Komodo knight odd match vs FM Victor Bolzoni : report

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Rebel, chrisw

User avatar
cdani
Posts: 2204
Joined: Sat Jan 18, 2014 10:24 am
Location: Andorra

Re: Komodo knight odd match vs FM Victor Bolzoni : report

Post by cdani »

thekingman wrote:I have long been in favor of a different kind of odds matches - initiative odds.
...
Both starting positions are arguably even more advantageous than knight odds, yet Stockfish won them both. Could a human do better?
Very interesting, the match and your idea!

I suppose a GM will do a lot better in your idea than Crafty, at least in the ideas developed to try to win. Clearly the side with initiative advantage had to open the position or press the rival one, and Crafty just played normal moves. Of course Crafty does not understand the position.
lkaufman
Posts: 5960
Joined: Sun Jan 10, 2010 6:15 am
Location: Maryland USA

Re: Komodo knight odd match vs FM Victor Bolzoni : report

Post by lkaufman »

thekingman wrote:Fascinating match, thank you very much for organizing it and sharing the results with us. I very much look forward to seeing how the second half goes on Sunday.

However, while I understand that material odds matches have a long history, in a lot of ways, they are not like normal chess at all - one side simply needs to successfully trade everything without blundering to win. I have long been in favor of a different kind of odds matches - initiative odds.

In an initiative odds match, one side starts with a lot of extra moves, or the opponent is forced to play a series of highly compromising moves before the game begins for real. This initiative can often be transformed into material advantage, but you need to work for it, and the result is a great duel between attacking and defensive technique - certainly a more "chess-like" battle!

I have suggested this sort of handicap myself. X number of moves (in the first four ranks) before Black moves makes sense. X might be 3 for a 2800 GM, 4 for a 2700, 5 for a 2600, 6 for a 2500 etc. Just my guess as to what might be fair. Exposed king is another way to do it as you say. Perhaps we'll try one of these ideas in the near future.

Here are two interesting such games between a roughly GM-strength program (Crafty) and Stockfish 6:

[pgn]
[Event "Stockfish vs Crafty 9 moves odds"]
[White "Crafty-23.4-win32"]
[Black "Stockfish-6-sse42"]
[Result "0-1"]

1. Nf3 Nf6 2. g3 Ng8 3. Bg2 Nf6 4. O-O Ng8 5. e4 Nf6 6. d3 Ng8 7. Be3 Nf6 8.
Nbd2 Ng8 9. c3 e6 10. Qb3 Nf6 11. e5 Nd5 12. Bg5 Be7 13. Ne4 Bxg5 14. Nexg5 Ne7
15. d4 h6 16. Ne4 O-O 17. Rfe1 Nbc6 18. Qa3 Rb8 19. c4 b5 20. Bf1 bxc4 21. Rac1
Bb7 22. Rxc4 Nd5 23. Nc5 Ba8 24. Bg2 d6 25. Nd3 Qd7 26. Rec1 Nb6 27. R4c2 Nd5
28. b3 a5 29. Rd2 Rfd8 30. Nb2 Nce7 31. Nc4 Bc6 32. Nxa5 Ra8 33. Rb2 Bb5 34. b4
Nc6 35. exd6 cxd6 36. Nd2 Rdc8 37. Bxd5 exd5 38. Rbc2 Nxd4 39. Rxc8+ Rxc8 40.
Rxc8+ Qxc8 41. Nab3 Nc2 42. Qb2 Bd3 43. a3 d4 44. Nc1 Bf5 45. Ncb3 d3 46. h4
Be6 47. Kh2 Bd5 48. Qc1 Qe6 49. Qf1 Bxb3 50. Qxd3 Ba4 51. Nf3 d5 52. Qc3 Qe2
53. Qc8+ Be8 54. Kg1 Ne1 55. Nd4 Qe4 56. Kf1 Nd3 57. Ne2 Kh7 58. Qc2 Qh1+ 59.
Ng1 Bb5 0-1

[/pgn]
Engines started on move 9

[pgn]
[Event "Stockfish vs Crafty Kd5 and move odds"]
[White "Stockfish-6-sse42"]
[Black "Crafty-23.4-win32"]
[Result "1-0"]

1. e3 Nf6 2. Ke2 Ng4 3. Kf3 Nh6 4. Ke4 Ng8 5. Kd5 Nc6 6. Ke4 Nf6+ 7. Kf3 d5 8.
Ke2 e5 9. Ke1 Bd6 10. h3 O-O 11. d3 d4 12. Nf3 Bd7 13. e4 Be6 14. Be2 Qe7 15.
Kf1 h6 16. Kg1 a5 17. Kh2 a4 18. Re1 Bb4 19. Bd2 Bc5 20. Na3 Bxa3 21. bxa3 Qxa3
22. Qc1 Qd6 23. Kg1 b5 24. Nh4 Rfe8 25. a3 Rab8 26. f4 exf4 27. Bxf4 Ne5 28.
Rf1 Qc5 29. Bxh6 Nxe4 30. Bxg7 Kxg7 31. Qf4 Nc4 32. Bg4 Ned6 33. Bxe6 Rxe6 34.
Qg4+ Rg6 35. Rxf7+ Kxf7 36. Qxg6+ Ke7 37. Qg7+ Kd8 38. Qf8+ Kd7 39. Qxb8 Ne3
40. Nf3 Qxc2 41. Qg8 Ndf5 42. Qg6 c5 43. h4 Ne7 44. Qg7 Qb2 45. Re1 Qa2 46. Rc1
Qxa3 47. Rb1 b4 48. Qf7 Qc3 49. Ne5+ Kd6 50. Qf6+ Kd5 51. Nf7 N3f5 52. Nd8 c4
53. Rf1 1-0
[/pgn]
Engines started on move 6

Both starting positions are arguably even more advantageous than knight odds, yet Stockfish won them both. Could a human do better? It would be an exciting match, that's for sure! So the next time anyone runs a human vs computer odds match, I would like to humbly request that you consider similar initiative odds rather than only pure material odds.
Komodo rules!
thekingman
Posts: 35
Joined: Mon Mar 16, 2015 6:17 am

Re: Komodo knight odd match vs FM Victor Bolzoni : report

Post by thekingman »

lkaufman wrote: I have suggested this sort of handicap myself. X number of moves (in the first four ranks) before Black moves makes sense. X might be 3 for a 2800 GM, 4 for a 2700, 5 for a 2600, 6 for a 2500 etc. Just my guess as to what might be fair. Exposed king is another way to do it as you say. Perhaps we'll try one of these ideas in the near future.
I hope so - that would be a phenomenal match to follow! One related consideration that comes to mind, though, is that to be effective, the human player needs to try and exploit the initative handicap on their own - aka, without computer-based preparation specific to the starting position. The early moves are key with an initiative handicap, and if they are computer assisted (through home preparation), the value of the handicap changes drastically.

One solution would be to simply ask them not to analyze at home. Another, potentially more robust solution would be to predetermine the opening moves, and reveal them only at the start of the game. This could also be used to ensure a variety of openings and strategies are tried (one game could start with, say, c4 d4 g3 Bg2, while the following could begin with e4 f4 Nf3 Bc4).

I apologize for diverting this thread, but this seemed as appropriate as anywhere for this discussion. Back on topic, I am very much looking forward to tomorrow's games!
User avatar
Guenther
Posts: 4605
Joined: Wed Oct 01, 2008 6:33 am
Location: Regensburg, Germany
Full name: Guenther Simon

Re: Komodo knight odd match vs FM Victor Bolzoni : report

Post by Guenther »

thekingman wrote: [Event "Stockfish vs Crafty 9 moves odds"]
[White "Crafty-23.4-win32"]
[Black "Stockfish-6-sse42"]
[Result "0-1"]

1. Nf3 Nf6 2. g3 Ng8 3. Bg2 Nf6 4. O-O Ng8 5. e4 Nf6 6. d3 Ng8 7. Be3 Nf6 8.
Nbd2 Ng8

Engines started on move 9

[Event "Stockfish vs Crafty Kd5 and move odds"]
[White "Stockfish-6-sse42"]
[Black "Crafty-23.4-win32"]
[Result "1-0"]

1. e3 Nf6 2. Ke2 Ng4 3. Kf3 Nh6 4. Ke4 Ng8 5. Kd5 Nc6
Engines started on move 6

Both starting positions are arguably even more advantageous than knight odds, yet Stockfish won them both. Could a human do better? It would be an exciting match, that's for sure! So the next time anyone runs a human vs computer odds match, I would like to humbly request that you consider similar initiative odds rather than only pure material odds.
Just for the record I disagree that both start positions are even more advantageous than Knight odds.
I would say the advantage is slightly less than two pawns. What advantage would Komodo 9 or newer give here?
(I wouldn't trust Stockfish here BTW, because its eval always adds too much for mobility and development in the early stage of a game, but that is known.)

Guenther
lkaufman
Posts: 5960
Joined: Sun Jan 10, 2010 6:15 am
Location: Maryland USA

Re: Komodo knight odd match vs FM Victor Bolzoni : report

Post by lkaufman »

thekingman wrote:
lkaufman wrote: I have suggested this sort of handicap myself. X number of moves (in the first four ranks) before Black moves makes sense. X might be 3 for a 2800 GM, 4 for a 2700, 5 for a 2600, 6 for a 2500 etc. Just my guess as to what might be fair. Exposed king is another way to do it as you say. Perhaps we'll try one of these ideas in the near future.
I hope so - that would be a phenomenal match to follow! One related consideration that comes to mind, though, is that to be effective, the human player needs to try and exploit the initative handicap on their own - aka, without computer-based preparation specific to the starting position. The early moves are key with an initiative handicap, and if they are computer assisted (through home preparation), the value of the handicap changes drastically.

One solution would be to simply ask them not to analyze at home. Another, potentially more robust solution would be to predetermine the opening moves, and reveal them only at the start of the game. This could also be used to ensure a variety of openings and strategies are tried (one game could start with, say, c4 d4 g3 Bg2, while the following could begin with e4 f4 Nf3 Bc4).

I apologize for diverting this thread, but this seemed as appropriate as anywhere for this discussion. Back on topic, I am very much looking forward to tomorrow's games!
What games are tomorrow? Our next scheduled match is about two weeks from now.
Komodo rules!
thekingman
Posts: 35
Joined: Mon Mar 16, 2015 6:17 am

Re: Komodo knight odd match vs FM Victor Bolzoni : report

Post by thekingman »

lkaufman wrote: What games are tomorrow? Our next scheduled match is about two weeks from now.
From the first post of this thread:
Vinvin wrote: 2nd day of the match will be next Sunday for the 3 last games. We try to begin 1 hour sooner and I'll look to use the Chessbase interface to avoid all this troubles.
Unless plans changed or I missed something, we should be in for three more knight odds games today!
Uri Blass
Posts: 10281
Joined: Thu Mar 09, 2006 12:37 am
Location: Tel-Aviv Israel

Re: Komodo knight odd match vs FM Victor Bolzoni : report

Post by Uri Blass »

thekingman wrote:
lkaufman wrote: What games are tomorrow? Our next scheduled match is about two weeks from now.
From the first post of this thread:
Vinvin wrote: 2nd day of the match will be next Sunday for the 3 last games. We try to begin 1 hour sooner and I'll look to use the Chessbase interface to avoid all this troubles.
Unless plans changed or I missed something, we should be in for three more knight odds games today!
next sunday was already a week ago

This is the thread of the 4th game and there were threads for the 5th game and the 6th game.



http://www.talkchess.com/forum/viewtopic.php?t=57125
lkaufman
Posts: 5960
Joined: Sun Jan 10, 2010 6:15 am
Location: Maryland USA

Re: Komodo knight odd match vs FM Victor Bolzoni : report

Post by lkaufman »

thekingman wrote:
lkaufman wrote: What games are tomorrow? Our next scheduled match is about two weeks from now.
From the first post of this thread:
Vinvin wrote: 2nd day of the match will be next Sunday for the 3 last games. We try to begin 1 hour sooner and I'll look to use the Chessbase interface to avoid all this troubles.
Unless plans changed or I missed something, we should be in for three more knight odds games today!
That match is finished; we played three games one Sunday and the other three a week later. We will have other matches soon, but only the match with GM Martin Petr in two weeks is definite at this time.
Komodo rules!
duncan
Posts: 12038
Joined: Mon Jul 07, 2008 10:50 pm

Re: Komodo knight odd match vs FM Victor Bolzoni : report

Post by duncan »

lkaufman wrote:
thekingman wrote:
lkaufman wrote: What games are tomorrow? Our next scheduled match is about two weeks from now.
From the first post of this thread:
Vinvin wrote: 2nd day of the match will be next Sunday for the 3 last games. We try to begin 1 hour sooner and I'll look to use the Chessbase interface to avoid all this troubles.
Unless plans changed or I missed something, we should be in for three more knight odds games today!
That match is finished; we played three games one Sunday and the other three a week later. We will have other matches soon, but only the match with GM Martin Petr in two weeks is definite at this time.
do you know what the handicap is ?
lkaufman
Posts: 5960
Joined: Sun Jan 10, 2010 6:15 am
Location: Maryland USA

Re: Komodo knight odd match vs FM Victor Bolzoni : report

Post by lkaufman »

duncan wrote:
lkaufman wrote:
thekingman wrote:
lkaufman wrote: What games are tomorrow? Our next scheduled match is about two weeks from now.
From the first post of this thread:
Vinvin wrote: 2nd day of the match will be next Sunday for the 3 last games. We try to begin 1 hour sooner and I'll look to use the Chessbase interface to avoid all this troubles.
Unless plans changed or I missed something, we should be in for three more knight odds games today!
That match is finished; we played three games one Sunday and the other three a week later. We will have other matches soon, but only the match with GM Martin Petr in two weeks is definite at this time.
do you know what the handicap is ?
I proposed again two pawns, but this time two games each with a2,b2; a2d2; and a2e2. These are probably the best pawns for White to remove other than a2,h2 which I felt was too small a handicap and rather weird. He hasn't had enough time to respond yet. This would be similar to the Kasparov two pawn handicap match, but against a much stronger opponent.
Komodo rules!