Three Knights vs Eight Pawns

[Milos]

this only comes to show how much restricted our beloved forum is in terms of chess quality.

4 out of 5 people, so 80%, voted this is a draw.

actually, black easily wins this.

It seems so. With good Komodo engine building an opening book and with the same Komodo in self-games on 2 threads, I got:


10''+ 0.1''"
67 Black wins, 33 draws

60''+ 0.6''
80 Black wins, 20 draws


The skewness of the results and the trend seems to confirm that Black wins here in serious games. Pretty interesting imbalance, engines are confused about initial scores in these positions, so one has to play games, if not a strong player himself. Even a strong player might be confused here.

You really believe small MC experiment at such a rudimentary short TC has any meaning???

Yes, and this is not MC, pretty good games on two threads (1 minute level on 2 threads is almost blitz).

Wow, how come no one actually thought of implementing this brilliant tactic for TCEC for example? Instead of thinking on early moves for half an hour, just pick top 3 candidates and play 100 games at 1' TC which overall wouldn't take more than 15mins on 44 cores for 3 moves and you'd get much better performance .

[Laskos]

this only comes to show how much restricted our beloved forum is in terms of chess quality.

4 out of 5 people, so 80%, voted this is a draw.

actually, black easily wins this.

It seems so. With good Komodo engine building an opening book and with the same Komodo in self-games on 2 threads, I got:


10''+ 0.1''"
67 Black wins, 33 draws

60''+ 0.6''
80 Black wins, 20 draws


The skewness of the results and the trend seems to confirm that Black wins here in serious games. Pretty interesting imbalance, engines are confused about initial scores in these positions, so one has to play games, if not a strong player himself. Even a strong player might be confused here.

You really believe small MC experiment at such a rudimentary short TC has any meaning???

Yes, and this is not MC, pretty good games on two threads (1 minute level on 2 threads is almost blitz).

Wow, how come no one actually thought of implementing this brilliant tactic for TCEC for example? Instead of thinking on early moves for half an hour, just pick top 3 candidates and play 100 games at 1' TC which overall wouldn't take more than 15mins on 44 cores for 3 moves and you'd get much better performance .

Yes, as HGM showed in similar, but even weirder disbalances, Stockfish might stay to evaluate the initial position on 44 cores whatever for 1 day, and still completely misevaluate it. But many fast games from his disbalances were very helpful. You seem to lack gut feeling in these disbalances. Openings are pretty irrelevant here, what matters are later stages of the game.

By the way, my 4-mover opening book I built in 10 minutes has 500+ uniques, reasonable positions. Now, at 120''+1.2'' (double time control), the score is 86 Black wins, 14 draws. The trend is pretty clear. It is still possible that real LTC might bring something new, but I would bet on Black win, and I voted so.

[Michel]

Yes, as HGM showed in similar, but even weirder disbalances, Stockfish might stay to evaluate the initial position on 44 cores whatever for 1 day, and still completely misevaluate it. But many fast games from his disbalances were very helpful. You seem to lack gut feeling in these disbalances. Openings are pretty irrelevant here, what matters are later stages of the game.

Are you using TBs? I was watching a game of SF against itself and it did not know how to evaluate KNNKP properly, leading to large misevaluations.

[Laskos]

Yes, as HGM showed in similar, but even weirder disbalances, Stockfish might stay to evaluate the initial position on 44 cores whatever for 1 day, and still completely misevaluate it. But many fast games from his disbalances were very helpful. You seem to lack gut feeling in these disbalances. Openings are pretty irrelevant here, what matters are later stages of the game.

Are you using TBs? I was watching a game of SF against itself and it did not know how to evaluate KNNKP properly, leading to large misevaluations.

I did not use TBs, but this is with Komodo 11.2. I don't know if Komodo has a similar flaw. Might try with TBs.

[Laskos]

Yes, as HGM showed in similar, but even weirder disbalances, Stockfish might stay to evaluate the initial position on 44 cores whatever for 1 day, and still completely misevaluate it. But many fast games from his disbalances were very helpful. You seem to lack gut feeling in these disbalances. Openings are pretty irrelevant here, what matters are later stages of the game.

Are you using TBs? I was watching a game of SF against itself and it did not know how to evaluate KNNKP properly, leading to large misevaluations.

I did not use TBs, but this is with Komodo 11.2. I don't know if Komodo has a similar flaw. Might try with TBs.

I uploaded here
http://s000.tinyupload.com/?file_id=159 ... 5594020110
an EPD file of 1800+ reasonable unique 4-mover openings from that initial position.

[Laskos]

Yes, as HGM showed in similar, but even weirder disbalances, Stockfish might stay to evaluate the initial position on 44 cores whatever for 1 day, and still completely misevaluate it. But many fast games from his disbalances were very helpful. You seem to lack gut feeling in these disbalances. Openings are pretty irrelevant here, what matters are later stages of the game.

Are you using TBs? I was watching a game of SF against itself and it did not know how to evaluate KNNKP properly, leading to large misevaluations.

I did not use TBs, but this is with Komodo 11.2. I don't know if Komodo has a similar flaw. Might try with TBs.

I left overnight Komodo 11.2 with 6-men Syzygy on SSD at 360''+ 3.6'' time control on two threads from the built opening suite, and the result in 100 games was 95 Black wins and 5 draws. The position is probably decided in favor of Black.

I modified the problem to this opening position, which seems truly borderline whether White can hold:

[d]1nn1k1n1/4p3/8/8/8/PPPPPPPP/8/4K3 w - - 0 1

On two threads, from a 3-mover book, the results were in 100 games:

10''+ 0.1''
50 Black wins, 50 sraws

60''+ 0.6''
51 Black wins, 49 draws

No scaling, and very hard too decel the decisiveness of the Black advantage, that already might depend on openings too. Will try to test at 360''+ 3.6'' in at least 40 games, maybe some trend will emerge.

[Nordlandia]

Charge of the light Brigade

5min + ponder enabled

K11.2.2 on 2-core vs SF on 8-core.

[pgn][Event "EXPERT-PC, Blitz 5m+5s"]
[Site "EXPERT-PC"]
[Date "2017.09.01"]
[Round "1"]
[White "Komodo 11.2.2 64-bit"]
[Black "Stockfish 310817 64 BMI2"]
[Result "1-0"]
[Annotator "-3.12;-6.80"]
[SetUp "1"]
[FEN "1q1qk1q1/3ppp2/8/8/8/2N2N2/3PPP2/1NNNKNN1 w - - 0 1"]
[PlyCount "59"]
[TimeControl "300+5"]

1. Na3 {-3.12/22 10 Both last book move} d5 {-6.80/28 27} 2. Nd3 {-2.85/24 13
(d4)} f6 {-7.49/30 41 (Qd7)} 3. Nab5 {-2.69/26 10 (Nc2)} Qdc8 {-6.63/29 6} 4.
Ng3 {-2.63/27 17 (Nde3)} e5 {-7.30/26 8} 5. Nh5 {-2.43/27 4 (e4)} Qg6 {-7.11/
29 16 (Kf7)} 6. Nxd5 {-1.94/26 10 (Nxf6+)} Qxb5 {-6.10/29 19} 7. Nhxf6+ {
-1.87/30 0} Kd8 {-6.03/38 48} 8. N5c3 {-1.82/35 0} Qbb8 {-5.81/40 18 (Qbb7)} 9.
Nfe4 {-1.57/29 10} Qbb7 {-5.81/40 0 (Qcg4)} 10. Ng3 {-1.53/29 15 (e3)} e4 {
-5.81/41 49} 11. Ncxe4 {-1.47/30 0} Qbg7 {-5.73/40 32 (Qgxe4)} 12. e3 {-1.26/
24 11 (Nde5)} Qa8 {-4.74/29 20 (Qa1)} 13. N1e2 {-1.17/27 21 (Nde5)} Qga1 {
-4.07/31 16 (Qgxe4)} 14. N2c3 {-1.04/23 23 (Ned4)} Ke8 {-4.61/27 9 (Qc8)} 15.
Nf4 {-0.84/24 23 (Nde5)} Qh6 {-4.42/28 5 (Qga6)} 16. d4 {-0.58/22 12 (Ne5)}
Q8a6 {-3.25/31 39 (Qb8)} 17. d5 {-0.30/21 21 (Ne5)} Qh8 {-3.16/27 10} 18. Ne6 {
-0.23/23 11 (Kd2)} Q6a5 {-1.11/33 23} 19. Nfd4 {-0.20/25 0} Kf7 {-0.31/26 6}
20. Ke2 {-0.23/21 4 (N6c5)} Q1a3 {-1.23/25 6 (Qc1)} 21. Kf3 {0.71/22 26 (Nc6)}
Kg8 {-2.91/26 3} 22. Ncb5 {1.13/23 25 (Kg2)} Qe7 {-0.28/27 7} 23. Ndc3 {
1.16/21 0} Qeh7 {0.00/31 5 (Qb6)} 24. Nbd6 {1.62/22 30} Qb4 {0.92/33 0 (Qa1)}
25. N6f5 {1.96/21 15} Qd7 {2.50/31 0 (Qe5)} 26. Nf4 {3.10/21 11 (d6)} Qdb7 {
6.23/27 15} 27. Nc6 {4.05/24 4} Qf8 {7.70/26 1 (Qhxc3)} 28. Ng6 {5.25/23 8}
Qbf7 {9.22/30 1} 29. Nxf8 {7.37/25 15} Kxf8 {12.19/32 0} 30. Ng5 {8.18/25 13}
1-0[/pgn]

https://lichess.org/tRQqLkbv

http://www.chessvariants.com/invention/ ... ht-brigade

[Lyudmil Tsvetkov]

Charge of the light Brigade

5min + ponder enabled

K11.2.2 on 2-core vs SF on 8-core.

[pgn][Event "EXPERT-PC, Blitz 5m+5s"]
[Site "EXPERT-PC"]
[Date "2017.09.01"]
[Round "1"]
[White "Komodo 11.2.2 64-bit"]
[Black "Stockfish 310817 64 BMI2"]
[Result "1-0"]
[Annotator "-3.12;-6.80"]
[SetUp "1"]
[FEN "1q1qk1q1/3ppp2/8/8/8/2N2N2/3PPP2/1NNNKNN1 w - - 0 1"]
[PlyCount "59"]
[TimeControl "300+5"]

1. Na3 {-3.12/22 10 Both last book move} d5 {-6.80/28 27} 2. Nd3 {-2.85/24 13
(d4)} f6 {-7.49/30 41 (Qd7)} 3. Nab5 {-2.69/26 10 (Nc2)} Qdc8 {-6.63/29 6} 4.
Ng3 {-2.63/27 17 (Nde3)} e5 {-7.30/26 8} 5. Nh5 {-2.43/27 4 (e4)} Qg6 {-7.11/
29 16 (Kf7)} 6. Nxd5 {-1.94/26 10 (Nxf6+)} Qxb5 {-6.10/29 19} 7. Nhxf6+ {
-1.87/30 0} Kd8 {-6.03/38 48} 8. N5c3 {-1.82/35 0} Qbb8 {-5.81/40 18 (Qbb7)} 9.
Nfe4 {-1.57/29 10} Qbb7 {-5.81/40 0 (Qcg4)} 10. Ng3 {-1.53/29 15 (e3)} e4 {
-5.81/41 49} 11. Ncxe4 {-1.47/30 0} Qbg7 {-5.73/40 32 (Qgxe4)} 12. e3 {-1.26/
24 11 (Nde5)} Qa8 {-4.74/29 20 (Qa1)} 13. N1e2 {-1.17/27 21 (Nde5)} Qga1 {
-4.07/31 16 (Qgxe4)} 14. N2c3 {-1.04/23 23 (Ned4)} Ke8 {-4.61/27 9 (Qc8)} 15.
Nf4 {-0.84/24 23 (Nde5)} Qh6 {-4.42/28 5 (Qga6)} 16. d4 {-0.58/22 12 (Ne5)}
Q8a6 {-3.25/31 39 (Qb8)} 17. d5 {-0.30/21 21 (Ne5)} Qh8 {-3.16/27 10} 18. Ne6 {
-0.23/23 11 (Kd2)} Q6a5 {-1.11/33 23} 19. Nfd4 {-0.20/25 0} Kf7 {-0.31/26 6}
20. Ke2 {-0.23/21 4 (N6c5)} Q1a3 {-1.23/25 6 (Qc1)} 21. Kf3 {0.71/22 26 (Nc6)}
Kg8 {-2.91/26 3} 22. Ncb5 {1.13/23 25 (Kg2)} Qe7 {-0.28/27 7} 23. Ndc3 {
1.16/21 0} Qeh7 {0.00/31 5 (Qb6)} 24. Nbd6 {1.62/22 30} Qb4 {0.92/33 0 (Qa1)}
25. N6f5 {1.96/21 15} Qd7 {2.50/31 0 (Qe5)} 26. Nf4 {3.10/21 11 (d6)} Qdb7 {
6.23/27 15} 27. Nc6 {4.05/24 4} Qf8 {7.70/26 1 (Qhxc3)} 28. Ng6 {5.25/23 8}
Qbf7 {9.22/30 1} 29. Nxf8 {7.37/25 15} Kxf8 {12.19/32 0} 30. Ng5 {8.18/25 13}
1-0[/pgn]

https://lichess.org/tRQqLkbv

http://www.chessvariants.com/invention/ ... ht-brigade

this is unrepresentative: too few pawns, too small pawn span, favouring greatly the knights.

[whereagles]

this is unrepresentative: too few pawns, too small pawn span, favouring greatly the knights.

and the revolt of the peasants is more realistic??

[Nordlandia]

<Elephantiasis effect>
Strong pieces lose value in the presence of opponent weaker pieces. And there is no way to relax, for lack of suitable trading partners. The queen's base value is devalued since there ain't any equal trading partners available.