [+0.68]: 16...Qe8 17.Nh5 Qe7 18.Bf2 Bxc4 19.h4 Nb3 20.Ra2 Na5 21.hxg5 Bxa2 22.Bxa2 Nc4 23.Bxc4 dxc4 24.Qc1 c5 25.gxh6 Kh7 26.Be3 cxd4 27.Nf6+ Kh8 28.Bxd4 Qb7 29.Kf2 Rfd8 30.Ne4 Qe7 31.Qd2 Qxa3 32.Nd6 Rd7 33.Nxc4 Qe7 34.Nd6 f6
2. [+1.18]: 16 ..Qd7 17.Nxg6 fxg6 18.Qxg6+ Qg7 19.Qxe6+ Qf7 20.Qxf7+ Rxf7 21.Bf2 Nxc4 22.Bd3 Raf8 23.Bxc4 Bxc4 24.Be3 Kg7 25.Kd2 Bb5 26.h4 g4 27.fxg4 Bd7 28.g5 h5 29.a4 a5 30.Kc2 Bf5+ 31.Kb3 Be4 32.Ra2 Kg6 33.Rb2 c6 34.Ra1 Re7 35.Re1 b5
3. [+1.69]: 16...gxh4 17.Nxg6 Re8 18.Nh8 Re7 19.f4 Nxc4 20.Qh7+ Kf8 21.Rf1 Ke8 22.Nxf7 Rxf7 23.Bg6 Qe7 24.f5 Nxe5 25.dxe5 Bxf1 26.Kxf1 Qf8 27.Rd1 Rd8 28.Kg1 Ke7 29.Bxf7 Qxf7 30.f6+ Ke8 31.Qxh6 c5 32.Qxh4 Kd7 33.h3 a5 34.Qg4 Kc7 35.Qe2 Qh7 36.Qf3 Rg8 37.g4 Rf8 38.Kg2 Qc2+ 39.Kg3 Qa2 40.f7 Qxa3 41.g5 a4
So Critter 1.6a somewhat agrees with your Komodo 3 evals, at least in its ranking of its top 3 moves.
And I'm also interested in the type of positions that pose problems to engines or, perhaps more to the point, which engines are better in which types of positions (open, semi-open, or closed) and in which phases of the game (opening, middlegame, endgame). I hope to start a project "soon" to determine this using the chessgames.com Sunday puzzles as the test positions. If interested, see my forum for some more details. I'm basically looking for people to help in analyzing the positions using different engines.
|Dec-26-12|| ||DanielBryant: I did poorly on Guess the Move with this game. I kept wanting to "resolve" the situation so the knight wouldn't remain on h8, but my amateur mind was biased against the dynamic possibilities if Kotov just let it be.|
|Apr-12-13|| ||perfidious: Kotov's knight performs feats of legerdemain in its most unusual journey from g1: h3-f4-g6-h8, where it perishes for the greater good.|
While White's remarkable attacking move 16.Nf4 has rightly got much attention, the finishing blow 30.Rxc7+ is also worthy of note.
This was Keres' only defeat entering the eighteenth and last round in the following critical game for his opponent: Bronstein vs Keres, 1950.
|Jul-14-19|| ||Jonathan Sarfati: Not the only time where Kotov played a winning Nh8 against Keres: see Kotov vs Keres, 1950|
< Earlier Kibitzing · PAGE 2 OF 2 ·