Nice TCEC position

[Jouni]

[d]8/6k1/1p2P3/2p3P1/P1Pp2K1/P7/6P1/8 w - - 1 46

Analysis by Stockfish 17:
...
46.Kg3 Kg6 47.Kf3 Kg7 48.Kg4 Kg6 49.Kf4 Kg7 50.Ke4 Kg6 51.Kd3 Kg7 52.Ke2 Kg6 53.Kd1 Kg7 54.Kc1 Kg6 55.Kb2 Kg7 56.Ka2 Kg6 57.Kb1 Kg7 58.Kb2
+- (2.72) Depth: 26/59 00:00:53 1615mN, tb=92

Analysis by Lc0 v0.31.0:
...
46.Kg3 Kg6 47.Kf2 Kg7 48.Kf3 Kg6 49.Kg3 Kg7 50.Kg4 Kg6 51.g3 Kg7 52.Kh5 d3 53.e7 Kf7 54.Kh6 d2 55.g6+ Kxe7 56.g7 d1Q 57.g8Q Qh1+ 58.Kg5 Qc1+
+/= (0.59) Depth: 14/34 00:00:30 151kN, tb=5

[Ajedrecista]

Hello Jouni:

Nice position indeed. 6P vs. 3P and black still saves the day because that d4 passed pawn is a monster that prevents the white king to cross to the fifth rank, adding that two of the white pawns are doubled. All that black must do is move the king from g7 to g6 and back to g7 and so on, unless white plays e7 or g6, when SF 17 immediately sees the draw. The Huntsman agrees with Lc0 and evaluates the position about +0.23 or so in my system since low depths.

Regards from Spain.

Ajedrecista.

[chesskobra]

I think if such a position were to be added to a test suite of positions, points should be awarded for just seeing that it is a draw, and not for finding the best move. Or, for example, best move could be '0' to mean that an engine sees eval 0 or 'offer a draw'. Interesting position as an endgame exercise at beginner level.

[fishpov]

I think, engines which have fortress detect are better to evaluate this kind of positon.
Here Crystal 8:

8/8/1p4k1/2p1P1P1/P1Pp2K1/P7/6P1/8 w - - 0 1

Analysis by Crystal 8:
...
6.e6 Kg7 7.Kf3 Kg6 8.Kf2 Kg7 9.Ke2 Kg6 10.Kd1 Kg7 11.Kc2 Kg6 12.Kb2 Kg7 13.g3 Kg6 14.Kc2 Kg7 15.Kd3
White is clearly better: +- (2.47) Depth: 6/21 00:00:03
...
6.Kf4 Kf7 7.Kf3 Kg6 8.Kg4 Kg7 9.a5 bxa5 10.Kf4 Kf7 11.Kg3 Kg6 12.e6 Kg7 13.Kf4 Kg6
White is slightly better: +/= (0.67 --) Depth: 12/55 00:00:04 140MN
...
6.Kf4 Kg7 7.e6 Kg6 8.Ke4 Kg7 9.Kf3 Kg6 10.Kf4 Kg7 11.Ke4 Kg6 12.Kd3 Kg7 13.Ke2 Kg6 14.Ke1 Kg7 15.Kf1 Kg6 16.Kg1 Kg7 17.g3 Kg6 18.Kf1 Kg7 19.Ke1 Kg6 20.Kd2 Kg7 21.Kc2 Kg6 22.Kd3 Kg7 23.g4 Kg6 24.Ke2 Kg7 25.Kf3 Kg6 26.Kf4 Kg7 27.Kf5 d3 28.g6 d2 29.e7 d1Q 30.e8Q î€Bf1+ 31.Ke6 Qxc4+ 32.Kd6 Qd4+ 33.Kc7 Qf6 34.Qb5 Qxg6 35.a5 bxa5 36.Qxc5 Qxg4 37.Qe5+ Kg6 38.Qe8+ Kg7 39.Qd7+ Qxd7+ 40.Kxd7 a4 41.Kc6 Kf6 42.Kb5 Ke6 43.Kxa4 Kd7 44.Kb5 Kc7 45.Ka6 Kc6 46.a4 Kc5 47.Ka5 Kc6 48.Kb4 Kb6 49.a5+ Kb7 50.Kb5 Ka7 51.a6 Kb8 52.Kc5 Ka7
The position is equal: = (0.00 --) Depth: 25/82 00:01:28 2644MN
6.Kf4 Kg7 7.Kf3 Kg6
The position is equal: = (0.00) Depth: 25/82 00:01:28 2645MN
6.Kf4 Kg7 7.Kf3 Kg6
The position is equal: = (0.00) Depth: 26/30 00:01:28 2646MN

(14.10.2024)

[nmcrazyim5]

Great position! Definitely deserves to be on test suites

[chessica]

[d]8/6k1/1p2P3/2p3P1/P1Pp2K1/P7/6P1/8 w - - 1 46

My slow PC:

Crystal 8:

8/112 00:52,698 27.216k 516k +0,33 Kg4-f4 Kg7-g6 Kf4-g3 Kg6-g7 Kg3-g4 Kg7-g6 g2-g3 Kg6-g7 Kg4-f5 d4-d3 g5-g6 d3-d2 e6-e7 d2-d1D e7-e8D Dd1-f1+ Kf5-e6 Df1xc4+ Ke6-d6 Dc4-d4+ Kd6-c7 Dd4-f6 De8-d7+ Kg7xg6
8/112 00:52,698 27.216k 516k +0,33 Kg4-f4 Kg7-g6 Kf4-g3 Kg6-g7 Kg3-g4 Kg7-g6 g2-g3 Kg6-g7 Kg4-f5 d4-d3 g5-g6 d3-d2 e6-e7 d2-d1D e7-e8D Dd1-f1+ Kf5-e6 Df1xc4+ Ke6-d6 Dc4-d4+ Kd6-c7 Dd4-f6 De8-d7+ Kg7xg6
8/88 00:52,698 27.216k 516k +0,62 Kg4-f3 Kg7-g6 Kf3-f2 Kg6-g7 Kf2-e2 Kg7-g6 Ke2-f1 Kg6-g7 Kf1-f2 Kg7-g6 g2-g3 Kg6-g7 Kf2-f3 Kg7-g6 Kf3-f4 Kg6-g7 Kf4-f5 d4-d3 g5-g6 d3-d2 e6-e7 d2-d1D e7-e8D Dd1-f1+ Kf5-e6 Df1xc4+ Ke6-d6 Dc4-d4+ Kd6-c7
---------------------------------------------------------------------------
9/86 01:23,966 51.183k 610k +0,11 Kg4-f4 Kg7-g6 Kf4-g3 Kg6-g7 Kg3-g4 Kg7-g6 g2-g3 Kg6-g7 a4-a5 b6xa5 Kg4-f4 Kg7-g6 Kf4-f3 Kg6-g7 Kf3-f2 Kg7-g6 g3-g4 Kg6-g7 Kf2-g3 Kg7-g6 Kg3-f3 Kg6-g7 Kf3-f4 Kg7-g6 Kf4-e4 Kg6-g7 g5-g6 Kg7xg6 g4-g5 a5-a4 Ke4-d3 Kg6-g7 e6-e7
9/110 01:23,966 51.183k 610k +0,31 Kg4-f3 Kg7-g6 Kf3-e2 Kg6-g7 Ke2-d3 Kg7-g6 Kd3-c2 Kg6-g7 Kc2-b2 Kg7-g6 Kb2-a1 Kg6-g7 Ka1-a2 Kg7-g6 Ka2-b2 Kg6-g7 g2-g3 Kg7-g6 Kb2-c2 Kg6-g7 Kc2-d3 Kg7-g6 Kd3-e2 Kg6-g7 Ke2-f3 Kg7-g6 Kf3-f4 Kg6-g7 Kf4-f5 d4-d3 g5-g6 d3-d2 e6-e7 d2-d1D e7-e8D Dd1-f1+ Kf5-e6 Df1xc4+ Ke6-d6 Dc4-d4+ Kd6-c7 Dd4-f6 Kc7-b7 Df6xg6 De8-e5+ Dg6-f6 De5-c7+ Kg7-h6 Dc7xb6 Df6xb6+ Kb7xb6 c5-c4
---------------------------------------------------------------------------
10/106 03:28,272 134.495k 646k +0,05 Kg4-f4 Kg7-g6 Kf4-g3 Kg6-g7 Kg3-g4 Kg7-g6 g2-g3 Kg6-g7 Kg4-f4 Kg7-g6 g3-g4 Kg6-g7 Kf4-f3 Kg7-g6 Kf3-e2 Kg6-g7 Ke2-d1 Kg7-g6 Kd1-e1 Kg6-g7 g5-g6 Kg7-f6 g6-g7 Kf6xg7 g4-g5 Kg7-g6 a4-a5 b6xa5 Ke1-d1 Kg6-g7 Kd1-d2 a5-a4 Kd2-d3 Kg7-g6 Kd3-e4 Kg6-g7 Ke4-f5 d4-d3 g5-g6 d3-d2 e6-e7 d2-d1D
10/112 03:28,272 134.495k 646k +0,20 Kg4-f3 Kg7-g6 Kf3-e4 Kg6-g7 Ke4-f5 d4-d3 g5-g6 d3-d2 e6-e7 d2-d1D e7-e8D Dd1-f1+ Kf5-e6 Df1xc4+ Ke6-d6 Dc4-d4+ Kd6-c7 Dd4-f6 Kc7-b7 Df6xg6 De8-b5 Dg6-e6 a4-a5 b6xa5 Db5xc5 De6-b3+ Kb7-a7
---------------------------------------------------------------------------
11/86 05:35,889 257.835k 768k 0,00 Kg4-f4 Kg7-g6 Kf4-f3 Kg6-g7
11/84 05:35,889 257.835k 768k +0,01 Kg4-f3 Kg7-g6 Kf3-f4 Kg6-g7 Kf4-f5 d4-d3 g5-g6 d3-d2 e6-e7 d2-d1D e7-e8D Dd1-f1+ Kf5-e6 Df1xc4+ Ke6-d6 Dc4-d4+ Kd6-c7 Dd4-f6 Kc7-b7 Df6xg6 De8-e5+ Kg7-h6 De5-d5 Dg6-f6 Dd5-b3 Kh6-g7 Db3-g3+ Df6-g6
---------------------------------------------------------------------------
12/28 05:37,675 259.581k 769k 0,00 Kg4-f4 Kg7-g6 Kf4-f3 Kg6-g7
12/34 05:37,675 259.581k 769k 0,00 Kg4-f3 Kg7-g6 Kf3-f4 Kg6-g7
---------------------------------------------------------------------------
13/28 05:40,586 262.133k 770k 0,00 Kg4-f4 Kg7-g6 Kf4-f3 Kg6-g7
13/34 05:40,586 262.133k 770k 0,00 Kg4-f3 Kg7-g6 Kf3-f4 Kg6-g7
---------------------------------------------------------------------------
14/28 05:45,374 266.298k 771k 0,00 Kg4-f4 Kg7-g6 Kf4-f3 Kg6-g7
14/34 05:45,374 266.298k 771k 0,00 Kg4-f3 Kg7-g6 Kf3-f4 Kg6-g7
---------------------------------------------------------------------------

[Dann Corbit]

Great position! Definitely deserves to be on test suites

The problem is that the key move is easy, it is the zero score that is hard.
All the strong programs will choose the right move. Test set calculation programs just look at the chosen key move for scoring the set.
I see an opposite problem a lot. There will be engines that choose the right key move, but score it as 0.0 for a checkmate. Hence, they are clueless on the right plan and might only draw in an actual game. Yet the design of the test set analysis program gives full credit. Perhaps a new type of scoring system like gradualtest {invented by Odd Gunnar Malin IIRC} could be established with a minimum or maximum score for problems. E.g. min=dm7 or max=0.
BTW, a wrong score for a draw is a real problem. In contests that require zero scores for a draw exit or that give a high score over an extended period a winning mark, it is possible to get wrong results or to have a game go on for 400 moves at slow time control when it was really drawn at 100 moves. Equally problematic is a zero score for a win. In a real game, the program with a zero score in a winning position simply does not understand the right plan.

[Vinvin]

Yes, the static evaluation is important : if a couple of move before chose this variation with a +2.50 eval (but really draw), this could prevent the engine to choose a winning variation with a +2.00 Eval.