"IQ test" for engines

[Uri Blass]

Here is stockfish's analysis after more than 2 hours

I do not see a plan in the pv but I can clearly see a huge score

[d]3k4/8/7p/2p1p1pP/1pPpPpP1/1P1P1P2/N7/2K5 w - - 0 1

Stockfish_17100511_x64_modern:

77/86 00:31 236,989k 7,493k +4.16 1.Kc2 Kd7 2.Nc1 Ke6 3.Ne2 Ke7 4.Ng1 Kd6 5.Nh3 Ke7 6.Kd1 Kf6 7.Kd2 Kg7 8.Ng1 Kh7 9.Kc2 Kg8 10.Ne2 Kh7 11.Kb1 Kg7 12.Ng1 Kg8 13.Kc2 Kh8 14.Nh3 Kg7 15.Kd2 Kf8 16.Nf2 Kf7 17.Ke1 Kf6 18.Nh3 Ke7 19.Ng1 Kd7 20.Kf1 Ke7 21.Ke2 Kd7 22.Nh3 Ke7 23.Nf2 Ke6 24.Kd1 Kf7 25.Kd2 Kf8 26.Nd1 Kf7 27.Nb2 Ke7 28.Na4 Kd6 29.Nb6 Ke6 30.Kc2 Kd6 31.Nd5 Ke6 32.Kd1 Kf7 33.Ke2 Ke6 34.Ke1 Kf7 35.Kd2 Ke6 36.Kc2 Kf7 37.Kb2 Ke6 38.Kb1 Kf7 39.Kc1 Ke6 40.Kd2 Kf7 41.Ke1 Ke6 42.Ke2 Kf7 43.Kf2 Ke6
78/94+ 01:42 832,138k 8,120k +4.23 1.Kb2
78/94+ 01:56 965,498k 8,299k +4.31 1.Kb2
78/94+ 01:57 979,855k 8,317k +4.42 1.Kb2
78/94+ 01:58 986,526k 8,325k +4.58 1.Kb2
78/94+ 02:55 1,549,447k 8,823k +4.80 1.Kb2
78/94+ 23:14 13,490,641k 9,673k +5.09 1.Kb2
78/94+ 29:24 17,166,351k 9,728k +5.47 1.Kb2
78/99+ 2:00:13 66,983,778k 9,286k +5.97 1.Kb2
78/99+ 2:00:13 66,984,758k 9,286k +6.62 1.Kb2
78/99+ 2:02:13 68,150,041k 9,293k +7.44 1.Kb2
78/99+ 2:02:22 68,232,120k 9,292k +8.49 1.Kb2
78/99+ 2:20:36 77,684,021k 9,208k +9.82 1.Kb2
78/99 2:21:29 78,135,607k 9,203k +9.82 1.Kb2 Kd7 2.Nc1 Ke8 3.Kc2 Kd8 4.Ne2 Ke7 5.Kb2 Kf6 6.Ka1 Kf7 7.Kb1 Kf8 8.Kb2 Kg7 9.Kc2 Kf6 10.Kb1 Kf7 11.Ng1 Kg8 12.Ka1 Kg7 13.Ka2 Kh8 14.Nh3 Kh7 15.Nf2 Kg8 16.Ka1 Kf7 17.Kb2 Kg8 18.Nd1 Kg7 19.Ka1 Kf7 20.Kb1 Ke7 21.Nb2 Kd7 22.Na4 Kd6 23.Kc1 Kc6 24.Nb2 Kd7 25.Kb1 Kc6 26.Ka1 Kd6 27.Na4 Kc6 28.Kb2 Kd6 29.Ka2 Kc6 30.Nb2 Kc7 31.Nd1 Kc6 32.Kb1 Kb7 33.Kc1 Kc6 34.Kd2 Kb7 35.Ke1 Kc8 36.Nf2 Kd8 37.Kd1 Ke8 38.Kd2 Kd7 39.Kc2 Kd6 40.Nh3 Ke6 41.Kb2 Kf6 42.Nf2

[Uri Blass]

Depth 79 and 80 and 81 are the same except different nonsense pv and maybe the pv is based on some tranposition and in a practical game it can find the right way.

[d]3k4/8/7p/2p1p1pP/1pPpPpP1/1P1P1P2/N7/2K5 w - - 0 1

Stockfish_17100511_x64_modern:

79/86 2:22:23 78,608,251k 9,201k +9.82 1.Kb2 Kd7 2.Nc1 Ke8 3.Kc2 Kd8 4.Kd1 Kc7 5.Ne2 Kd6 6.Kc1 Ke7 7.Kc2 Kd7 8.Ng1 Kd6 9.Kb1 Ke7 10.Ne2 Kd7 11.Ka2 Ke7 12.Kb2 Kf6 13.Kc2 Ke7 14.Kd2 Kd6 15.Kd1 Kc7 16.Ke1 Kd8 17.Ng1 Ke7 18.Nh3 Kf7 19.Kd1 Ke6 20.Ng1 Kd7 21.Kc1 Ke7 22.Kb1 Kd7 23.Nh3 Ke7 24.Ka1 Ke8 25.Nf2 Kf7 26.Kb2 Kg8 27.Kc1 Kf7 28.Kb1 Ke7 29.Nh3 Ke6 30.Kc2 Kf6 31.Kb2 Ke7 32.Ka2 Ke8 33.Nf2 Ke7 34.Kb1 Kf7 35.Ka1 Kg8 36.Nh3 Kh7 37.Ka2 Kg8 38.Kb1 Kf8 39.Kc2 Kg8 40.Nf2 Kg7
80/92 2:22:34 78,708,017k 9,201k +9.82 1.Kb2 Kd7 2.Nc1 Ke8 3.Kc2 Kd8 4.Kd1 Kc7 5.Ne2 Kd6 6.Kc1 Ke7 7.Ng1 Kf6 8.Kd1 Kg7 9.Ne2 Kh8 10.Kc1 Kg8 11.Ng1 Kg7 12.Kd1 Kf6 13.Kc2 Ke7 14.Kb1 Kf8 15.Nh3 Kg8 16.Kc2 Kh7 17.Nf2 Kg8 18.Nd1 Kf8 19.Kb1 Kf7 20.Kb2 Ke7 21.Nf2 Kf7 22.Ka2 Kg7 23.Nd1 Kf7 24.Kb1 Ke8 25.Nb2 Kd7 26.Ka1 Kd6 27.Nd1 Ke7 28.Kb2 Kf7 29.Kc1 Ke7 30.Nf2 Kf7 31.Nh3 Kg8 32.Kb1 Kf8 33.Ng1 Kg8 34.Ne2 Kg7 35.Ka2 Kh7 36.Ka1 Kh8 37.Kb2 Kh7 38.Ng1 Kh8 39.Kc2 Kh7 40.Ne2 Kh8 41.Kd1 Kg7 42.Kd2 Kf7 43.Kc2 Ke7 44.Ng1
81/89 2:30:15 82,967,240k 9,203k +9.82 1.Kb2 Kd7 2.Nc1 Ke8 3.Kc2 Kd8 4.Kd1 Kc7 5.Ne2 Kd6 6.Kc1 Ke7 7.Ng1 Kf6 8.Kb2 Kg7 9.Nh3 Kh7 10.Nf2 Kg8 11.Kb1 Kf7 12.Ka1 Kg8 13.Nd1 Kf7 14.Nb2 Ke7 15.Na4 Kd6 16.Kb1 Kc6 17.Kb2 Kd6 18.Kc2 Kc6 19.Kc1 Kd6 20.Kd2 Kc6 21.Nb2 Kd7 22.Kc1 Ke6 23.Kc2 Kd7 24.Nd1 Ke7 25.Kb2 Kd7 26.Kc1 Ke7 27.Nf2 Ke6 28.Nh3 Kf7 29.Kd2 Kg7 30.Kd1 Kh7 31.Kc1 Kg8 32.Kc2 Kh7 33.Ng1 Kh8 34.Kc1 Kh7 35.Kd1 Kg7 36.Ne2 Kh8 37.Kc1 Kg8 38.Kb1 Kg7 39.Ng1 Kg8 40.Nh3 Kf8 41.Kb2 Kg8 42.Nf2 Kf7

[lech]

For sure this is not very difficult to find for human but for chess computer there is, I think, no actually program which is able to find the way for winning this.

[d]3k4/8/7p/2p1p1pP/1pPpPpP1/1P1P1P2/N7/2K5 w - - 0 1
For this position I let Sting SF 8.8 one day with big evaluation but no plan

If Sting returns a big gain it probably means that sees the only one way to solution. Because it is a very long PV line Sting can have a problem to display it correctly.
As I earlier wrote this example is rather very technical only and Stockfish (maybe other engines too), thanks to the fact that Black has only King, skips the null-move reductions and can solve it.

For me it is an important position to motivate me to change (Sting) the way (code) to remember and return PV lines.

[Uri Blass]

Stockfish has a fail high at depth 82 and I stopped the search.

I think that it has to see the right line with that big score

[jhellis3]

[d]3k4/8/7p/2p1p1pP/1pPpPpP1/1P1P1P2/N7/2K5 w - - 0 1

The correct path forward for this puzzle is understanding that placing the knight on h4 creates two threats and black cannot cover both with just the king, so he is forced to capture the knight on h4, but this allows white to open the position and win the game.

The two threats of the knight on h4 are:

1) knight to f5, followed by Nxh6

2) knight to f5, followed by knight to d6 to b7 to Nxc5

If black defends the h pawn, the c5 pawn will fall. Thus black must take the knight on h4.

Getting the knight to h4 and the white king in position to recapture just takes many moves, which is why engines struggle with it.

[Michel]

The correct path forward for this puzzle is understanding that placing the knight on h4 creates two threats

In fact only one threat is sufficient to win: Nxh6. If black defends the h6 pawn then she is in zugzwang and it is sufficient to move the white king to capture the pawn.

[jhellis3]

In fact only one threat is sufficient to win

How do you propose to queen the h-pawn (without capturing other pawns first)?

[d]7k/8/8/2p1pNpP/1pPpPpP1/1P1P1P2/6K1/8 w - -

[Michel]

How do you propose to queen the h-pawn (without capturing other pawns first)?

It is totally clear that once the h-pawn has been captured the position is won since the black king has to stay in the square of the passed pawn and so the knight is free to gobble up the other pawns (why should it not be allowed to do this?).

[jhellis3]

The knight taking other pawns is creating more than one threat .

You said one threat was sufficient to win. But if you can win as white taking only the h-pawn, I will eat as many hats as you see fit.

IOW, white must take another pawn (have more than 1 threat), to win the game. This means my OP is correct, and your "correction" rather not so much.

[lech]

For sure this is not very difficult to find for human but for chess computer there is, I think, no actually program which is able to find the way for winning this.

[d]3k4/8/7p/2p1p1pP/1pPpPpP1/1P1P1P2/N7/2K5 w - - 0 1
For this position I let Sting SF 8.8 one day with big evaluation but no plan

The main problem for Sting was not in PV lines though (but it should be modified too).

Note PV info depth 55 after 4 minutes!

modified Sting SF 8.8; old dual core; 1024M Hash

Code: Select all

info depth 54 multipv 1 score cp 533 nodes 186132397 nps 1257787 time 147984 pv
c1d2 d8e8 d2e2 e8d8 e2f1 d8e8 f1g2 e8f8 g2h1 f8f7 h1g1 f7e7 g1f2 e7d8 f2e2 d8d7
a2c1 d7e7 e2d1 e7f7 c1e2 f7f8 d1c1 f8e7 c1b1 e7f7 b1b2 f7e6 b2a1 e6e7 a1b2 e7f7
e2g1 f7e6 b2b1 e6e7 b1c2 e7e8 g1e2 e8f7 c2d2 f7e6 e2g1 e6f7 d2e2 f7e6 e2f1 e6f7
g1e2 f7e7 f1e1 e7f7 e2c1 f7e7 e1d2 e7f8 c1e2 f8f7 d2c2 f7e6 c2b2 e6f7
info depth 55
info currmove c1d2 currmovenumber 1
info currmove c1b2 currmovenumber 2
info depth 55 multipv 1 score cp 545 lowerbound nodes 250612911 nps 1258671 time
 199109 pv c1b2
info currmove c1b2 currmovenumber 1
info depth 55 multipv 1 score cp 557 lowerbound nodes 251008029 nps 1258482 time
 199453 pv c1b2
info currmove c1b2 currmovenumber 1
info currmove c1d2 currmovenumber 2
info currmove c1c2 currmovenumber 3
info currmove c1b1 currmovenumber 4
info currmove c1d1 currmovenumber 5
info currmove a2c3 currmovenumber 6
info currmove a2b4 currmovenumber 7
info depth 55 multipv 1 score cp 557 nodes 320194423 nps 1269826 time 252156 pv
c1b2 d8e7 a2c1 e7d7 b2b1 d7e6 b1a2 e6e7 a2a1 e7f7 c1e2 f7f6 a1b2 f6e6 b2b1 e6e7
b1c2 e7f7 c2c1 f7e7 c1b1 e7f6 e2g1 f6e6 g1h3 e6f7 b1c1 f7f8 c1c2 f8e7 h3f2 e7d7
c2c1 d7e6 f2d1 e6d7 c1b1 d7e7 b1c2 e7d6 c2d2 d6d7 d1b2 d7c6 b2a4 c6d6 a4b6 d6e7
b6d5 e7e6 d5c7 e6d7 c7b5 d7c6 d2d1 c6d7 b5a3 d7e6 a3c2 e6d7 d1d2 d7e6 d2e1 e6d7
e1f2 d7e7 c2e1 e7d6 f2f1 d6c6 f1g2 c6d7 g2h3 d7e6 e1g2 e6d6 g2h4 g5h4 h3h4 d6e7
g4g5 e7f7 g5h6 f7g8 h4g4 g8f7 g4f5 f7g8 f5e5 g8h7 e5f4 h7h6 f4g4 h6h7 e4e5
info depth 56
info currmove c1b2 currmovenumber 1
info currmove c1d2 currmovenumber 2
info currmove c1c2 currmovenumber 3
info currmove c1b1 currmovenumber 4
info currmove c1d1 currmovenumber 5
info currmove a2c3 currmovenumber 6
info currmove a2b4 currmovenumber 7
info depth 56 multipv 1 score cp 545 upperbound nodes 357147771 nps 1274462 time
 280234 pv c1b2
info currmove c1b2 currmovenumber 1
info depth 56 multipv 1 score cp 569 lowerbound nodes 369326140 nps 1275945 time
 289453 pv c1b2
info currmove c1b2 currmovenumber 1
info currmove c1d2 currmovenumber 2
info currmove c1c2 currmovenumber 3
info currmove c1b1 currmovenumber 4
info currmove c1d1 currmovenumber 5
info currmove a2c3 currmovenumber 6
info currmove a2b4 currmovenumber 7
info depth 56 multipv 1 score cp 533 upperbound nodes 419046493 nps 1269418 time
 330109 pv c1b2
info currmove c1b2 currmovenumber 1
info depth 56 multipv 1 score cp 587 lowerbound nodes 431186977 nps 1267034 time
 340312 pv c1b2
info currmove c1b2 currmovenumber 1
info currmove c1d2 currmovenumber 2
info currmove c1c2 currmovenumber 3
info currmove c1d1 currmovenumber 4
info currmove c1b1 currmovenumber 5
info currmove a2c3 currmovenumber 6
info currmove a2b4 currmovenumber 7
info depth 56 multipv 1 score cp 573 nodes 460340617 nps 1256261 time 366437 pv
c1b2 d8e7 a2c1 e7d7 c1e2 d7c7 b2b1 c7d7 b1c2 d7d6 e2g1 d6d7 g1e2
info depth 57
info currmove c1b2 currmovenumber 1
info depth 57 multipv 1 score cp 585 lowerbound nodes 464542545 nps 1257008 time
 369562 pv c1b2
info currmove c1b2 currmovenumber 1
info currmove c1d2 currmovenumber 2
info currmove c1b1 currmovenumber 3
info currmove c1c2 currmovenumber 4
info currmove c1d1 currmovenumber 5
info currmove a2c3 currmovenumber 6
info currmove a2b4 currmovenumber 7
info depth 57 multipv 1 score cp 573 nodes 495080666 nps 1243043 time 398281 pv
c1b2 d8e7 a2c1 e7d7 c1e2 d7e7 e2c1
info depth 58
info currmove c1b2 currmovenumber 1
info depth 58 multipv 1 score cp 581 lowerbound nodes 501252274 nps 1244623 time
 402734 pv c1b2
info currmove c1b2 currmovenumber 1
info depth 58 multipv 1 score cp 589 lowerbound nodes 502748524 nps 1244621 time
 403937 pv c1b2
info currmove c1b2 currmovenumber 1
info depth 58 multipv 1 score cp 602 lowerbound nodes 535884996 nps 1250653 time
 428484 pv c1b2
info currmove c1b2 currmovenumber 1
info depth 58 multipv 1 score cp 620 lowerbound nodes 549758598 nps 1252163 time
 439047 pv c1b2
info currmove c1b2 currmovenumber 1
info depth 58 multipv 1 score cp 647 lowerbound nodes 635523540 nps 1260179 time
 504312 pv c1b2
info currmove c1b2 currmovenumber 1
info currmove c1d2 currmovenumber 2
info depth 58 multipv 1 score cp 688 lowerbound nodes 715875050 nps 1239275 time
 577656 pv c1d2
info currmove c1d2 currmovenumber 1
info currmove c1b2 currmovenumber 2
info currmove c1c2 currmovenumber 3
info currmove c1d1 currmovenumber 4
info currmove c1b1 currmovenumber 5
info currmove a2c3 currmovenumber 6
info currmove a2b4 currmovenumber 7
info depth 58 multipv 1 score cp 686 nodes 819564395 nps 1252080 time 654562 pv
c1d2 d8c7 d2d1 c7d7 a2c1 d7e7 d1e1 e7f8 e1e2 f8e7 e2d1 e7d8 d1d2 d8d7 d2d1
info depth 59
info currmove c1d2 currmovenumber 1
info depth 59 multipv 1 score cp 698 lowerbound nodes 830043525 nps 1253457 time
 662203 pv c1d2
info currmove c1d2 currmovenumber 1
info currmove c1b2 currmovenumber 2
info currmove c1c2 currmovenumber 3
info currmove c1b1 currmovenumber 4
info currmove c1d1 currmovenumber 5
info currmove a2c3 currmovenumber 6
info currmove a2b4 currmovenumber 7
info depth 59 multipv 1 score cp 698 nodes 980912589 nps 1233197 time 795422 pv
c1d2 d8c7 d2d1 c7d7 a2c1 d7e7 d1c2 e7e8 c2b1 e8f7 b1a2 f7e8 c1e2 e8f7 e2g1 f7e7
a2a1 e7f7 a1b1 f7f8 g1h3 f8e7 b1c2 e7f7 c2d2 f7e7 d2c1 e7f8 h3f2 f8f7 f2d1 f7e6
d1b2 e6d6 b2a4 d6c6 c1b2 c6d6 a4b6 d6e7 b6c8 e7d7 c8a7 d7c7 a7b5 c7c6 b5a3 c6d7
b2a2 d7e7 a3c2 e7f8 a2b1 f8g8 c2e1 g8h7 b1c2 h7h8 c2c1 h8g8 c1d1 g8f8 d1c2 f8e7
c2c1 e7d6 c1b1 d6c6 b1b2 c6d7 b2c1 d7d6
info depth 60
info currmove c1d2 currmovenumber 1
info depth 60 multipv 1 score cp 711 lowerbound nodes 982878472 nps 1232860 time
 797234 pv c1d2
info currmove c1d2 currmovenumber 1
info depth 60 multipv 1 score cp 723 lowerbound nodes 984862531 nps 1232618 time
 799000 pv c1d2
info currmove c1d2 currmovenumber 1
info depth 60 multipv 1 score cp 741 lowerbound nodes 994094077 nps 1233009 time
 806234 pv c1d2
info currmove c1d2 currmovenumber 1