Return Match for Komodo

[Lyudmil Tsvetkov]

I can understand that a human, even if strong GM, loses to a top engine.
Humans make mistakes, sometimes many.

What really puzzles me, however, is the total lack of resistance. I really can not explain to myself why the score in all the games drops all the time instead of increasing. Advantage is so big, that it should increase continuously, and by move 30 the game should be over.

Ok, humans make mistakes, but why not increase your advantage up till move 10 or 15, and then make a losing mistake? Why the advantage constantly drops starting from the first plies?

Just to show how big really the handicap advantage is, below a game of mine against Komodo 8:

[pgn][Event "Blitz 1m"]
[Site "Microsoft"]
[Date "2016.08.02"]
[Round "?"]
[White "Komodo 8 64-bit"]
[Black "Lyudmil Tsvetkov, Dell xps"]
[Result "0-1"]
[Annotator "owner"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/P1PPP1PP/RNBQKBNR w KQkq - 0 1"]
[PlyCount "59"]
[EventDate "2016.??.??"]
[TimeControl "60"]

{512MB, OWNER-PC} 1. d4 {-1.36/15 2} g6 {10} 2. e4 {-1.14/14 1} d6 {2} 3. Nf3 {
-1.26/16 2} Bg7 {2} 4. c3 {-1.34/15 1} Nf6 {8} 5. Nbd2 {-1.35/18 1} O-O {3} 6.
Bd3 {-1.35/17 1} Nc6 {3} 7. O-O {-1.10/14 1} e5 {3} 8. d5 {-1.24/14 2} Ne7 {4}
9. Rb1 {-1.36/14 2} h6 {3} 10. Nc4 {-1.25/15 5} Ne8 {7} 11. Ne3 {-1.15/12 2} f5
{7} 12. Nd2 {-1.26/17 1} f4 {4} 13. Ng4 {-1.35/18 0} h5 {7} 14. Nf2 {-1.09/15 1
} Nf6 {5} 15. Nf3 {-1.07/16 2} Bh6 {6} 16. Qb3 {-1.18/16 0} Kh8 {10} 17. h4 {
-1.11/15 3} Nh7 {37} 18. Bb2 {-0.86/14 2} g5 {27} 19. hxg5 {-0.90/13 1} Nxg5 {1
} 20. Nh4 {-1.04/16 0} Rg8 {15} 21. Qd1 {-1.13/16 2} f3 {50} 22. Nxf3 {-1.79/
16 2} Ng6 {14} 23. Ne1 {-2.00/16 2} Nf4 {48} 24. Bc1 {-4.24/23 0} Ngh3+ {47}
25. Nxh3 {-5.11/27 0} Nxh3+ {2} 26. Kh1 {-5.48/21 2} Qh4 {97} 27. Bxh6 {-6.05/
25 0} Nf2+ {7} 28. Kg1 {-6.02/23 0} Nxd1 {8} 29. Rxd1 {-6.09/24 0} Bg4 {53} 30.
Rb1 {-3.82/17 1} 0-1

[/pgn]

[Lyudmil Tsvetkov]

Above game was with default contempt=-7, but from my experience Komodo with contempt=-70 for example is the same engine that only makes weaker moves.

[Guenther]

I can understand that a human, even if strong GM, loses to a top engine.
Humans make mistakes, sometimes many.

What really puzzles me, however, is the total lack of resistance. I really can not explain to myself why the score in all the games drops all the time instead of increasing. Advantage is so big, that it should increase continuously, and by move 30 the game should be over.

Ok, humans make mistakes, but why not increase your advantage up till move 10 or 15, and then make a losing mistake? Why the advantage constantly drops starting from the first plies?

Just to show how big really the handicap advantage is, below a game of mine against Komodo 8:

[pgn][Event "Blitz 1m"]
[Site "Microsoft"]
[Date "2016.08.02"]
[Round "?"]
[White "Komodo 8 64-bit"]
[Black "Lyudmil Tsvetkov, Dell xps"]
[Result "0-1"]
[Annotator "owner"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/P1PPP1PP/RNBQKBNR w KQkq - 0 1"]
[PlyCount "59"]
[EventDate "2016.??.??"]
[TimeControl "60"]

{512MB, OWNER-PC} 1. d4 {-1.36/15 2} g6 {10} 2. e4 {-1.14/14 1} d6 {2} 3. Nf3 {
-1.26/16 2} Bg7 {2} 4. c3 {-1.34/15 1} Nf6 {8} 5. Nbd2 {-1.35/18 1} O-O {3} 6.
Bd3 {-1.35/17 1} Nc6 {3} 7. O-O {-1.10/14 1} e5 {3} 8. d5 {-1.24/14 2} Ne7 {4}
9. Rb1 {-1.36/14 2} h6 {3} 10. Nc4 {-1.25/15 5} Ne8 {7} 11. Ne3 {-1.15/12 2} f5
{7} 12. Nd2 {-1.26/17 1} f4 {4} 13. Ng4 {-1.35/18 0} h5 {7} 14. Nf2 {-1.09/15 1
} Nf6 {5} 15. Nf3 {-1.07/16 2} Bh6 {6} 16. Qb3 {-1.18/16 0} Kh8 {10} 17. h4 {
-1.11/15 3} Nh7 {37} 18. Bb2 {-0.86/14 2} g5 {27} 19. hxg5 {-0.90/13 1} Nxg5 {1
} 20. Nh4 {-1.04/16 0} Rg8 {15} 21. Qd1 {-1.13/16 2} f3 {50} 22. Nxf3 {-1.79/
16 2} Ng6 {14} 23. Ne1 {-2.00/16 2} Nf4 {48} 24. Bc1 {-4.24/23 0} Ngh3+ {47}
25. Nxh3 {-5.11/27 0} Nxh3+ {2} 26. Kh1 {-5.48/21 2} Qh4 {97} 27. Bxh6 {-6.05/
25 0} Nf2+ {7} 28. Kg1 {-6.02/23 0} Nxd1 {8} 29. Rxd1 {-6.09/24 0} Bg4 {53} 30.
Rb1 {-3.82/17 1} 0-1

[/pgn]

Well, as usual you forgot to mention a few details, e.g White additionally had to suffer a 10:1 - 16:1 time handicap which can be calculated from the search times.
Like in your handicap test games when you decided when the engine has to move to fit your expectation...
I really don't want to know what else is manipulated or wrong in your test environment.

[Laskos]

I can understand that a human, even if strong GM, loses to a top engine.
Humans make mistakes, sometimes many.

What really puzzles me, however, is the total lack of resistance. I really can not explain to myself why the score in all the games drops all the time instead of increasing. Advantage is so big, that it should increase continuously, and by move 30 the game should be over.

Ok, humans make mistakes, but why not increase your advantage up till move 10 or 15, and then make a losing mistake? Why the advantage constantly drops starting from the first plies?

Just to show how big really the handicap advantage is, below a game of mine against Komodo 8:

[pgn][Event "Blitz 1m"]
[Site "Microsoft"]
[Date "2016.08.02"]
[Round "?"]
[White "Komodo 8 64-bit"]
[Black "Lyudmil Tsvetkov, Dell xps"]
[Result "0-1"]
[Annotator "owner"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/P1PPP1PP/RNBQKBNR w KQkq - 0 1"]
[PlyCount "59"]
[EventDate "2016.??.??"]
[TimeControl "60"]

{512MB, OWNER-PC} 1. d4 {-1.36/15 2} g6 {10} 2. e4 {-1.14/14 1} d6 {2} 3. Nf3 {
-1.26/16 2} Bg7 {2} 4. c3 {-1.34/15 1} Nf6 {8} 5. Nbd2 {-1.35/18 1} O-O {3} 6.
Bd3 {-1.35/17 1} Nc6 {3} 7. O-O {-1.10/14 1} e5 {3} 8. d5 {-1.24/14 2} Ne7 {4}
9. Rb1 {-1.36/14 2} h6 {3} 10. Nc4 {-1.25/15 5} Ne8 {7} 11. Ne3 {-1.15/12 2} f5
{7} 12. Nd2 {-1.26/17 1} f4 {4} 13. Ng4 {-1.35/18 0} h5 {7} 14. Nf2 {-1.09/15 1
} Nf6 {5} 15. Nf3 {-1.07/16 2} Bh6 {6} 16. Qb3 {-1.18/16 0} Kh8 {10} 17. h4 {
-1.11/15 3} Nh7 {37} 18. Bb2 {-0.86/14 2} g5 {27} 19. hxg5 {-0.90/13 1} Nxg5 {1
} 20. Nh4 {-1.04/16 0} Rg8 {15} 21. Qd1 {-1.13/16 2} f3 {50} 22. Nxf3 {-1.79/
16 2} Ng6 {14} 23. Ne1 {-2.00/16 2} Nf4 {48} 24. Bc1 {-4.24/23 0} Ngh3+ {47}
25. Nxh3 {-5.11/27 0} Nxh3+ {2} 26. Kh1 {-5.48/21 2} Qh4 {97} 27. Bxh6 {-6.05/
25 0} Nf2+ {7} 28. Kg1 {-6.02/23 0} Nxd1 {8} 29. Rxd1 {-6.09/24 0} Bg4 {53} 30.
Rb1 {-3.82/17 1} 0-1

[/pgn]

Well, as usual you forgot to mention a few details, e.g White additionally had to suffer a 10:1 - 16:1 time handicap which can be calculated from the search times.
Like in your handicap test games when you decided when the engine has to move to fit your expectation...
I really don't want to know what else is manipulated or wrong in your test environment.

I guess he saw the engine output throughout the game, which is huge. Not a single above 0.50 blunder. Petr Neuman made such in each single game with longer thinking time.

[Jesse Gersenson]

Contempt, the type which helps Komodo play odds matches better, was introduced publically in Komodo 9.2. It is not in Komodo 8 which Lyudmil Tsvetkov is using.

Average search depth reached in Neuman match was 31, and just 17 in Tsvetkov game.

The starting position is nearly correct -- b2+f2 is very close. Try again with the actual handicap used, b2+e2.

[velmarin]

Contempt, the type which helps Komodo play odds matches better, was introduced publically in Komodo 9.2. It is not in Komodo 8 which Lyudmil Tsvetkov is using.

Average search depth reached in Neuman match was 31, and just 17 in Tsvetkov game.

The starting position is nearly correct -- b2+f2 is very close. Try again with the actual handicap used, b2+e2.

It would be like, if someone wants to win at home, simply wins, you can stop the clock, use other engine, "Undo Move".
Already rains wet envelope, good players, have your elo rating, more modesty in general and do not tend to be boastful.

[Guenther]

Well, as usual you forgot to mention a few details, e.g White additionally had to suffer a 10:1 - 16:1 time handicap which can be calculated from the search times.
Like in your handicap test games when you decided when the engine has to move to fit your expectation...
I really don't want to know what else is manipulated or wrong in your test environment.

I guess he saw the engine output throughout the game, which is huge. Not a single above 0.50 blunder. Petr Neuman made such in each single game with longer thinking time.

This could be true too of course, but my point was that he obviously set the program to play all moves in 1min, when he was allowed to overstep the time up to infinity.
This amounted to a time handicap of >10:1 additionally for the program.

Code: Select all

01.  d4   -1.36/15  2  01... g6   10
02.  e4   -1.14/14  1  02... d6    2
03.  Nf3  -1.26/16  2  03... Bg7   2
04.  c3   -1.34/15  1  04... Nf6   8
05.  Nbd2 -1.35/18  1  05... O-O   3
06.  Bd3  -1.35/17  1  06... Nc6   3
07.  O-O  -1.10/14  1  07... e5    3
08.  d5   -1.24/14  2  08... Ne7   4
09.  Rb1  -1.36/14  2  09... h6    3
10.  Nc4  -1.25/15  5  10... Ne8   7
11.  Ne3  -1.15/12  2  11... f5    7
12.  Nd2  -1.26/17  1  12... f4    4
13.  Ng4  -1.35/18  0  13... h5    7
14.  Nf2  -1.09/15  1  14... Nf6   5
15.  Nf3  -1.07/16  2  15... Bh6   6
16.  Qb3  -1.18/16  0  16... Kh8  10
17.  h4   -1.11/15  3  17... Nh7  37
18.  Bb2  -0.86/14  2  18... g5   27
19.  hxg5 -0.90/13  1  19... Nxg5  1
20.  Nh4  -1.04/16  0  20... Rg8  15
21.  Qd1  -1.13/16  2  21... f3   50
22.  Nxf3 -1.79/16  2  22... Ng6  14
23.  Ne1  -2.00/16  2  23... Nf4  48
24.  Bc1  -4.24/23  0  24... Ngh3 47
25.  Nxh3 -5.11/27  0  25... Nxh3  2
26.  Kh1  -5.48/21  2  26... Qh4  97
27.  Bxh6 -6.05/25  0  27... Nf2   7
28.  Kg1  -6.02/23  0  28... Nxd1  8
29.  Rxd1 -6.09/24  0  29... Bg4  53
30.  Rb1  -3.82/17  1      
                   39            490

Edit:
The depths are strange too, as if he had played ponder off first and after White was already lost he changed to ponder on ;-)

I think it is a waste of time to read his posts...

[Lyudmil Tsvetkov]

I can understand that a human, even if strong GM, loses to a top engine.
Humans make mistakes, sometimes many.

What really puzzles me, however, is the total lack of resistance. I really can not explain to myself why the score in all the games drops all the time instead of increasing. Advantage is so big, that it should increase continuously, and by move 30 the game should be over.

Ok, humans make mistakes, but why not increase your advantage up till move 10 or 15, and then make a losing mistake? Why the advantage constantly drops starting from the first plies?

Just to show how big really the handicap advantage is, below a game of mine against Komodo 8:

[pgn][Event "Blitz 1m"]
[Site "Microsoft"]
[Date "2016.08.02"]
[Round "?"]
[White "Komodo 8 64-bit"]
[Black "Lyudmil Tsvetkov, Dell xps"]
[Result "0-1"]
[Annotator "owner"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/P1PPP1PP/RNBQKBNR w KQkq - 0 1"]
[PlyCount "59"]
[EventDate "2016.??.??"]
[TimeControl "60"]

{512MB, OWNER-PC} 1. d4 {-1.36/15 2} g6 {10} 2. e4 {-1.14/14 1} d6 {2} 3. Nf3 {
-1.26/16 2} Bg7 {2} 4. c3 {-1.34/15 1} Nf6 {8} 5. Nbd2 {-1.35/18 1} O-O {3} 6.
Bd3 {-1.35/17 1} Nc6 {3} 7. O-O {-1.10/14 1} e5 {3} 8. d5 {-1.24/14 2} Ne7 {4}
9. Rb1 {-1.36/14 2} h6 {3} 10. Nc4 {-1.25/15 5} Ne8 {7} 11. Ne3 {-1.15/12 2} f5
{7} 12. Nd2 {-1.26/17 1} f4 {4} 13. Ng4 {-1.35/18 0} h5 {7} 14. Nf2 {-1.09/15 1
} Nf6 {5} 15. Nf3 {-1.07/16 2} Bh6 {6} 16. Qb3 {-1.18/16 0} Kh8 {10} 17. h4 {
-1.11/15 3} Nh7 {37} 18. Bb2 {-0.86/14 2} g5 {27} 19. hxg5 {-0.90/13 1} Nxg5 {1
} 20. Nh4 {-1.04/16 0} Rg8 {15} 21. Qd1 {-1.13/16 2} f3 {50} 22. Nxf3 {-1.79/
16 2} Ng6 {14} 23. Ne1 {-2.00/16 2} Nf4 {48} 24. Bc1 {-4.24/23 0} Ngh3+ {47}
25. Nxh3 {-5.11/27 0} Nxh3+ {2} 26. Kh1 {-5.48/21 2} Qh4 {97} 27. Bxh6 {-6.05/
25 0} Nf2+ {7} 28. Kg1 {-6.02/23 0} Nxd1 {8} 29. Rxd1 {-6.09/24 0} Bg4 {53} 30.
Rb1 {-3.82/17 1} 0-1

[/pgn]

Well, as usual you forgot to mention a few details, e.g White additionally had to suffer a 10:1 - 16:1 time handicap which can be calculated from the search times.
Like in your handicap test games when you decided when the engine has to move to fit your expectation...
I really don't want to know what else is manipulated or wrong in your test environment.

I guess he saw the engine output throughout the game, which is huge. Not a single above 0.50 blunder. Petr Neuman made such in each single game with longer thinking time.

Rigth, Kai,

I have been looking at the engine output, this gives considerable edge to the human.
The reason though is not that I wanted to do so, but simply that it stayed like that after I had been play-analysing some positions.

I really do not think though that this is a decisive factor with such a huge handicap.

[Lyudmil Tsvetkov]

I can understand that a human, even if strong GM, loses to a top engine.
Humans make mistakes, sometimes many.

What really puzzles me, however, is the total lack of resistance. I really can not explain to myself why the score in all the games drops all the time instead of increasing. Advantage is so big, that it should increase continuously, and by move 30 the game should be over.

Ok, humans make mistakes, but why not increase your advantage up till move 10 or 15, and then make a losing mistake? Why the advantage constantly drops starting from the first plies?

Just to show how big really the handicap advantage is, below a game of mine against Komodo 8:

[pgn][Event "Blitz 1m"]
[Site "Microsoft"]
[Date "2016.08.02"]
[Round "?"]
[White "Komodo 8 64-bit"]
[Black "Lyudmil Tsvetkov, Dell xps"]
[Result "0-1"]
[Annotator "owner"]
[SetUp "1"]
[FEN "rnbqkbnr/pppppppp/8/8/8/8/P1PPP1PP/RNBQKBNR w KQkq - 0 1"]
[PlyCount "59"]
[EventDate "2016.??.??"]
[TimeControl "60"]

{512MB, OWNER-PC} 1. d4 {-1.36/15 2} g6 {10} 2. e4 {-1.14/14 1} d6 {2} 3. Nf3 {
-1.26/16 2} Bg7 {2} 4. c3 {-1.34/15 1} Nf6 {8} 5. Nbd2 {-1.35/18 1} O-O {3} 6.
Bd3 {-1.35/17 1} Nc6 {3} 7. O-O {-1.10/14 1} e5 {3} 8. d5 {-1.24/14 2} Ne7 {4}
9. Rb1 {-1.36/14 2} h6 {3} 10. Nc4 {-1.25/15 5} Ne8 {7} 11. Ne3 {-1.15/12 2} f5
{7} 12. Nd2 {-1.26/17 1} f4 {4} 13. Ng4 {-1.35/18 0} h5 {7} 14. Nf2 {-1.09/15 1
} Nf6 {5} 15. Nf3 {-1.07/16 2} Bh6 {6} 16. Qb3 {-1.18/16 0} Kh8 {10} 17. h4 {
-1.11/15 3} Nh7 {37} 18. Bb2 {-0.86/14 2} g5 {27} 19. hxg5 {-0.90/13 1} Nxg5 {1
} 20. Nh4 {-1.04/16 0} Rg8 {15} 21. Qd1 {-1.13/16 2} f3 {50} 22. Nxf3 {-1.79/
16 2} Ng6 {14} 23. Ne1 {-2.00/16 2} Nf4 {48} 24. Bc1 {-4.24/23 0} Ngh3+ {47}
25. Nxh3 {-5.11/27 0} Nxh3+ {2} 26. Kh1 {-5.48/21 2} Qh4 {97} 27. Bxh6 {-6.05/
25 0} Nf2+ {7} 28. Kg1 {-6.02/23 0} Nxd1 {8} 29. Rxd1 {-6.09/24 0} Bg4 {53} 30.
Rb1 {-3.82/17 1} 0-1

[/pgn]

Well, as usual you forgot to mention a few details, e.g White additionally had to suffer a 10:1 - 16:1 time handicap which can be calculated from the search times.
Like in your handicap test games when you decided when the engine has to move to fit your expectation...
I really don't want to know what else is manipulated or wrong in your test environment.

I have always said that I am giving myself more time when playing top engines - usually ratio is 1:2 or 1:3, when I am playing blitz games, but this time I simply forgot to change Komodo thinking time from 1' to my usual 2' + 2''. I had been running some gauntlets before, and as you know very well, Fritz settings for tournament mode are valid for the entire interface until changed.

I can assure you, no other tricks though.

Of course, I can win and have been winning many games at blitz time control with absolutely the same conditions for both sides, i.e. not looking at the eval, and not overstepping TC, but as this requires too much tension and gathering-up, consuming too much energy, I usually simply prefer to play it leisurely at time handicap.

[Frank Brenner]

In Game 4 Stockfish also plays the knight sac 11 Sxe6 with contempt = 0.

I dont believe that "contempt" was the feature that is responsible for this beautiful open tactical chess game (nr 4).

The reason was: luck

To have more exciting games like game 4 in the future in Computerchess Komodo has to learn the idea of raising the potential complexity of a position.

I want to explain what i mean with that:

Most dead draw positions are of very low (zero) potential of complexity.

The Starting position (no move played so far) has a high potential of complexity

A complicated position in the middle game that is game theoretic a draw, but white to move has to find the next 3 very sharp and only moves could be of moderate low potential complexity or of high potential complexity depending on the final position that appears on the board when white succesful finds the 3 right moves and plays the right path.

If white findes the 3 moves and the final position is a very much simplificated position the initial position is maybe of more or les low potential complexity. The potential complexity could be 3 or so on a interval between [0,20]

If the final position after finding the 3 moves is still complex, the initial position also have a very high potential complexity maybe 10 or so.


At the moment Komodo only deals with material. K trys to avoid exchanges ... but this feature is only a very low approximation of the goal to raise the potential complexity of a position.



The problem is: how to identify the potential complexity and how how to give it a number of 0....50 .. ?

Actually i have no real idea for that.

When i look at the live tcec games i see a lot of 0.00 zero evaluations.
It is hard find properties of those positions that identify this positions as positions of low potential complexity.
I think there is one other thing to deal with: making progress.

I agree with you, I've often thought about how to measure drawishness (opposite of complexity) in a position. It's not easy. The problem of progress is also critical, but there at least we took a step forward with Komodo 10.1. Regarding game 4, maybe the knight sacrifice wasn't so shocking, but the earlier Ng5 move setting it up was pretty hard to find. Did Stockfish play that? Komodo only did it with high Contempt.

asmfish shows 10. Sg5 only in the first second and after this it changes it mind and never goes back to 10 Sg5.

The reason is because stockfish is the opinion that after 10. Sg5 10 ... Sd7? from Neuman was a bad move (that enables 10 Sxe6!)

In n-best mode Stockfish recommend (after 10. Sg5) 0-0 with +1,41 instead of Sb8-d7 +0,87


Even with contempt = 100 stockfish does not play 10.Sg5



In tactics the top engine (K or SF) is about 800 ELo stronger than the bast humans, but in positional play it is maybe between -1200 elo and +300 elo, average maybe +150 elo stronger.
In some cases the human is very sure about a theoretic draw while the engine displays +0,7 for a long time of moves and just makes move repetitions (in fact due to contempt or due to +0,7 they the engine dont make repetitions instead it makes permutation of non-pawns to avoid the 0,0 evaluation of a real repetition)

For games engine- vs human i would suggest to write an separate engine, that takes into account that (GM) humans are weak in tactics but very good in making simplifications to keep a drawish position.

(of course that does not apply to correspondence chess)