An idea for a different type of tournament

[Uri Blass]

Let chess engines start from the following position and play 2 games with white and black.
The winner is the side that mate faster.
If both side need the same number of moves to mate then it is a draw.


[d]rnb1kbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1

I do not know if stockfish is going not to be best but I suspect it is not going to be best because of the poor evaluation it has
Of course I do not expect strong engines to see a forced mate but I expect them at least to see a significant improvement in the evaluation and I see nothing of it with stockfish.
13.28/1 13.26/2 13.40/3 and only 13.35/40

I wonder how much time stockfish needs to understand that at least it wins a pawn relative to depth 1
so it can show an evaluation of 14.28

Note that 1.Nc3 seems to me not the best move and stockfish needs a long time to change it's mind to 1.e4


FEN: rnb1kbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1

Stockfish_20020515_x64_modern:
Found 145 tablebases
1/1 00:00 42 42k +13.28 1.Nc3
2/2 00:00 123 62k +13.26 1.Nc3 a6
3/3 00:00 267 134k +13.40 1.Nc3 d6 2.Nd5
4/4 00:00 387 194k +13.64 1.Nc3 c6 2.Ne4
5/5 00:00 1k 590k +13.84 1.Nf3 c6 2.Nc3 d5
6/6 00:00 4k 1,271k +13.53 1.Nf3 Nf6 2.Nc3 Na6 3.Ne5 d6
7/9 00:00 8k 1,553k +13.28 1.Nf3 Nc6 2.Nc3 Nf6 3.d3 g6 4.Be3 d6
8/11 00:00 14k 1,519k +13.37 1.e3 Nc6 2.Nf3 Nf6 3.Be2 d6 4.Nc3 Bf5
9/12 00:00 28k 2,182k +13.30 1.Nf3 Nc6 2.Nc3 Nf6 3.e4 d6 4.Bd3 a6 5.0-0 Bg4
10/15 00:00 69k 2,463k +13.40 1.e3 Nc6 2.Nc3 d6 3.Nb5 Kd8 4.Nf3 a6 5.Nc3 Bg4 6.h3 Be6
11/20 00:00 96k 2,587k +13.32 1.Nc3 Nc6 2.e3 d6 3.Nb5 Kd8 4.Nf3 Nf6 5.Bc4 a6 6.Nc3 Be6 7.Bxe6 fxe6 8.0-0 Kc8 9.Ne2
12/22 00:00 170k 2,618k +13.28 1.Nc3 Nc6 2.Nb5 Kd8 3.Nh3 a6 4.Ng5 Nh6 5.Nc3 e6 6.d4 d5 7.Be3 Bd7
13/20 00:00 327k 2,974k +13.23 1.Nc3 Nf6 2.e3 e6 3.Nh3 Be7 4.Nb5 Bd8 5.Be2 Nc6 6.0-0 0-0 7.Nf4 Ne5 8.d3
14/22 00:00 470k 3,015k +13.27 1.Nc3 Nf6 2.e3 e6 3.Nh3 Be7 4.d4 0-0 5.Be2 Nc6 6.Nf4 g5 7.Nd3 d6 8.Bd2 Bd7 9.Kf1
15/22 00:00 745k 3,280k +13.27 1.Nc3 Nf6 2.e3 e6 3.Nh3 Be7 4.d4 0-0 5.Be2 d6 6.Nf4 Re8 7.0-0 Nbd7 8.Bd2 c5 9.Bd3 cxd4 10.exd4
16/22 00:00 1,042k 3,427k +13.24 1.Nc3 Nf6 2.Nf3 e6 3.e4 d6 4.Bc4 Be7 5.0-0 0-0 6.d4 Nc6 7.Qe2 Re8 8.Bf4 Bd7 9.a3 h6 10.Qe3 a6 11.Rfd1
17/27 00:00 1,308k 3,525k +13.27 1.Nc3 Nf6 2.e4 d6 3.Nf3 e6 4.d4 Be7 5.Bf4 0-0 6.Bc4 Nc6 7.Qe2 Bd7 8.0-0-0 Rad8 9.Kb1 a6 10.h3 Nb4 11.a3
18/23 00:00 1,587k 3,566k +13.28 1.Nc3 Nf6 2.e4 d6 3.Nf3 e6 4.d4 Be7 5.Bf4 0-0 6.Bc4 Nc6 7.Qe2 a6 8.a4 Na5 9.Bd3 Nc6 10.0-0-0 Bd7 11.Rd2 Ng4
19/28 00:00 2,475k 3,694k +13.27 1.Nc3 Nf6 2.e4 d6 3.Nf3 e6 4.d4 Be7 5.Bf4 0-0 6.Bc4 Nc6 7.0-0 a6 8.d5 Na5 9.Bd3 Nd7 10.Be3 Nc5 11.Bxc5 dxc5 12.Qe2 h6
20/28 00:01 3,787k 3,694k +13.30 1.Nc3 Nf6 2.e4 d6 3.Nf3 e6 4.d4 Be7 5.Bf4 0-0 6.Be2 a6 7.0-0 b5 8.Qd3 Nbd7 9.e5 dxe5 10.Nxe5 Nxe5 11.Bxe5 c5 12.Ne4 Nxe4 13.Qxe4
21/27 00:01 4,750k 3,231k +13.33 1.Nc3 Nf6 2.e4 d6 3.Nf3 e6 4.d4 Be7 5.Bf4 0-0 6.Be2 a6 7.0-0 b5 8.Qd3 Nbd7 9.e5 dxe5 10.Nxe5 Nxe5 11.Bxe5 c6 12.Bf3 Nd5 13.Nxd5 cxd5 14.c3
22/29 00:01 5,729k 3,179k +13.30 1.Nc3 Nf6 2.e4 d6 3.Nf3 e6 4.d4 Be7 5.Bc4 0-0 6.Qe2 a6 7.a4 Nc6 8.0-0 Bd7 9.Bf4 Nb4 10.Bb3 Nc6 11.g3 Rad8 12.Kg2 h6 13.Rad1
23/31 00:02 7,826k 3,364k +13.30 1.Nc3 Nf6 2.e4 d6 3.Nf3 e6 4.d4 Be7 5.Bf4 0-0 6.Bd3 Nc6 7.a3 Re8 8.0-0 Bd8 9.Bd2 Bd7 10.g3 g6 11.Qe2 Kg7 12.Kg2 Rc8 13.Rad1 Re7
24/34 00:03 12,365k 3,581k +13.31 1.Nc3 d6 2.Nf3 Nf6 3.e4 e6 4.d4 Be7 5.h3 a6 6.Be2 Nc6 7.a3 0-0 8.0-0 h6 9.Bd3 g5 10.Be3 Kg7 11.Qe2 Bd7 12.Rad1 Rfc8 13.Rd2 Rd8
25/37 00:04 17,173k 3,686k +13.36 1.Nc3 d6 2.Nf3 Nf6 3.e4 e6 4.d4 Be7 5.Bd3 0-0 6.0-0 Nc6 7.a3 Bd7 8.h3 h5 9.Bf4 Bd8 10.Qe2 g6 11.Rad1 a6 12.Rd2 Kg7 13.g3 Be7 14.Kg2
26/36 00:09 36,101k 3,828k +13.30 1.Nc3 Nf6 2.e4 d6 3.Nf3 e6 4.d4 a6 5.Bf4 Nc6 6.h3 g6 7.Bc4 b5 8.Bd3 b4 9.Ne2 a5 10.0-0 Ba6 11.a3 Bxd3 12.Qxd3 0-0-0 13.Qa6+ Kd7 14.axb4 axb4 15.Qd3 Bg7
27/36 00:10 39,680k 3,822k +13.31 1.Nc3 Nf6 2.e4 d6 3.Nf3 e6 4.d4 a6 5.Bd3 Nc6 6.0-0 g6 7.a3 h5 8.Be3 Bg7 9.h3 Bd7 10.Qe2 Kf8 11.Rad1 Bh6 12.e5 Bxe3 13.Qxe3 dxe5 14.dxe5 Nd5 15.Nxd5
28/39 00:16 64,903k 3,828k +13.29 1.Nc3 Nf6 2.e4 d6 3.Nf3 a6 4.d4 e6 5.Bf4 Nc6 6.Bd3 g6 7.0-0 Bg7 8.Be3 h5 9.h3 Bd7 10.a3 0-0-0 11.g3 Rhe8 12.Qe2 Re7 13.Kg2 Kb8 14.Rad1 a5 15.Rd2 b6 16.Rfd1
29/38 00:19 74,754k 3,838k +13.29 1.Nc3 Nf6 2.e4 d6 3.Nf3 a6 4.d4 e6 5.Bf4 Nc6 6.h3 b5 7.a3 Bb7 8.Qd2 0-0-0 9.Bd3 h6 10.Rd1 g5 11.Bh2 Nd7 12.Bg3 Nb6 13.d5 exd5 14.Nxd5 Nxd5 15.exd5 Ne7
30/40 00:24 94,959k 3,861k +13.29 1.Nc3 Nf6 2.e4 d6 3.Nf3 a6 4.d4 e6 5.Bf4 Nc6 6.a3 g6 7.Bc4 Bg7 8.0-0 0-0 9.h3 h5 10.Be3 Rd8 11.Qe2 e5 12.d5 Nb8 13.b4 Nbd7 14.Ng5 Nb6 15.Bxb6 cxb6 16.Bd3 Nd7 17.Rad1
31/40 00:31 123,399k 3,865k +13.27 1.Nc3 Nf6 2.e4 d6 3.Nf3 a6 4.d4 e6 5.Bf4 Nc6 6.a3 g6 7.Bc4 Bg7 8.0-0 0-0 9.h3 Nd7 10.Be3 Nb6 11.Be2 f5 12.exf5 Rxf5 13.Bd3 Rh5 14.Qe2 Bd7 15.Rad1 Nd5 16.Nxd5 Rxd5 17.c3
32/48 03:47 838,979k 3,694k +13.32 1.e4 a6 2.Nf3 d6 3.d4 e6 4.Nc3 Nf6 5.Bf4 Be7 6.e5 dxe5 7.Nxe5 0-0 8.Qf3 c6 9.Nc4 Nbd7 10.0-0-0 g6 11.Bd3 Nd5 12.Nxd5 cxd5 13.Nd6 Nf6 14.Nxc8 Rfxc8 15.g4 h5 16.h3 Kg7 17.c3 Rc6 18.Bg5 hxg4 19.hxg4
33/46 03:52 856,836k 3,684k +13.32 1.e4 a6 2.Nf3 d6 3.d4 e6 4.Nc3 Nc6 5.Bd3 Bd7 6.d5 exd5 7.Nxd5 0-0-0 8.0-0 Nge7 9.Be3 f6 10.Nd4 Nxd5 11.exd5 Ne5 12.Be2 Re8 13.Bd2 Nf7 14.c4 Ng5 15.Bf4 Be7 16.f3 g6 17.Qd2 h5 18.Be3 h4
34/47 04:07 904,744k 3,654k +13.30 1.e4 a6 2.Nf3 d6 3.d4 e6 4.Nc3 Nc6 5.Bd3 Bd7 6.d5 exd5 7.Nxd5 0-0-0 8.0-0 Nge7 9.Be3 f6 10.Bc4 Ne5 11.Be2 Re8 12.Nxe5 fxe5 13.c4 Nxd5 14.cxd5 g6 15.Bd3 Bg7 16.g3 h5 17.Qf3 Bh6 18.Bxh6 Rxh6 19.Qe3 Rh7
35/50 04:39 1,014,891k 3,629k +13.32 1.e4 a6 2.Nf3 d6 3.d4 e6 4.Nc3 Nf6 5.Bf4 Be7 6.e5 dxe5 7.Nxe5 0-0 8.Qf3 c6 9.Nc4 Nbd7 10.0-0-0 g6 11.Bd3 Nd5 12.Nxd5 cxd5 13.Nd6 Nf6 14.Nxc8 Rfxc8 15.g4 h5 16.h3 Re8 17.g5 Nd7 18.h4 b5 19.c3 b4 20.cxb4 Bxb4 21.Qg3
36/51 05:17 1,153,412k 3,628k +13.35 1.e4 a6 2.Nf3 d6 3.d4 e6 4.Nc3 Nf6 5.Bf4 Nbd7 6.a4 Be7 7.Bd3 0-0 8.0-0 Nh5 9.Bc1 Nhf6 10.a5 e5 11.Be2 exd4 12.Nxd4 g6 13.Be3 Re8 14.Bd2 Nc5 15.f3 h5 16.Be3 Bd7 17.Bc4 Ne6 18.Qe2 Nxd4 19.Bxd4 Be6 20.Ra4 Bxc4
37/49 05:57 1,305,225k 3,649k +13.32 1.e4 a6 2.Nf3 d6 3.d4 e6 4.Nc3 Nf6 5.Bf4 Nbd7 6.a4 b6 7.Bd3 Bb7 8.Qe2 a5 9.Nb5 Rc8 10.0-0 Be7 11.Rad1 0-0 12.h3 g6 13.b3 Kg7 14.Nc3 c6 15.Qd2 Rh8 16.Rde1 h6 17.h4 Ng4 18.g3 Rhe8 19.Kg2 Rc7 20.e5 dxe5 21.dxe5
38/54 09:09 1,999,499k 3,638k +13.32 1.e4 d6 2.Nf3 e6 3.d4 Nf6 4.Nc3 Be7 5.Bd3 0-0 6.0-0 c6 7.Qe2 Nbd7 8.Bf4 Re8 9.h3 Bf8 10.e5 dxe5 11.Nxe5 Nxe5 12.Bxe5 Nd7 13.Bg3 Nf6 14.Ne4 Nxe4 15.Qxe4 g6 16.c3 Kg7 17.Bf4 Bd7 18.f3 b5 19.Rf2 Be7 20.Rd1 c5 21.Bc7
39/54 09:24 2,056,824k 3,641k +13.31 1.e4 d6 2.Nf3 e6 3.d4 Nf6 4.Nc3 Be7 5.Bd3 0-0 6.0-0 c6 7.Qe2 Nbd7 8.Bf4 Re8 9.h3 Bf8 10.e5 dxe5 11.Nxe5 Nxe5 12.Bxe5 Nd7 13.Bg3 Nf6 14.Ne4 Nxe4 15.Qxe4 g6 16.c3 Kg7 17.Bf4 h5 18.f3 Be7 19.g4 hxg4 20.fxg4 Kg8 21.g5 c5 22.Kg2 cxd4
40/55 13:05 2,871,119k 3,656k +13.35 1.e4 d6 2.Nf3 e6 3.d4 Nd7 4.c4 a6 5.Nc3 b6 6.Bf4 Bb7 7.Bd3 h6 8.a3 Ne7 9.h4 e5 10.Be3 exd4 11.Nxd4 0-0-0 12.f4 Re8 13.0-0 h5 14.b4 g6 15.Rc1 Nf6 16.Nde2 Bg7 17.Bd2 Kb8 18.Qc2 a5 19.g3 axb4 20.axb4 Nc6 21.Nd5 Nxd5 22.cxd5

[lkaufman]

I don't know which engine would be best, but it is safe to predict that Lc0 and most if not all of the NN engines would be the worst. I think it would be more interesting to do this with just knight odds instead of queen odds, because once the eval gets up to about two pieces plus or the equivalent it would be normal to resign or adjudicate the game a win. Playing well when the game isn't quite at the resign threshold is important, but how you play when the game is trivially won doesn't much matter. Also it would tell us which engine is likely strongest against top human players. since they are approaching knight odds level vs. top engines.

[phhnguyen]

Let chess engines start from the following position and play 2 games with white and black.
The winner is the side that mate faster.
If both side need the same number of moves to mate then it is a draw.

IMO, it is not a new type of tournaments but a different way to calculate a tournament result (such as giving scores 1/0.5/0 or 3/1/0 for w/d/l) after the tournament completed. You still need to select a typical tournament type such as round-robin/knockout/swiss/gauntlet... before starting. All most all model chess GUIs can easily start tournament games from a given FEN, swap sides with multi rounds thus it is not a big deal either.

[mmt]

I was thinking about this position - it might seem surprising to many people that we still can't prove a mate up a queen and a tempo.

[jp]

I was thinking about this position - it might seem surprising to many people that we still can't prove a mate up a queen and a tempo.

Less ambitiously, maybe we should start with odds of all Black's pieces (only K and Ps).
What is the quickest mate, and what is the mate tree proving it?

With no human input, SF10 at depth 27 gets this PV.

[pgn][FEN "4k3/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQ - 0 1"]

1. e3 h6 2. Qh5 c6 3. Qe5 d6 4. Qxg7 f6 5. Qg8+ Kd7 6. Qb8 b5 7. Qb7+ Ke6 8. Qxc6 a6 9. Qxa6 f5 10. Qxb5 Kf6 11. Bc4 f4 12. exf4 e5 13. fxe5+ dxe5 14. Nf3 e4 15. Qe5+ Kg6 16. Qxe4+ Kg7 *[/pgn]

If you make SF10 look at the position after 4. Qxg7, it claims mate by move 15:

4... e6 5. Qg8+ Ke7 6. Qb8 e5 7. Qxb7+ Ke6 8. Qxc6 Ke7 9. Qc7+ Ke6 10. Bb5 e4 11. Be8 d5 12. f4 exf3 13. Nxf3 Kf5 14. Qxf7+ Kg4 15. Qg6#

The same from the position after 3. Qe5:

4... h5 5. Nf3 a6 6. Qg8+ Kd7 7. Qxf7 h4 8. Nxh4 c5 9. Nf5 Kc6 10. Qxe7 b5 11. Qxd6+ Kb7 12. Qxc5 Ka8 13. Nd6 b4 14. Qc8+ Ka7 15. Qb7#

And after 2. Qh5:

2... a6 3. Qe5 d6 4. Qxg7 e5 5. Qg8+ Ke7 6. Qc8 c6 7. Qxb7+ Ke8 8. Bc4 d5 9. Qxc6+ Kf8 10. Qxd5 Ke7 11. Qxf7+ Kd6 12. Bd5 e4 13. Nc3 a5 14. Nxe4+ Ke5 15. Qe6#


So is it really mate in 15, and can we construct the tree?

And if you just leave SF (e.g. 11 or dev) running on its own, at what depth does it start announcing a mate?

[jp]

odds of all Black's pieces (only K and Ps)

Komodo 13.2 at depth 25 gets this PV.

[pgn]
[FEN "4k3/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQ - 0 1"]

1. e4 h6 2. Qh5 a6 3. Qe5 d6 4. Qxg7 a5 5. Bb5+ c6 6. Be2 f5 7. exf5 Kd8 8. f6 exf6 9. Qxb7 d5
10. Qxc6 Ke7 11. Qxd5 Kf8 12. Qxa5 Kg7 13. d3 Kh7 14. Qc7+ *[/pgn]

Slower mates arise from this, maybe...

[Alayan]

I ran Stockfish dev on the position with only king and pawns for black, several days ago when this was discussed in TCEC chat.

Stockfish displayed a mate score after I let it search for some time.

EDIT :

info depth 33 seldepth 42 multipv 1 score mate 39 nodes 944199624 nps 4600690 hashfull 989 tbhits 0 time 205230 pv e2e3 g7g6 d1f3 c7c6 f3f4 d7d6 f4h6 g6g5 h6h7 a7a6 h7g8 e8d7 g8f7 g5g4 g1e2 d7c8 f7e8 c8c7 e8e7 c7b8 e7d6 b8a8 d6d8 a8a7 d8c7 c6c5 c7c5 a7b8 e2d4 b7b5 c5b6 b8c8 b6a6 c8b8 a6b5 b8a8 b5a5 a8b8 a5e5 b8a7 e5c7 a7a8

info depth 35 seldepth 36 multipv 1 score mate 18 nodes 1073437129 nps 4615980 hashfull 997 tbhits 0 time 232548 pv e2e3 g7g6 d1f3 c7c6 f3f4 d7d6 f4h6 e8d7 h6h7 d7e8 h7g8 e8d7 g8f7 a7a5 g1f3 a5a4 f7g6 a4a3 b1a3 c6c5 g6f7 d7d8 f7f8 d8d7 f1b5 d7c7 f8e7 c7b8 e7d6 b8a7 f3e5 c5c4 e5d7 b7b6 d6b8

info depth 38 seldepth 36 multipv 1 score mate 17 nodes 1243478381 nps 4635693 hashfull 999 tbhits 0 time 268240 pv e2e3 g7g6 d1f3 c7c6 f3f4 d7d6 f4h6 e8d7 h6h7 f7f6 h7g6 a7a6 g1f3 d6d5 g6f7 c6c5 f7d5 d7c8 d5e6 c8b8 e6e7 b8a7 e7c5 a7b8 f3d4 f6f5 d4f5 b7b6 c5e7 b8a8 f1a6 b6b5 e7b7

I remembered it had 2. Qg4 in its mate PV the first time around, so I made it search from there for a time and it came with

info depth 34 seldepth 33 multipv 1 score mate -15 nodes 318192400 nps 3535117 hashfull 681 tbhits 0 time 90009 pv a7a5 g4d4 d7d6 d4h8 e8d7 h8h7 d7e8 g1f3 a5a4 h7g8 e8d7 g8f7 a4a3 b1a3 d7d8 f7e6 g6g5 f3g5 c7c6 g5f7 d8c7 e6e7 c7b6 f7d6 b6a5 a3c4 a5a4 b2b3 a4b4 c1a3

Which is also a mate in 17 from the initial position.

Restarting again from the inital position without clearing hash :

info depth 42 seldepth 34 multipv 1 score mate 16 nodes 409167870 nps 4772193 hashfull 852 tbhits 0 time 85740 pv e2e3 g7g6 d1f3 c7c6 f3f4 d7d6 f4h6 f7f5 h6h7 e8d8 h7g6 f5f4 g1f3 f4e3 d2e3 a7a6 f3g5 b7b5 g6e6 d6d5 g5f7 d8c7 e6e7 c7b6 f7d6 b6a5 e7a7 a5a4 a7a6 a4b4 a6a3

Bringing out the knights early can probably allow a slightly faster mate.

[jp]

If we believe all the SF(10 & dev) mate announcements, then maybe ...h6 & ...a6, are not the best moves for Black after 1.e3 & 2.Qh5.

This is not (yet) an example of SF getting a mate announcement wrong, even if it ends up mate no quicker than 16, because when SF10 went for 1.e3 h6, 2.Qh5 a6, etc. on its own, it didn't announce the mate in 15, only from the position with ...a6 on the board.

[mmt]

I ran Stockfish dev on the position with only king and pawns for black, several days ago when this was discussed in TCEC chat.

Stockfish displayed a mate score after I let it search for some time.

SF Matefinder doesn't get it faster. But we found a mate here so we can now try with one more piece for black.

[jp]

SF Matefinder doesn't get it faster.

Did it show the same mate or a different one?

But we found a mate here so

Do you think SF mate announcements are 100% reliable?