Chessqueen wrote: ↑Mon Apr 27, 2020 5:17 amlkaufman wrote: ↑Mon Apr 27, 2020 4:59 amChessqueen wrote: ↑Mon Apr 27, 2020 4:08 amMy main question is since GM Lenderman is about 100+ rating over Jorge Sammour is the difference in speed of your 32 cores will give Komodo 13 another 150 rating points than using my computer?Chessqueen wrote: ↑Mon Apr 27, 2020 3:45 am[Event "Blitz:10'+5""]Chessqueen wrote: ↑Mon Apr 27, 2020 3:36 amlkaufman wrote: ↑Sun Apr 26, 2020 7:38 pm Was this game 15 + 10 or 10 + 5? What was final position Komodo eval shown? Regarding bishop handicap, with GM level players bishop pair is a clear advantage even in blitz, maybe only in bullet would it make sense to remove a bishop, but then the computer would win in either case.
The time control was 10'+5" Jorge analyzed another way that the game could had gone but he was running out of time.
[Site "MyTown"]
[Date "2020.04.26"]
[Round "?"]
[White "komodo 13.3 64-bit -13.3-64b"]
[Black "Jorge Sammour"]
[Result "1/2-1/2"]
[Annotator "ChessQueen"]
[SetUp "1"]
[FEN "rbnqknbr/pppp1ppp/8/8/8/8/PPPPPPPP/RBNQKN1R w KQkq - 0 1"]
[PlyCount "5"]
[TimeControl "600+5"]
1. e4 {[%emt 0:00:52]} c6 {[%emt 0:00:52]} 2. c3 {
[%emt 0:00:29]} d5 {[%emt 0:01:32]} 3. exd5 {[%emt 0:00:28]} Qxd5 {
[%emt 0:02:03]} 4. Ne3 {[%emt 0:00:00]} Qd7 {[%emt 0:00:15]} 5. d4 {
[%emt 0:01:18]} Ne7 {[%emt 0:02:47]} 6. O-O {[%emt 0:00:00]} Ne6 {
[%emt 0:00:31]} 7. Nd3 {[%emt 0:01:53]} f6 {[%emt 0:00:12]} 8. a4 {
[%emt 0:01:47]} Bd6 {[%emt 0:00:21]} 9. Re1 {[%emt 0:01:26]} O-O-O {
[%emt 0:00:40]} 10. a5 {[%emt 0:00:50]} h5 {[%emt 0:00:36]} 11. Ba2 {
[%emt 0:00:40]} Nf4 {[%emt 0:01:43]} 12. Bxg8 {[%emt 0:00:39]} Nxd3 {
[%emt 0:00:43]} 13. Qxd3 {[%emt 0:00:17]} Rdxg8 {[%emt 0:00:58]} 14. a6 {
[%emt 0:00:50]} h4 {[%emt 0:00:53]} 15. axb7+ {[%emt 0:00:27]} Kxb7 {
[%emt 0:00:52]} 16. Qa6+ {[%emt 0:00:08]} Ka8 {[%emt 0:01:42]}
[%emt 0:00:00]} (17. Nc4 {[%emt 0:00:44]} Nc8 {[%emt 0:00:06]} 18. Na5 {
[%emt 0:00:40]} Re8 {[%emt 0:00:07]} 19. Nxc6 {[%emt 0:00:17]} Rxe1+ {
[%emt 0:00:09]} 20. Rxe1 {[%emt 0:00:17]} Qb7 {[%emt 0:00:26]} 21. Qxb7+ {
[%emt 0:00:32]} Kxb7 {[%emt 0:00:10]} 22. Na5+ {[%emt 0:00:30]} Kb6 {
[%emt 0:00:08]} 23. Ra1 {[%emt 0:01:05]} Bc7 {[%emt 0:00:00]} 24. Nc4+ {
[%emt 0:00:22]} Kb5 {[%emt 0:00:30]} 25. Ne3 {[%emt 0:00:00]} a5 {
[%emt 0:00:14]} 26. g3 {[%emt 0:00:14]} Kc6 {[%emt 0:00:21]} 27. Rc1 {
[%emt 0:00:02]} h3 {[%emt 0:00:59]} 28. Rc2 {[%emt 0:00:00]} Ne7 {
[%emt 0:00:48]} 29. Kf1 {[%emt 0:00:25]} a4 {[%emt 0:01:37]} 30. Ke2 {
[%emt 0:00:30]} Re8 {[%emt 0:01:51]} 31. Kd3 {[%emt 0:03:05]} Nd5 {
[%emt 0:01:56]} 32. Rc1 {[%emt 0:01:29]} Nxe3 {[%emt 0:01:04]} 33. fxe3 {
[%emt 0:00:00]} Rb8 {[%emt 0:00:27]} 34. Rc2 {[%emt 0:00:20]} f5 {
[%emt 0:01:38]} 35. Rf2 {[%emt 0:00:00]} Bd8 {[%emt 0:00:54]} 36. Kc2 {
[%emt 0:00:37]} g6 {[%emt 0:00:41]} 37. Kd3 {[%emt 0:00:18]} Kd5 {
[%emt 0:01:44]} 38. Rd2 {[%emt 0:00:21]})) 1/2-1/2
Well, yes Komodo will gain at least another 150 elo, probably more, but it's not that simple, because if both sides play 150 elo stronger then the odds-receiver should score more, since reducing the error rate for both players will obviously favor the one getting the handicap. So if Komodo and Sammour are really equal on your computer under the Lenderman conditions, then we might be in trouble. But we don't know that, as your first game was played with Contempt 16, and the second with bishop for pawn handicap. When you say he "analyzed" another continuation after the game, do you really mean he "played another game against Komodo from move 7". I mean if he analyzed it with Komodo on both sides surely Black would have won, and also the same if he analyzed it for both sides without Komodo.
If you can tell me the nodes per second Komodo gets in the initial position (normal chess, no handicap) on your computer with Threads set to whatever you used during this game, I can give you a much more accurate answer to how many elo stronger it will play on the Threadripper.
I am only getting 1545 kN/s in the initial position, No he analyzed it against Komodo 13 from move 7 on.........
On my 8 core laptop I get 13,000 kn/s in in initial position, on my 32 core Threadripper I get 57,500 KN/s (using 32 Threads; the count goes up to 81,000 kn/s if I use 64 but using more threads than cores artificially inflates the node count so I would use the lower figure). There is some wasting of nodes when running a lot of cores so let's say my threadripper is maybe close to 32 times as fast as your machine, so five doublings. The general rule for rapid games is something like 60 elo per doubling, so that would be 300 elo. Maybe 300 elo for Komodo offsets 150 elo for the human in a handicap game, hard to say. So perhaps Komodo on my threadripper vs Lenderman might do about as well as it does on your machine vs. Sammour. But we will have a newer, stronger Komodo, and the option to use MCTS or even an NN version if we think it will help. Still, based on your limited data and his comments, Komodo is probably an underdog in the overall Lenderman match, especially in the FRC games at full knight odds.
By the way, the analysis you quote is a bit strange, in that it is hard to imagine any strong player, human or engine, not playing 14...b6 but allowing 15.axb7+. Perhaps he wasn't trying to make best moves for Black for some reason. Or were you perhaps playing some or all of the Black moves yourself in this analysis?
I strongly believe that any FIDE 2225 thru 2275 can crash the champion Carlsen with a knight odds