Shocking misevaluation by SF

Spliffjiffer · Post by **Spliffjiffer** » Sat Aug 05, 2017 4:45 am

im talking about the sinquefield cup and am refering to the game between anand vs carlsen, played on friday 04.08.2017:

SF's evaluation of the endgame was so disapointing wrong that i build up following position to verify the impact of such a big misevaluation...quickly i came accross this, very similar to the endgame occured in the game:
[d]R7/4k3/5pp1/7p/7P/p4NP1/3n1PK1/r7 b - - 0 58

here SF stubbornly wants to tade the knights and enters the completely drawn endgame as occured in the game

...this is really bad play by SF because black can "simply" maintain good winning-chances by playing 1...nc4!? for example in my opinion, ne4 might be worth a try too...i didnt investigate

for example:
1. ... Nc4 2. Nd4 Kd6 3. Rd8+ Kc7 4. Ra8 Ra2 5. Ra4 Nd6 6. Ra6 Kd7
7. Ra8 f5 8. Ra7+ Kc8 9. Kf3 Rd2 10. Nc6 Rc2 11. Ne7+ Kd8 12. Nxg6 Nb5
13. Ra8+ Kc7 14. Nf4 a2 15. Nd5+ Kb7 16. Ra4 Rd2 17. Nb4 Rd4 18. Ra5 Rxb4 19.Rxa2 and this should be winning !

with all the respect to the SF-team and the great achievements they have done (i love you !)...here is work to do !

btw: the 2nd best move, according to SF, (nb3?!) leads to nowhere else as to its main-line with nxf3 i think

tpoppins · Post by **tpoppins** » Sat Aug 05, 2017 6:05 am

Spliffjiffer wrote:..quickly i came accross this, very similar to the endgame occured in the game:
[d]R7/4k3/5pp1/7p/7P/p4NP1/3n1PK1/r7 b - - 0 58
here SF stubbornly wants to trade the knights and enters the completely drawn endgame as occurred in the game :-(

A bad example. It takes asmFish + 6-men Syzygy less than 200 MN to go over -3.00 and not much more to reach -10.00. A completely drawn endgame this certainly is not.

Furthermore nothing like that came up in the actual game. By the time the black a-pawn became passed the knights had long been off the board, and they came off the board not as the result of a direct exchange.

[pgn][Event "Sinquefield Cup"]
[Site "0:15:33-0:47:33"]
[Date "2017.08.04"]
[EventDate "2017.08.02"]
[Round "3"]
[Result "1/2-1/2"]
[White "Anand"]
[Black "Carlsen"]
[ECO "C78"]
[WhiteElo "?"]
[BlackElo "?"]
[PlyCount "2"]

1.e4 e5 2.Nf3 Nc6 3.Bb5 a6 4.Ba4 Nf6 5.O-O b5 6.Bb3 Bc5 7.Nc3 O-O 8.d3 d6 9.Nd5 h6 10.c3 Rb8 11.Re1 Ba7 12.Be3 Bxe3 13.Nxe3 Re8 14.a4 b4 15.a5 bxc3 16.bxc3 Be6 17.Bxe6 Rxe6 18.Nd5 Ne7 19.Nxf6+ Rxf6 20.d4 Ng6 21.g3 Qc8 22.Re3 Rb5 23.Nd2 h5 24.h4 Qh3 25.Qf1 Qg4 26.Qe2 Qxe2 27.Rxe2 Re6 28.Nf1 Nf8 29.Rc2 exd4 30.cxd4 c5 31.dxc5 dxc5 32.Nd2 Rd6 33.Nc4 Rd4 34.Kg2 Ne6 35.Ra3 g6 36.Nd2 Kg7 37.Rac3 Rdb4 38.Ra2 Rd4 39.Rac2 Ra4 40.Nb3 Rxb3 41.Rxb3 Nd4 42.Rcb2 Nxb3 43.Rxb3 Rxe4 44.Rb6 Re6 45.Rb7 c4 46.Rc7 Re5 47.Rxc4 Rxa5 48.Rc6 Ra2 49.Kf3 a5 50.Ra6 a4 51.Ke3 a3 52.Kf3 f6 53.Ra7+ Kf8 54.Kg2 Ra1 55.Kf3 Ke8 56.Ra6 Ke7 57.Kg2 Kf7 58.Kf3 Ra2 59.Kg2 g5 60.g4 gxh4 61.gxh5 Ra1 62.Ra7+ Kg8 63.h6 Kh8 64.Kh2 f5 65.f4 a2 66.Kg2 1/2-1/2
[/pgn]

Lyudmil Tsvetkov · Post by **Lyudmil Tsvetkov** » Sat Aug 05, 2017 6:10 am

Spliffjiffer wrote:im talking about the sinquefield cup and am refering to the game between anand vs carlsen, played on friday 04.08.2017:

SF's evaluation of the endgame was so disapointing wrong that i build up following position to verify the impact of such a big misevaluation...quickly i came accross this, very similar to the endgame occured in the game:
[d]R7/4k3/5pp1/7p/7P/p4NP1/3n1PK1/r7 b - - 0 58

here SF stubbornly wants to tade the knights and enters the completely drawn endgame as occured in the game
...this is really bad play by SF because black can "simply" maintain good winning-chances by playing 1...nc4!? for example in my opinion, ne4 might be worth a try too...i didnt investigate

for example:
1. ... Nc4 2. Nd4 Kd6 3. Rd8+ Kc7 4. Ra8 Ra2 5. Ra4 Nd6 6. Ra6 Kd7
7. Ra8 f5 8. Ra7+ Kc8 9. Kf3 Rd2 10. Nc6 Rc2 11. Ne7+ Kd8 12. Nxg6 Nb5
13. Ra8+ Kc7 14. Nf4 a2 15. Nd5+ Kb7 16. Ra4 Rd2 17. Nb4 Rd4 18. Ra5 Rxb4 19.Rxa2 and this should be winning !

with all the respect to the SF-team and the great achievements they have done (i love you !)...here is work to do !

btw: the 2nd best move, according to SF, (nb3?!) leads to nowhere else as to its main-line with nxf3 i think

exchanging knights is the easiest way to a win.

for example, this one is won:

[d]8/1R6/5pp1/7p/5K1P/p5P1/k4P2/r7 w - - 0 8

only real mistake is pushing the black pawn to a2, pawn on a2 with rook on a1, that is already a draw.

with pawn on a3 still, the black king avoids permanent checks by hiding on a2, with an easy win.

did Carlsen drew this?

Spliffjiffer · Post by **Spliffjiffer** » Sat Aug 05, 2017 6:32 am

oh, i misplaced the white rook wrongly on a8

...it belongs to a6 instead...ill try to build a reasonable example to show the impact of such a misevaluation...when i hopefully have soon time for that...sry again

shrapnel · Post by **shrapnel** » Sat Aug 05, 2017 6:33 am

Nothing surprising.
Stockfish always plays safe..... for Draws.
Strong Engine, no doubt, but most boring Engine.
Komodo and Houdini play more brilliant, real chess, but unfortunately also lose more often ; so everyone uses Stockfish or its derivatives.

MikeB · Post by **MikeB** » Sun Aug 06, 2017 5:51 am

Spliffjiffer wrote:oh, i misplaced the white rook wrongly on a8 ...it belongs to a6 instead...ill try to build a reasonable example to show the impact of such a misevaluation...when i hopefully have soon time for that...sry again

with R on a6, current dev-SF takes about about 1:36 to see Nb3, although perhaps Nxf3 here is still winning ( did not check it)
[d]8/4k3/R4pp1/7p/7P/p4NP1/3n1PK1/r7 b - - 0 1

Code: Select all

dep	score	nodes	time	&#40;not shown&#58;  tbhits	knps	seldep&#41;
 48	 -3.97!	5.80G	3&#58;33.73	Nb3! 
 48	 -3.67!	4.61G	2&#58;49.73	Nb3! 
 48	 -3.46!	3.49G	2&#58;08.30	Nb3! 
 48	 -3.30!	2.64G	1&#58;36.87	Nb3! 
 48	 -3.22?	380.7M	0&#58;12.87	Nxf3 Kxf3? 
 48	 -3.30?	353.7M	0&#58;11.92	Nxf3 Kxf3? 
 47	 -3.37 	99.2M  	0&#58;03.45	Nxf3 Ra7+ Ke6 Ra6+ Kd5 Kxf3 a2 Ke2 Kc4 Ra7 Kb5 Ra8 Kb4 Ra6 Kc5 Ra5+ Kd4 Ra6 Kc3 Ra7 Kc4 Ra6 Kd5 Ra7 Kc6 Ra8 Kd6 Ra6+ Kc7 Ra3 Kb7 Kf3 Kc6 Ra6+ Kc7 Ra4 Kb6 Ra3 Kb5 Ra7 Kb4 Ra6 Kb3 Rb6+ Ka3 Ra6+ Kb2 Rb6+ Kc1 Ra6 Kd2 Ra3 Kc2 Ra6 
 46	 -3.37 	97.1M  	0&#58;03.37	Nxf3 Ra7+ Ke6 Ra6+ Kd5 Kxf3 a2 Ke2 Kc4 Ra7 Kb5 Ra8 Kb4 Ra6 Kc5 Ra5+ Kb6 Ra8 Kb5 Ra7 Kc6 Ra8 Kd6 Ra6+ Kc7 Ra3 Kb7 Kf3 Kc6 Ra6+ Kc7 Ra4 Kb6 Ra3 Kb5 Ra7 Kb4 Ra6 Kb3 Rb6+ Ka3 Ra6+ Kb2 Rb6+ Kc1 Ra6 Kd2 Ra3 Kc2 Ra6 Kc3 Ra8 Kb3 Rb8+ 
 45	 -3.37 	89.6M  	0&#58;03.11	Nxf3 Ra7+ Ke6 Ra6+ Kd5 Kxf3 a2 Ke2 Kc4 Ra7 Kb5 Ra8 Kb4 Ra6 Kc5 Ra5+ Kb6 Ra8 Kb5 Ra7 Kc6 Ra8 Kd6 Ra6+ Kc7 Ra3 Kb7 Kf3 Kc6 Ra6+ Kc7 Ra4 Kb6 Ra3 Kb5 Ra7 Kb4 Ra6 Kb3 Rb6+ Kc2 Rc6+ Kb2 Rb6+ Kc3 Rc6+ Kd4 Ra6 Kd3 Ra4 Kc2 Ra6 
 44	 -3.37 	88.9M  	0&#58;03.09	Nxf3 Ra7+ Ke6 Ra6+ Kd5 Kxf3 a2 Ke2 Kc4 Ra7 Kb5 Ra8 Kb4 Ra6 Kc5 Ra5+ Kb6 Ra8 Kb5 Ra7 Kc6 Ra8 Kd6 Ra6+ Kc7 Ra3 Kb7 Kf3 Kc6 Ra6+ Kc7 Ra4 Kb6 Ra3 Kb5 Ra7 Kb4 Ra6 Kb3 Rb6+ Kc2 Rc6+ Kb2 Rb6+ Kc3 Rc6+ Kd4 Ra6 Kd3 Ra4 Kc2 Ra6

latest -dev-SF-McB likes Nb3 almost from the beginning - but it may be luck

Code: Select all

dep	score	nodes	time	&#40;not shown&#58;  tbhits	knps	seldep&#41;
 39	 -4.60 	4.82G	3&#58;04.81	Nb3 Ra8 Kd6 Rb8 Na5 Ra8 Nc4 Nd4 Ra2 Kf3 Kc5 Ne6+ Kb4 Nd4 Rb2 Ne2 Ne5+ Ke3 Ng4+ Kf3 Kb3 Ra7 Rd2 Rb7+ Kc2 Rc7+ Kb2 Rb7+ Ka1 Nc1 Rb2 Ra7 Rxf2+ Ke4 a2 Kd5 Kb1 Nxa2 Rxa2 Rc7 Ra6 Rc6 Rxc6 Kxc6 Kb2 Kd7 Ne5+ Ke7 
 38	 -4.52 	3.58G	2&#58;17.52	Nb3 Rb6 Nc5 Rc6 Nd3 Ra6 a2 Ra4 f5 Nd2 Ne1+ Kh2 Kd7 Ra3 Kc7 Nf1 Nd3 Ne3 Nb4 Ra4 Rc1 Nd5+ Nxd5 Rxa2 Rc6 Kg2 Kd6 Ra5 Ke6 Ra2 Nc3 Rc2 Kd5 Rd2+ Ke4 Rd8 Ke5 Re8+ Kf6 f3 Ra6 Rd8 Ke5 Re8+ Kd4 Kh3 Nd5 Rd8 Rb6 Kg2 Ra6 Kf2 
 37	 -4.42 	2.59G	1&#58;39.63	Nb3 Rb6 Nc5 Rc6 Nd3 Ra6 a2 Ra4 f5 Nd2 Ne1+ Kh2 Kd7 Ra3 Kc7 Nf1 Nd3 Ne3 Nb4 Ra4 Rc1 Nd5+ Nxd5 Rxa2 Rc6 Kg2 Kd6 Ra5 Ke6 Kf3 Rc3+ Ke2 Rc2+ Kf3 Nb4 Ke3 Rb2 Ra7 Ke5 Ra5+ Kf6 Ra4 Nd5+ Kf3 Nc3 Ra5 Ne4 
 36	 -4.42 	2.50G	1&#58;36.10	Nb3 Rb6 Nc5 Rb5 Ne4 Nd4 Rd1 Nc2 a2 Ra5 Nc3 Ra7+ Ke6 Ra6+ Rd6 Ra3 Rd2 Nb4 Rb2 Nxa2 Nxa2 Ra6+ Kf7 Ra7+ Kf8 Ra6 Kg7 Ra7+ Kh6 Ra3 Nb4 Re3 Rd2 Re6 Nd5 Re1 f5 Kf1 Nb4 Re2 Rxe2 Kxe2 Nd5 Kf1 Kg7 Ke1 Kf6 Kf1 Nc3 
 35	 -4.32 	2.37G	1&#58;30.86	Nb3 Rb6 Nc5 Rc6 Nd3 Ra6 a2 Ra4 f5 Nd2 Ne1+ Kh2 Kd7 Ra3 Kc7 Nb3 Rb1 Rxa2 Rxb3 Ra7+ Rb7 Ra3 Kd7 Kg1 Rb1 Ra5 Ke6 Kf1 Nc2+ Kg2 Rb6 Kf3 Nd4+ Ke3 Nb3 Ra3 Kf6 f3 Re6+ Kd3 Nc1+ Kd4 Ne2+ Kd5 Nxg3 
 34	 -3.94 	1.31G	0&#58;49.82	Nb3 Ra8 Kd6 Rb8 Na5 Ra8 Nc4 Rd8+ Kc7 Ra8 Ra2 Nd4 Kd6 Rc8 Kd5 Nb5 Rb2 Rd8+ Ke6 Nd4+ Ke7 Nc6+ Kf7 Ra8 a2 Nd4 Ke7 Kf3 Kd6 Ke4 Kc5 Kd3 Rd2+ Kc3 Rxd4 Rxa2 Rd6 Ra4 Ne5 f4 Ng4 Ra5+ Kb6 Ra2 Ne3 
 34	 -3.69!	816.6M	0&#58;31.16	Nb3! 
 33	 -3.32 	435.3M	0&#58;16.39	Nb3 Ra4 Kd6 Nd4 Nxd4 Rxd4+ Kc5 Ra4 a2 Ra7 Kc4 Ra8 Kb3 Rb8+ Kc3 Ra8 Kb4 Rb8+ Ka4 Ra8+ Kb5 Ra7 Kb4 Ra8 Kc3 Kf3 Kd2 Kg2 Ke1 Ra5 Kd1 Kf3 Kd2 Kg2 Kc2 Ra8 Kd3 Ra6 
 33	 -3.23?	245.6M	0&#58;09.28	Nb3 Ng1? 
 32	 -3.30 	235.8M	0&#58;08.92	Nb3 Ng1 Nc1 Kf3 a2 Ne2 Nxe2 Kxe2 Kd7 Ra3 Kc6 Ra6+ Kb5 Ra8 Kc4 Rc8+ Kb3 Rb8+ Ka3 Ra8+ Kb4 Rb8+ Ka4 Ra8+ Kb5 Rb8+ Kc6 Ra8 g5 Ra6+ Kd5 Ra8 Ke5 Ra5+ Kd4 Ra8 Kc3 Rc8+ Kb4 Rb8+ Ka5 Ra8+ Kb5 Rb8+ Kc6 
 31	 -3.31 	216.8M	0&#58;08.23	Nb3 Ng1 Nc1 Kf3 a2 Ne2 Nxe2 Kxe2 Kd7 Ra3 Kc6 Ra6+ Kb7 Ra3 Kb6 Rb3+ Kc6 Ra3 Kc5 Ra6 Kc4 Ra7 Kd5 Ra5+ Kd6 Kf3 Kc7 Ra7+ Kb6 Ra8 Kc5 Rc8+ Kd4 Rd8+ Kc3 Rc8+ Kb3 Rb8+ Kc4 Rc8+ Kd5 Rd8+ Ke6 Ra8 
 31	 -3.27!	184.4M	0&#58;07.03	Nb3! 
 31	 -2.97!	82.7M  	0&#58;03.16	Nb3! 
 30	 -2.67 	58.9M  	0&#58;02.26	Nb3 Ng1 Nc1 Kf3 a2 Ne2 Rb1 Ra7+ Kd6 Ra6+ Kd7 Nxc1 a1=Q Rxa1 Rxa1 Ne2 g5 Kg2 Ke6 hxg5 fxg5 Nd4+ Kf6 Nf3 Ra2 Nd4 Ra4 Nf3 Kg6 Nd2 Kf5 Kf1 Ra2 Nc4 Ke6 Ne3 Ra4 Kg2 
 29	 -2.67 	56.3M  	0&#58;02.16	Nb3 Ng1 Nc1 Kf3 a2 Ne2 Rb1 Ra7+ Kd6 Ra6+ Kd7 Nxc1 a1=Q Rxa1 Rxa1 Ne2 g5 Kg2 Ke6 Nd4+ Ke5 Nf3+ Kf5 hxg5 fxg5 Nd4+ Kf6 Nf3 Ra2 Nd4 Ra4 Nf3 Kg6 Nd2 Kf5 Kf1 Ra2 Nc4 
 28	 -2.67 	52.1M  	0&#58;02.00	Nb3 Ng1 Nc1 Kf3 a2 Ne2 Rb1 Ra7+ Kd6 Ra6+ Kd7 Nxc1 a1=Q Rxa1 Rxa1 Ne2 g5 Kg2 Ke6 Nd4+ Ke5 Nf3+ Kf5 hxg5 fxg5 Nd4+ Kf6 Nf3 Ra2 Nd4 Ra4 Nf3 Kg6 Nd2 Kf5 Kf1 Ra2 Nc4 
 27	 -2.67 	37.3M  	0&#58;01.43	Nb3 Ng1 Nc1 Kf3 a2 Ne2 Rb1 Ra7+ Kd6 Ra6+ Kd7 Nxc1 a1=Q Rxa1 Rxa1 Ne2 g5 Kg2 Ke6 Nd4+ Ke5 Nf3+ Kf5 hxg5 fxg5 Nd4+ Kf6 Nf3 Ra2 Nd4 Ra4 Nf3 Kg6 Nd2 Kf5 Kf1 
 26	 -2.67 	32.3M  	0&#58;01.23	Nb3 Ng1 Nc1 Kf3 a2 Ne2 Rb1 Ra7+ Kd6 Ra6+ Kd7 Nxc1 a1=Q Rxa1 Rxa1 Ne2 g5 Kg2 Ke6 Nd4+ Ke5 Nf3+ Kf5 hxg5 fxg5 Nd4+ Kf6 Nf3 Ra2 Nd4 Ra4 Nf3 Kg6 Nd2 Kf5 Kf1 
 25	 -2.67 	29.0M  	0&#58;01.10	Nb3 Ng1 Nc1 Kf3 a2 Ne2 Rb1 Ra7+ Kd6 Ra6+ Kd7 Nxc1 a1=Q Rxa1 Rxa1 Nd3 g5 hxg5 fxg5 Nc5+ Ke7 Kg2 Ra3 Kh3 Kf6 Ne4+ Kg6 Kg2 Kf5 Nd2 Ra2 Nf3 Kf6 Nd4 Ra3 Nf3 Ra4 Nd2 Ke6 
 24	 -2.64 	13.5M  	0&#58;00.52	Nb3 Ng1 Nc1 Kf3 a2 Ne2 Rb1 Ra7+ Kf8 Nxc1 a1=Q Rxa1 Rxa1 Ne2 g5 hxg5 fxg5 Kg2 Kf7 Nc3 Ra3 Ne4 Kg6 Kf1 Kf5 Nd6+ Ke6 Nc4 Ra7 Kg2 Ra4 Nb6 Rd4 Nc8 
 23	 -2.40 	8.94M  	0&#58;00.35	Nb3 Ng1 Nc1 Kf3 a2 Ne2 Rb1 Ra7+ Kf8 Nxc1 a1=Q Rxa1 Rxa1 Ne2 g5 hxg5 fxg5 Kg2 Kf7 Nc3 Kf6 Ne4+ Kf5 Nd6+ Ke5 Nf7+ Kf6 Nd6 Ra2 Ne4+ Kg6 Kh3 Ra6 Kg2 
 22	 -2.45 	5.62M  	0&#58;00.22	Nb3 Ng1 Nc1 Kf3 a2 Ne2 Rb1 Ra7+ Ke6 Ra6+ Kf7 Ra7+ Kf8 Nxc1 a1=Q Rxa1 Rxa1 Ne2 Ra3+ Kg2 g5 Nd4 Kf7 Ne2 Kg7 hxg5 fxg5 Nd4 Ra4 
 21	 -2.42 	4.88M  	0&#58;00.19	Nb3 Ng1 Nc1 Kf3 a2 Ne2 Rb1 Ra7+ Kf8 Nxc1 a1=Q Rxa1 Rxa1 Ne2 Ra3+ Kg2 g5 hxg5 fxg5 Nd4 Kf7 Nc2 Ra2 Nb4 Rd2 Na6 Ke6 Nc5+ Kf5 
 20	 -2.30 	2.72M  	0&#58;00.11	Nb3 Ng1 Nc1 Kf3 a2 Ne2 Rb1 Ra7+ Kf8 Ra8+ Kg7 Ra7+ Kh6 Nxc1 a1=Q Rxa1 Rxa1 Ne2 Ra3+ Kg2 g5 hxg5+ fxg5 Nd4 Ra4 Nf5+ Kg6 Ne3 Kf6 f4 Ra2+ Kf1 Rh2 Nc4 gxf4 
 19	 -2.25 	2.20M  	0&#58;00.09	Nb3 Ng1 Nc1 Kf3 a2 Ne2 Rb1 Ra7+ Kd6 Ra6+ Ke5 Ra5+ Ke6 Ra6+ Kf7 Ra7+ Kf8 Ra8+ Kg7 Ra7+ Kh6 Nxc1 a1=Q Rxa1 Rxa1 Ne2 Ra3+ Kg2 
 18	 -2.32 	2.00M  	0&#58;00.08	Nb3 Ng1 Nc1 Kf3 a2 Ne2 Rb1 Ra7+ Kd6 Ra6+ Ke5 Ra5+ Ke6 Ra6+ Kf7 Ra7+ Kf8 Ra8+ Kg7 Ra7+ Kh6 Nxc1 a1=Q Rxa1 Rxa1 Ne2 Ra3+ Kg2 Kg7 Nf4 Kf7 Kh3 Rf3 Kg2 Rb3 Ne2 Ra3 
 17	 -2.23 	1.88M  	0&#58;00.08	Nb3 Ng1 Nc1 Kf3 a2 Ne2 Rb1 Ra7+ Kf8 Ra8+ Kg7 Ra7+ Kh6 Nxc1 a1=Q Rxa1 Rxa1 Ne2 Ra3+ Kg2 Kg7 Nf4 Kf7 Ne2 f5 Nf4 
 16	 -2.22 	1.77M  	0&#58;00.07	Nb3 Ng1 Nc1 Kf3 a2 Ne2 Rb1 Ra7+ Kd6 Ra6+ Kd7 Ra7+ Ke8 Ra8+ Kf7 Ra7+ Kf8 Ra8+ Kg7 Ra7+ Kh6 Nxc1 a1=Q Rxa1 Rxa1 
 15	 -2.31 	1.44M  	0&#58;00.06	Nb3 Ng1 Nc1 Kf3 a2 Ne2 Rb1 Ra7+ Kf8 Ra8+ Kg7 Ra7+ Kh6 Nxc1 a1=Q Rxa1 Rxa1 Ne2 Ra3+ Kg2 Kg7 Nf4 f5 Nd5 Kf7 Ne3 Rd3 Kf3 Rb3 
 14	 -2.32 	1.18M  	0&#58;00.05	Nb3 Ng1 Nc1 Kf3 a2 Ne2 Rb1 Ra7+ Kf8 Nxc1 a1=Q Rxa1 Rxa1 Nd3 Ra3 Ke3 Kf7 Ke4 Ra4+ Kd5 Ra2 Ke4 Ra3 Nf4 Rb3 Nd5 
 13	 -2.20 	863733	0&#58;00.04	Nb3 Ng1 Nc1 Kf3 a2 Ne2 Rb1 Ra7+ Kf8 Nxc1 a1=Q Rxa1 Rxa1 Nd3 Ra3 Ke4 g5 Nb4 Kf7 Nd5 Rb3 hxg5 
 12	 -2.28 	398259	0&#58;00.02	Nb3 Ng1 Nc1 Kf3 a2 Ra7+ Kd6 Ra6+ Kc5 Ne2 Kb5 Ra7 Nxe2 Kxe2 Kb6 
 11	 -2.35 	317899	0&#58;00.02	Nb3 Ng1 Nc1 Kf3 a2 Ra7+ Kd6 Ne2 Rb1 Ra6+ Ke5 Ra5+ Ke6 Nxc1 a1=Q Rxa1 Rxa1 
 10	 -2.19 	141795	0&#58;00.01	Nb3 Ra7+ Ke6 Ra6+ Kf5 Rb6 Nc1 Nd4+ Ke5 Nc2 Ra2 Ne3 Nd3 
  9	 -1.99 	57959  	0&#58;00.00	Nb3 Ra7+ Kf8 Ra6 a2 Rxf6+ Ke7 Ra6 Nc5 Ra7+ Kf6 Nd4 
  8	 -1.67 	38127  	0&#58;00.00	Nxf3 Kxf3 a2 Ke2 g5 Ra5 gxh4 gxh4 
  7	 -2.07 	20039  	0&#58;00.00	Nb3 Ra7+ Ke6 Ra6+ Kf5 Rb6 Nc5 Nd4+ Ke5 Nc2 
  6	 -3.12 	13509  	0&#58;00.00	Nb3 Nh2 Nc5 Ra5 Nd3 Ra6 
  5	 -3.09 	4668    	0&#58;00.00	Nb3 Ra7+ Kd6 Ng1 a2 Nf3 
  4	 -9.69 	1907    	0&#58;00.00	Nb3 Nh2 Nc5 g4 
  3	 -3.26 	1168    	0&#58;00.00	Nb3 Ne1 Rxe1 Rxa3 
  2	 -3.87 	596      	0&#58;00.00	Ne4 Ng1 
  1	 -2.39 	227      	0&#58;00.00	Ne4 
  1	Found 510 tablebases 
  0	#

one can see from the results that dev-SF-McB does less pruning - so it's not the evaluation function ( in my opinion) ,but it's the search

this is just one position , for the record dev-SF is stronger than dev-SF-McB , but they are close...

Spliffjiffer · Post by **Spliffjiffer** » Sun Aug 06, 2017 8:34 am

tested with the latest asmfish (05.08.2017), no TB's :

on my pc asmfish didnt refused to take on f3 with 40/2h time-control

[d]8/4k3/R4pp1/6np/7P/p4NP1/5PK1/r7 b - - 0 58

if 1...nxf3 its a dead draw, no surprise of course

for example:

1...nxf3 2. kxf3 a2, 3. kf4 kd7, 4.ra3 kc6, 5. rc3 kb5, 6. rc2 kb4, 7. re2 kc3, 8.ke4 f5, 9.kf4 kd3, 10.rb2 and draw

1...nxf3 2. kxf3 ra2, 3.ke3 is also completely draw of course

now to the thing thats called "playing chess and try to play for a win"

after realizing that the rook-endgame is a totally dead draw, there is only one way to go for a win...keeping the knight on the board, at least for now !

i think 1...ne6 gives good winning chances and its due to SF's completely wrong evaluation that it is not switching to this move immediately or at least quickly

...thats the sad thing i wanted to talk about...not wether the position has occured in the anand-carlsen game or not !...of course i apologize for posting the wrong position !

just one variation... to show how alive the position is instead of a dead draw !

1...ne6 !?, 2. nd2 nc5, 3. ra7+ ke6, 4. ra8 a2, 5. ra3 kd5, 6. ra5 kc6, 7. kf3 nb7, 8. ra8 kb5 9. ne4 na5, 10. nc3+ kc4, 11. rc8+ kd3, 12. nxa2 rxa2 13. rd8 kc3, 14. rd6 nc4, 15. rxf6 kd4...

[d]8/8/5Rp1/7p/2nk3P/5KP1/r4P2/8 w - - 1 73

you might think that this is a fortress and black can make no progress...well, you are not the only one to think like that...SF is on flat-line course and is pretty clueless about how to proceed...but its winning !
i didnt let play SF vs itself so i dont really know wether it will find the win on the verge of the 50 move rule or not but for myself SF was not really a help to do the job

...another problem where SF will surely improve in the future, so give it a smile

lets go on with it...

16. rf4+ kd5 17. rf6 ne5+ 18. kg2 ng4, 19. rf4 ke6 20. kf1 ke7, 21. kg2 ne5, 22. re4 kf6, 23. kg1 rd2, 24. rf4+ kg7, 25. re4 nd3 26. re7+ kf6, 27. re8 kf5, 28. rf8+ ke6, 29. kf1 ra2 30. rf3 ne5, 31. rf4 ke7, 32. re4 kf6, 33. kg1 rd2, 34. rf4+ kg7 35. re4 nd3, 36. re7+ kf6, 37. re8 kf5, 38. rf8+ ke6, 39. kf1 ra2, 40. rf3 ra1+, 41. ke2 ne5, 42. rf4 ke7, 43. re4 kf6, 44. re3 ng4, 45. re8 kg7, 46. re7+ kh6...

[d]8/4R3/6pk/7p/6nP/6P1/4KP2/r7 w - - 63 104

the rest SF "understands" pretty well

...only resigning it still doesnt want to

Lyudmil Tsvetkov · Post by **Lyudmil Tsvetkov** » Sun Aug 06, 2017 9:28 am

Spliffjiffer wrote:tested with the latest asmfish (05.08.2017), no TB's :

on my pc asmfish didnt refused to take on f3 with 40/2h time-control

[d]8/4k3/R4pp1/6np/7P/p4NP1/5PK1/r7 b - - 0 58

if 1...nxf3 its a dead draw, no surprise of course
for example:

1...nxf3 2. kxf3 a2, 3. kf4 kd7, 4.ra3 kc6, 5. rc3 kb5, 6. rc2 kb4, 7. re2 kc3, 8.ke4 f5, 9.kf4 kd3, 10.rb2 and draw

1...nxf3 2. kxf3 ra2, 3.ke3 is also completely draw of course

now to the thing thats called "playing chess and try to play for a win"

after realizing that the rook-endgame is a totally dead draw, there is only one way to go for a win...keeping the knight on the board, at least for now !

i think 1...ne6 gives good winning chances and its due to SF's completely wrong evaluation that it is not switching to this move immediately or at least quickly ...thats the sad thing i wanted to talk about...not wether the position has occured in the anand-carlsen game or not !...of course i apologize for posting the wrong position !

just one variation... to show how alive the position is instead of a dead draw !

1...ne6 !?, 2. nd2 nc5, 3. ra7+ ke6, 4. ra8 a2, 5. ra3 kd5, 6. ra5 kc6, 7. kf3 nb7, 8. ra8 kb5 9. ne4 na5, 10. nc3+ kc4, 11. rc8+ kd3, 12. nxa2 rxa2 13. rd8 kc3, 14. rd6 nc4, 15. rxf6 kd4...

[d]8/8/5Rp1/7p/2nk3P/5KP1/r4P2/8 w - - 1 73

you might think that this is a fortress and black can make no progress...well, you are not the only one to think like that...SF is on flat-line course and is pretty clueless about how to proceed...but its winning !
i didnt let play SF vs itself so i dont really know wether it will find the win on the verge of the 50 move rule or not but for myself SF was not really a help to do the job ...another problem where SF will surely improve in the future, so give it a smile
lets go on with it...

16. rf4+ kd5 17. rf6 ne5+ 18. kg2 ng4, 19. rf4 ke6 20. kf1 ke7, 21. kg2 ne5, 22. re4 kf6, 23. kg1 rd2, 24. rf4+ kg7, 25. re4 nd3 26. re7+ kf6, 27. re8 kf5, 28. rf8+ ke6, 29. kf1 ra2 30. rf3 ne5, 31. rf4 ke7, 32. re4 kf6, 33. kg1 rd2, 34. rf4+ kg7 35. re4 nd3, 36. re7+ kf6, 37. re8 kf5, 38. rf8+ ke6, 39. kf1 ra2, 40. rf3 ra1+, 41. ke2 ne5, 42. rf4 ke7, 43. re4 kf6, 44. re3 ng4, 45. re8 kg7, 46. re7+ kh6...

[d]8/4R3/6pk/7p/6nP/6P1/4KP2/r7 w - - 63 104

the rest SF "understands" pretty well ...only resigning it still doesnt want to

that one is already a draw(unless some magical solution still exists)

the problem with SF solving this is that it is difficult to come up with a general rule for correcting evaluation.

- this happens only in very simple endings, with 3 vs 3 on the king side, or 2 vs 2, or 1 vs 1, and a single passed pawn more that is on a2, with the rook on a1
- with pawn on a3, it will depend on whether the king is cut or not, as with Ra6 vs Ra8
- if the pawns on the king side are not symmetrical, than it is unclear when the stronger side wins and when not, all will depend on minor details
- etc., etc.

besides, SF will be reluctant to implement such a specific rule for very small elo gain.

Komodo has the very same problem, I suppose Houdini too.

zullil · Post by **zullil** » Sun Aug 06, 2017 1:51 pm

Spliffjiffer wrote:tested with the latest asmfish (05.08.2017), no TB's :

on my pc asmfish didnt refused to take on f3 with 40/2h time-control

[d]8/4k3/R4pp1/6np/7P/p4NP1/5PK1/r7 b - - 0 58

I'm sure that the Stockfish team would gladly accept code that raises Elo by improving evaluation of rook and pawn endgame positions.

The latest asmFish (with 6-man tables) does switch from Nxf3 to Ne6, though as you point out, it takes some time:

Code: Select all

-3.54 58... Ne6 59. Nd2 Nc5 60. Ra5 a2 61. Ra7+ Kd6 62. Ra5 Ke6 63. Ra3 Kd7 64. Ra5 Kc6 65. Kf3 Ne6 66. Ke2 Kb6 67. Nc4+ Kc7 68. Nd2 Nd4+ 69. Kd3 Nc6 70. Ra4 Nb4+ 71. Rxb4 Rd1 72. Ra4 a1=Q 73. Rxa1 Rxa1 74. Ne4 Ra6 75. Ke2 Kd8 76. Nc3 Ra5 77. Kd3 Ra3 78. Kd2 Kd7 79. Nb5 Ra5 80. Nd4 Kd6 81. Kd3 Ra4 82. Nf3 Ra2 83. Ke3 Kd5 84. Ne1 Ra3+ 85. Kf4 Ra4+ 86. Ke3 Ke5 87. Nd3+ Kf5 &#40;depth 43, 0&#58;03&#58;45&#41;

zullil · Post by **zullil** » Sun Aug 06, 2017 9:00 pm

zullil wrote:
Spliffjiffer wrote:tested with the latest asmfish (05.08.2017), no TB's :

on my pc asmfish didnt refused to take on f3 with 40/2h time-control

[d]8/4k3/R4pp1/6np/7P/p4NP1/5PK1/r7 b - - 0 58

I'm sure that the Stockfish team would gladly accept code that raises Elo by improving evaluation of rook and pawn endgame positions.

The latest asmFish (with 6-man tables) does switch from Nxf3 to Ne6, though as you point out, it takes some time:
Code: Select all
-3.54 58... Ne6 59. Nd2 Nc5 60. Ra5 a2 61. Ra7+ Kd6 62. Ra5 Ke6 63. Ra3 Kd7 64. Ra5 Kc6 65. Kf3 Ne6 66. Ke2 Kb6 67. Nc4+ Kc7 68. Nd2 Nd4+ 69. Kd3 Nc6 70. Ra4 Nb4+ 71. Rxb4 Rd1 72. Ra4 a1=Q 73. Rxa1 Rxa1 74. Ne4 Ra6 75. Ke2 Kd8 76. Nc3 Ra5 77. Kd3 Ra3 78. Kd2 Kd7 79. Nb5 Ra5 80. Nd4 Kd6 81. Kd3 Ra4 82. Nf3 Ra2 83. Ke3 Kd5 84. Ne1 Ra3+ 85. Kf4 Ra4+ 86. Ke3 Ke5 87. Nd3+ Kf5 &#40;depth 43, 0&#58;03&#58;45&#41;

asmFish sees no progress for Black with 58... Ne6. But perhaps the Fish is blind. Do you believe Ne6 is winning?

Code: Select all

-2.32 58... Ne6 59. Nd2 a2 60. Ra7+ Kd8 61. Nf1 Nc5 62. Ne3 Nd3 63. Nd5 Ne1+ 64. Kf1 f5 65. Ra8+ Kd7 66. Nc3 Nf3+ 67. Ke2 Re1+ 68. Kxf3 a1=Q 69. Rxa1 Rxa1 70. Nd5 Ra3+ 71. Kg2 Ke6 72. Nf4+ Kf7 73. Nd5 Ra6 74. Nf4 Ra4 75. Kf3 Ra3+ 76. Kg2 Ra2 77. Kf3 Kf6 78. Nd5+ Ke5 79. Nf4 Ra3+ 80. Kg2 Ra6 81. Kg1 Rc6 82. Kg2 Ke4 83. Ne2 Rd6 84. Nf4 Ke5 85. Kf3 Rb6 86. Nh3 Kf6 87. Kg2 Ra6 88. Ng5 Ra1 89. Nh3 Ra3 90. Kh2 Ra5 91. Kg2 Ra7 92. Kh2 Ke5 93. Kg2 Ra4 94. Kf3 Ra1 95. Nf4 Kf6 96. Nd5+ Ke6 97. Nf4+ Kf7 98. Kg2 Re1 99. Kf3 Re4 100. Kg2 Rd4 101. Kf3 Kf6 102. Kg2 Rd2 103. Nh3 Rc2 104. Kf3 Rc3+ 105. Kg2 Rc7 106. Nf4 Rc6 107. Nh3 Rd6 108. Kh2 Rd8 &#40;depth 78, 6&#58;10&#58;40&#41;

Shocking misevaluation by SF

Shocking misevaluation by SF

Re: Shocking misevaluation by SF

Re: Shocking misevaluation by SF

Re: Shocking misevaluation by SF

Re: Shocking misevaluation by SF

Re: Shocking misevaluation by SF

Re: Shocking misevaluation by SF

Re: Shocking misevaluation by SF

Re: Shocking misevaluation by SF

Re: Shocking misevaluation by SF