nice tactical position

[jdart]

From a recent online Arasan game:

[d]r1bq1r2/bp3pk1/2pp4/p6p/P1BPPP1N/6Bn/1P4pK/2RQR3 w - - 0 22

White played Nxg2 and won. But on longer analysis, it appears f5! and allowing Black to promote is even better. Stockfish and Houdini are very fast on this. But other engines have difficulty (Fritz 16 for example).

--Jon

[Dann Corbit]

Current dev SF took 37 plies for me:

Code: Select all

Searching: r1bq1r2/bp3pk1/2pp4/p6p/P1BPPP1N/6Bn/1P4pK/2RQR3 w - - 0 1
infinite: 0 time: 0 increment: 0 moves to go: 0

 1   +0.85   00:00    4311  Qxh5 
 2   +2.97   00:00   12661  Qxh5 g1=Q+ Rxg1 
 3   +0.75   00:00   20875  Qxh5 Bxd4 Kxg2 Bxb2 
 4   +0.85   00:00   31043  Qxh5 Bxd4 Kxg2 
 5   +0.95   00:00   39901  Qxh5 Bxd4 f5 Ng5 Kxg2 d5 exd5 cxd5 
 6   +1.59   00:00   50109  Qxh5 Bxd4 Nxg2 Bxb2 
 7   +4.35   00:00   58458  Qxh5 Bxd4 
 8   +2.70   00:00   69431  Qxh5 Bxd4 Kxg2 Qf6 f5 Rh8 Qg4+ Qg5 Kxh3 Qxg4+ Kxg4 
 9   +1.10   00:00   89529  Qxh5 Bxd4 Kxg2 Qf6 f5 Qg5 Qxg5+ Nxg5 Bf4 Nh7 Bxd6 
10   +1.25   00:00  106177  Qxh5 Bxd4 f5 Ng5 Kxg2 d5 exd5 cxd5 Bb3 
11   +0.64   00:00  218602  Qxh5 Bxd4 Kxg2 Qf6 f5 Qg5 Qxg5+ Nxg5 Bxd6 Re8 e5 Bd7 Kg3 
12   +0.73   00:00  267888  Qxh5 Bxd4 Kxg2 Qf6 f5 Qg5 Qxg5+ Nxg5 Rcd1 c5 Nf3 Nxf3 Kxf3 Rd8 Rd2 
13   +0.68   00:00  488505  Qxh5 Bxd4 Kxg2 d5 Bxd5 Qf6 f5 Qg5 Qxg5+ Nxg5 Bc4 Rh8 e5 Bxb2 Rcd1 
14   +0.85   00:00  827218  Qxh5 Bxd4 Kxg2 Qf6 f5 Qg5 Qxg5+ Nxg5 Rcd1 Bxb2 Bxd6 Re8 e5 Bd7 e6 Nxe6 fxe6 
15   +0.77   00:00   1043K  Qxh5 Bxd4 Kxg2 Qf6 f5 Qg5 Qxg5+ Nxg5 Rcd1 Be5 Bxe5+ dxe5 Nf3 b5 axb5 cxb5 Nxg5 bxc4 Rh1 Ba6 
16   +1.08   00:00   1344K  Qxh5 Bxd4 Kxg2 d5 Bxd5 Qf6 f5 Qg5 Qxg5+ Nxg5 Bc4 b5 axb5 Bb7 Rcd1 cxb5 Rxd4 bxc4 f6+ Kh7 Rxc4 
17   +0.80   00:00   4108K  Nxg2 Qf6 f5 Ng5 Bf4 d5 Rc3 Qxd4 Rg3 Qxd1 Rxg5+ Kf6 Rxd1 dxc4 Rxh5 Bb8 Bxb8 Rxb8 Ne3 Rg8 Nxc4 Rg4 
18   +0.97   00:00   5811K  Nxg2 d5 Bh4 Qd6 exd5 Bg4 Be2 f5 dxc6 bxc6 Bf3 Qxd4 Rxc6 Qxd1 Bxd1 Bg1+ Rxg1 Nxg1 Rc7+ Kg8 Bxg4 
19   +1.24   00:01  10366K  Nxg2 Bg4 Be2 Rh8 Bh4 Qd7 f5 f6 e5 dxe5 dxe5 Bxf5 Qxd7+ Bxd7 Bxf6+ Kh6 Bxh8 Rxh8 Rcd1 Bg4 Rd6+ Kg5 Bd3 
20   +1.49   00:01  14484K  Nxg2 Bg4 Be2 Rh8 Bxg4 hxg4 Qxg4+ Kf8 Nh4 Nf2 Bxf2 Bxd4 Bg3 Bxb2 Rb1 Bf6 Kh3 Rg8 Qf3 Rh8 Qf2 Qd7+ f5 Qe7 Rb3 Kg8 Rd3 Re8 Rdd1 
21   +0.95   00:04  36632K  Nxg2 Bg4 Be2 Qd7 f5 Rh8 Bh4 Rag8 Bxg4 hxg4 Qxg4+ Kf8 Qxh3 Bxd4 Rcd1 Be5+ Kg1 f6 Kf2 Qh7 Rg1 b5 axb5 Qb7 bxc6 Qxc6 
22   +1.58   00:05  43181K  Nxg2 Bg4 Be2 Qd7 Bh4 Rh8 Rc3 Rag8 Qd2 Bb6 Rg3 Kf8 Bc4 Bd8 Qe3 Be7 Rf1 Bd8 d5 Be7 dxc6 bxc6 Rxh3 Bxh3 Qxh3 Qxh3+ Kxh3 
23   +1.42   00:06  54712K  Nxg2 d5 Bh4 Nf2 Bxf2 dxc4 f5 Rg8 Nf4 Kf8 Rg1 Rxg1 Qxg1 Qf6 Rxc4 Bb8 Qg3 Qg7 Qh4 Qg4 Qxg4 hxg4 Kg3 Bxf4+ Kxf4 Bd7 Rc3 Re8 e5 b5 Kxg4 bxa4 
24   +1.45   00:08  74234K  Nxg2 d5 Bh4 Nf2 Bxf2 dxc4 f5 Rg8 Rg1 Kf8 Nf4 Rxg1 Qxg1 Qf6 Rxc4 Bd7 Nxh5 Qh6 Qg4 Re8 Rc3 Bxd4 Bxd4 Qd2+ Kh3 Qxd4 Qg5 Rxe4 Nf6 Bxf5+ Qxf5 Re3+ Rxe3 Qxe3+ Kg4 Qd4+ Kg5 Qd2+ Qf4 Qxf4+ Kxf4 b5 b3 bxa4 
25   +1.39   00:09  84384K  Nxg2 d5 Bh4 Nf2 Bxf2 dxc4 f5 Rg8 Rg1 Kf8 Nf4 Rxg1 Qxg1 Qf6 Rxc4 Bb8 Qg3 Qg7 Rc1 Bd7 Qh4 Qg4 Qxg4 hxg4 Kg3 Ke7 Kxg4 Bxf4 Bh4+ Bg5 Kxg5 Rh8 Kg4+ Kf8 Bg3 Ke7 Kf3 f6 Kf4 b5 axb5 cxb5 Rc7 Kd8 
26   +1.83   00:14 126192K  Nxg2 d5 Bh4 Nf2 Bxf2 dxc4 Ne3 Rh8 f5 Kf8 Nxc4 Bb8+ Ne5 f6 Bh4 Qxd4 Qxd4 Bxe5+ Qxe5 fxe5 Bf6 Rg8 Rcd1 Kf7 Bxe5 Ra6 Rg1 Rxg1 Rxg1 Bd7 Rg7+ Ke8 Kg3 c5 Kf4 Rb6 Bc3 Bxa4 Rh7 Bc2 Rxh5 
27   +1.57   00:21 184835K  Nxg2 Bg4 Be2 Qd7 Bh4 f6 f5 Rh8 e5 dxe5 dxe5 Bxf5 Bxf6+ Kh6 Nh4 Rhg8 Rf1 Bg1+ Rxg1 Rxg1 Qxd7 Bxd7 Rxg1 Nxg1 Kxg1 Rg8+ Kf2 Be6 Bf3 Rf8 Be4 Rxf6+ exf6 Kg5 Ng6 Kxf6 Nf4 Bg4 Ke3 
28   +1.82   00:32 282613K  Nxg2 Bg4 Be2 Qd7 Bh4 f5 Rc3 Rae8 Rxh3 Bxh3 Kxh3 Rxe4 Bf2 Rg8 Rg1 Kf8 Bf3 Qf7 Bxe4 fxe4 Qe1 Qb3+ Qe3 Qxe3+ Bxe3 Ke7 f5 d5 Kh4 Kf6 Kxh5 
29   +2.08   00:42 364649K  Nxg2 Bg4 Be2 Qd7 Bh4 f6 Ne3 Rh8 Rf1 Rag8 Rc3 d5 e5 Rh6 Bxf6+ Kf8 Bh4 Rh7 Rd3 Bb6 Qe1 Bd8 Bd1 Bxh4 Qxh4 Ke8 Qe1 Qd8 Rb3 Rf7 f5 Ng5 Be2 Bxe2 Qxe2 
30   +2.00   01:01 534573K  Nxg2 Bg4 Be2 Qd7 Bh4 f5 Rc3 Rae8 Rxh3 Bxh3 Kxh3 Rxe4 Bf2 Rg8 Rg1 Kf6 Bf3 d5 Kh2 Qh7 Bxe4 fxe4 Ne3 Rxg1 Kxg1 Ke6 Kh2 Kd7 Bg3 h4 Bf2 Kc7 f5 Bb6 f6 
31   +1.89   01:12 629344K  Nxg2 Bg4 Be2 Qd7 Bh4 f5 Rc3 Rae8 Rxh3 Bxh3 Kxh3 Rxe4 Bf2 Rg8 Nh4 Kf8 Bf3 Rxe1 Qxe1 Bb6 d5 Bxf2 Qxf2 Qg7 Nxf5 Qg6 Qc2 Rg7 Qe4 Rf7 Nd4 Qxe4 Bxe4 Ke7 dxc6 bxc6 Nxc6+ Kf6 Nd4 d5 Bf3 Rd7 Kh4 Rb7 Bxd5 Rxb2 Kxh5 
32   +1.92   01:32 808349K  Nxg2 Bg4 Be2 Qd7 Bh4 f5 Rc3 Rae8 Rxh3 Bxh3 Kxh3 Rxe4 Bf2 Rg8 Rg1 Kf6 Bf3 d5 Kh2 Re6 Ne3 Rxg1 Qxg1 Qg7 Bh4+ Kf7 Bxh5+ Kf8 Qxg7+ Kxg7 Nxf5+ Kf8 Bf3 Bb6 Kh3 Bc7 Bg3 Kf7 Kh4 Kf8 Kg4 Bd8 Bh4 Bxh4 Nxh4 Kf7 
33   +2.19   01:55   1011M  Nxg2 Bg4 Be2 Qd7 Bh4 Rh8 Rc3 Rae8 Qd2 Rh6 Bd1 f5 exf5 Rxe1 Bxe1 Qxf5 Rb3 Qf7 Rxh3 Bxh3 Kxh3 Qd5 Qd3 Rf6 Bc3 Qf5+ Qxf5 Rxf5 d5+ Kh6 dxc6 bxc6 Bf3 d5 Bxa5 Bd4 Nh4 Rf6 Bd2 Kg7 a5 c5 Kg3 Bxb2 
34   +2.48   02:38   1379M  Nxg2 Bg4 Be2 Qd7 Bh4 f5 Rc3 Rae8 Rxh3 Bxh3 Kxh3 Rxe4 d5 cxd5 Qxd5 Qc6 Qd3 d5 Bf3 Bb6 Bxe4 fxe4 Qd1 Qe6+ Kh2 d4 Qe2 d3 Qxd3 exd3 Rxe6 d2 Rd6 Bc7 Rd7+ Rf7 Rxd2 Bxf4+ Nxf4 Rxf4 Rd7+ Kg6 Be1 Rxa4 Rxb7 Ra2 Rb6+ Kf7 Bc3 a4 Rf6+ Ke8 Re6+ Kd7 Ra6 a3 Ra7+ Kc6 Rxa3 Rxa3 bxa3 
35   +2.37   04:14   2186M  Nxg2 Bg4 Be2 Qd7 Bh4 Rh8 Rc3 Bb6 Qd2 Bd8 f5 d5 Rg3 Nf2 Qf4 Bxh4 Nxh4 Qe7 f6+ Qxf6 Nf5+ Kf8 Qxf2 dxe4 Bxg4 hxg4+ Kg2 Rh5 Rf1 Rd8 Rxg4 Rd5 Ng3 Qxf2+ Rxf2 Rhg5 Rxe4 b5 Ref4 Rd7 Kh3 Rg8 Ne4 
36   +2.35   05:14   2715M  Nxg2 Bg4 Be2 Qd7 Bh4 f5 Rc3 fxe4 Bxg4 hxg4 Rg3 Nf2 Qd2 Rf7 Qxf2 d5 Rf1 Rg8 f5 Bb8 Kg1 Kf8 Rb3 Rh7 Bg3 Rgh8 Nh4 Rxh4 Bxb8 g3 Bxg3 Qg7 Qe3 R8h5 Kf2 Rxf5+ Ke1 Rxf1+ Kxf1 Rh6 Ke2 Rf6 Qg1 Ke8 Bf2 Qxg1 Bxg1 Rf7 Bf2 Rh7 Be1 Rh2+ Kd1 b5 axb5 a4 Rb4 cxb5 Rxb5 Kf7 Bd2 Ke6 Bg5 Rh8 Kd2 Rg8 
37   +4.68   08:42   4522M  f5 g1=Q+ Rxg1 Nxg1 Qxh5 Bxd4 Qg4+ Kf6 Ng6 Nf3+ Kg2 Qe8 Nxf8 Qxf8 Kxf3 Ke7 Rd1 Qh8 Qg5+ Kd7 Bxf7 Be5 Bxe5 Qxe5 Be6+ Kc7 Qe7+ Kb6 Qxd6 Qxb2 Qd8+ Ka7 Qxa5+ Kb8 Qd8 Ka7 Bxc8 Rxc8 Qd4+ Qxd4 Rxd4 Re8 Kf4 c5 Rd2 b5 axb5 Kb6 e5 Kxb5 f6 Kc6 Kf5 Rh8 e6 c4 

[jjoshua2]

Leelenstein 12.1 finds it in 1 node!

Code: Select all

go nodes 1
info depth 1 seldepth 1 time 36 nodes 1 score cp 204
info string h4g2  (874 ) N:       0 (+ 0) (P: 12.51%) (Q:  1.00000) (D:  0.000) (U: 0.40330) (Q+U:  1.40330) (V:  -.----)  C:00000000
info string h2g2  (397 ) N:       0 (+ 0) (P: 12.55%) (Q:  1.00000) (D:  0.000) (U: 0.40467) (Q+U:  1.40467) (V:  -.----)  C:00000000
info string f4f5  (830 ) N:       0 (+ 0) (P: 18.57%) (Q:  1.00000) (D:  0.000) (U: 0.59865) (Q+U:  1.59865) (V:  -.----)  C:00000000
bestmove f4f5

T40B.4-160 does not, but does find it with go nodes 10. LS 10.0 also finds it in 1 node. DQ 2.0 didn't find it within 10k nodes. J13B.2-136 did not with 1,10,100 but did with 1000 nodes.

[jp]

T40B.4-160 does not, but does find it with go nodes 10. LS 10.0 also finds it in 1 node. DQ 2.0 didn't find it within 10k nodes. J13B.2-136 did not with 1,10,100 but did with 1000 nodes.

Okay, but what PVs do they come up with? Previously we've seen examples where everyone is excited that it gets the first move "instantly", but overlook that it doesn't get the correct follow-up moves and even after huge calculation can't see the win.

[jjoshua2]

Well there is no pv at 1 node, but I can let it get a batch and then think a few seconds here...
111 nodes

Code: Select all

f4f5 g2g1q e1g1 h3g1 d1g1 f8g8 g1e3

LS 12.1 go 1 million nodes finishes at 545176 nodes with this PV

Code: Select all

f4f5 g2g1q e1g1 h3g1 d1h5 a7d4 c1d1 d8f6 d1d4 f8h8 h5g4 g7f8 d4d6 h8h4 g4h4 g1f3 h2h3 c8f5 e4f5 f3h4 d6f6 h4f3 f6f7

I'm not sure how reliable PV is though once it goes more than a few moves. The second half is pretty much random as there are few playouts out there, and thus cpuct is low on those trees so it might not see tactics. But the question is does it play good moves at the root once it gets out there.

[Eelco de Groot]

Nice test for the humans is why Black can not take the Knight with the pawn on move 5, because 6.Qxg6 (or 27. Qxg6) does not seem to work.

[pgn] [Event "?"] [Site "?"] [Date "2019.12.9"] [Time "8:38:56"] [Round "?"] [White "Eelco de Groot"] [Black "Chess Tiger 2007.1 engine - Gambit style"] [Result "*"] [Setup "1"] [FEN "r1bq1r2/bp3pk1/2pp4/p6p/P1BPPP1N/6Bn/1P4pK/2RQR3 w - - 0 22"] {0 22} 22. f5 g1R 23. Rxg1 Nxg1 24. Qxh5 Bxd4 25. Qg4+ Kf6 26. Ng6 fxg6 $2 *[/pgn]


I only have my old AMD Athlon, the first one from around 2005 at the moment, so was testing this with Chess Tiger 2007 Old ChessPartner interface. Chess Tiger has some sort of limit that the search explodes after 19 iterations. Imagine what it could do if Christophe dusted it off and polished Tiger up a bit. Deep Tiger.

n