TalkChess.com

Is it possible to quantify the advantage of two knights over two bishops of same color.

The power of the two bishops is often because one can reach the squares the other cannot. With a mostly empty board, one of the two bishops can often control any square on the chessboard in one move. But if the two bishops are the same color, they lose this property. The two same-colored bishops can't do much that one bishop couldn't do, and really can't do much that a bishop plus another minor piece couldn't do.

Two knights is less redundant than two bishops of same color. But how much does the this advantage constitute - half a pawn or more?

Let use this position. How big is white's edge here?

[d]2b1k1b1/pppppppp/8/8/8/8/PPPPPPPP/2N1K1N1 w - - 0 1

I know about this engine game

[pgn][Event "Charactors"] [Site "MacBook Pro Early 2015"] [Date "2018.01.11"] [Round "1"] [White "Rodent III Henny"] [Black "Rodent III Scorpion"] [Result "1/2-1/2"] [WhiteElo ""] [BlackElo ""] [ECO ""] 1.e4 c5 2.Nf3 g6 3.d4 cxd4 4.Nxd4 Nc6 5.c4 Bg7 6.Nc2 Qb6 7.Nc3 Bxc3 8.bxc3 Nf6 9.f3 Qa5 10.Be2 h5 11.O-O h4 12.Be3 Qxc3 13.Rb1 h3 14.Rb5 hxg2 15.Bd2 Qxc2 16.Qxc2 Nd4 17.Qd3 gxf1=B 18.Qxd4 Bxe2 19.Re5 d6 20.Rxe7 Kxe7 21.Bg5 Kd7 22.Bxf6 Re8 23.c5 Re6 24.Qa4 Kc7 25.cxd6 Rxd6 26.Be5 Be6 27.Qd4 Rad8 28.Qc5 Kb8 29.Bxd6 Ka8 30.Be5 Rc8 31.Qe3 Bb5 32.a3 b6 33.Kf2 Kb7 34.Qd2 Rc5 35.Bd6 Rc4 36.h4 Ka6 37.Bb8 Rc8 38.Be5 Rc5 39.Bd4 Rc4 40.Ke3 Ra4 41.Qb2 Rc4 42.Bf6 Ra4 43.Kf4 Bd3 44.Bc3 Rc4 45.Qd2 Bb1 46.a4 f5 47.Ke5 Rc6 48.a5 fxe4 49.fxe4 Bb3 50.axb6 Re6 51.Kf4 Rxe4 52.Kg5 axb6 53.Qd8 B1c2 54.Kh6 Re6 55.Bd4 Kb5 56.Kg5 Be4 57.Qc7 Rc6 58.Qe5 Bbd5 59.Be3 Rc4 60.Kh6 Rc6 61.Qd4 Ka6 62.Qa4 Kb7 63.Qa1 Re6 64.Qb2 Bc4 65.Qg7 Ka6 66.Qc7 Bed5 67.Bf2 Rc6 68.Qe7 Re6 69.Qd8 Be4 70.Qc8 Kb5 71.Qd7 Bc6 72.Qd4 B6d5 73.Qa1 Bb3 74.Qc3 Rc6 75.Qd3 Bdc4 76.Qd7 Bc2 77.Kg7 B2b3 78.Qa7 Bc2 79.Qa1 Bd5 80.Qa3 Bce4 81.Bd4 Rc4 82.Be3 Rc2 83.Qa1 Rc4 84.Qb2 Rb4 85.Qe2 Kc6 86.Qa6 Rb3 87.Bg1 Rb4 88.Qa7 Rb1 89.Bh2 Kc5 90.Qa4 b5 91.Qa6 Kb4 92.Bd6 Kc4 93.Qa3 b4 94.Qa2 Kd4 95.Qf2 Kc3 96.Be5 Kd3 97.Qd4 Kc2 98.Bd6 b3 99.Qf2 Kd3 100.Bb4 Rd1 101.Ba5 Kc4 102.Qe3 Rd3 103.Qc1 Kd4 104.Bb6 Ke5 105.Qg5 Kd6 106.Qd8 Ke6 107.Qf6 Kd7 108.Qe5 Bc6 109.Ba5 Bcd5 110.Qc7 Ke6 111.Qf7 Kd6 112.Qf6 Kd7 113.Kh6 Rf3 114.Qd8 Kc6 115.Qe8 Kc5 116.Qe7 Kb5 117.Qa3 Kc4 118.Qb4 Kd3 119.Qc3 Ke2 120.Qe1 Kd3 121.Qd2 Kc4 122.Qc1 Kd3 123.Qd1 Kc4 124.Qd2 Bc6 125.Kg7 Bcd5 126.Qc1 Kb5 127.Qa3 g5 128.h5 Kc6 129.Qa4 Kd6 130.Kh6 g4 131.Qb4 Kd7 132.Qb6 Bc6 133.Qd4 Bcd5 134.Qg7 Kc6 135.Qc7 Kb5 136.Qb6 Kc4 137.Qb4 Kd3 138.Qd2 Kc4 139.Kg5 g3 140.h6 Bc6 141.Qb4 Kd5 142.Bc7 g2 143.Bh2 Rg3 144.Kf6 b2 145.Qxb2 Kc4 146.Bg1 Rh3 147.Ke7 Bcd5 148.Qd4 Kb3 149.Qd1 Ka2 150.Bd4 Ka3 151.Qc1 Ka4 152.Qb2 Rxh6 153.Bc5 Rh7 154.Kf6 Rb7 155.Qa3 Kb5 156.Ke5 Kc6 157.Bd4 Rd7 158.Qc5 Kb7 159.Bg1 Bc6 160.Kf6 Bed5 161.Qb6 Kc8 162.Bh2 Rf7 163.Kg5 Kd7 164.Qc7 Ke8 165.Qb8 Ke7 166.Qe5 Kf8 167.Qh8 Ke7 168.Bg1 Rf1 169.Bc5 Kd7 170.Qh7 Bf7 171.Qd3 Kc7 172.Qd6 Kb7 173.Qe7 Kc8 174.Kh6 Bc4 175.Qa7 Bb7 176.Bd4 Bcd5 177.Kg7 Kd7 178.Qc5 Ke6 179.Be3 Bbc6 180.Bf2 Rb1 181.Bg3 Rb7 182.Kh8 Kd7 183.Bh2 Ke6 184.Qd6 Kf5 185.Qf4 Kg6 186.Qg4 Kf6 187.Qh4 Kg6 188.Qg3 Kf7 189.Bg1 Be4 190.Bc5 Ke6 191.Qh3 Kd5 192.Qe3 Rh7 193.Kg8 Ke6 194.Kf8 Rh8 195.Kg7 Rh5 196.Bd4 Bcd5 197.Kf8 Rf5 198.Kg7 Rf1 199.Qh3 Rf5 200.Kh6 Kd7 201.Kg7 Ke6 202.Kh6 Kf7 1/2-1/2 [/pgn]

With all material present two same-colored Bishops (the 'anti-pair') are equally strong as two Kights or Knight + Bishop, and about half a Pawn weaker than two Bishops on different shades. To measure it for this end-game position you woul have to do a few hundred games of engine self-play.

hgm wrote: ↑Fri Sep 06, 2019 10:08 pm With all material present two same-colored Bishops (the 'anti-pair') are equally strong as two Kights or Knight + Bishop, and about half a Pawn weaker than two Bishops on different shades. To measure it for this end-game position you woul have to do a few hundred games of engine self-play.

What time control do you recommend. And what contemt value should i use, default?

It shouldn't depend on TC which side is better, so you can use a fast TC to quickly play many games. I would expect contempt to be detrimental against an equally strong opponent (as in self-play). Furthermore, if both engines would play with the same non-zero contempt, it would not really be self-play, as they would differ in opinion about the (white POV) score for repetitions. To be consistent they would need opposite contempt. And under those conditions the value of the contempt would not matter that much, as there would always be one side that thinks a draw is a poor result.

So in case using stockfish developmental version of Mon Aug 26. One side use +24 while the other use -24. TC of 15s+ 0.85s may be suffient i suppose.

What time control do you usually use when generating selfplay games in large numbers?

I just read this article about bishop vs knight.

http://blog.abrok.eu/bishop-vs-knight/

hgm wrote: ↑Sat Sep 07, 2019 10:08 am It shouldn't depend on TC which side is better, so you can use a fast TC to quickly play many games. I would expect contempt to be detrimental against an equally strong opponent (as in self-play). Furthermore, if both engines would play with the same non-zero contempt, it would not really be self-play, as they would differ in opinion about the (white POV) score for repetitions. To be consistent they would need opposite contempt. And under those conditions the value of the contempt would not matter that much, as there would always be one side that thinks a draw is a poor result.

How do you know that it should not depend on TC which side is better?

I think that the only way to know is by playing games at different time control and see that you get similiar results.
Without games between engines you cannot know.

Raphexon wrote: ↑Sat Sep 07, 2019 12:10 pm I just read this article about bishop vs knight.

http://blog.abrok.eu/bishop-vs-knight/

Thanks for the link. He didn't should examine Queen + Knight tandem against Queen + Bishop.