Why do the engines pick this losing move?

[reflectionofpower]

I was reading this article https://en.chessbase.com/post/why-chess-players-blunder and curiously I set up the position they were talking about - Carlsen-Aronian, 6th Norway Blitz, 2018 (8)

[d]8/6p1/8/7p/3kP1p1/5P1P/2K2P2/8 w - - 0 1 and all the engines I have pick 52.hxg4?? which is the move Carlsen used and it's completely losing as you can't stop the passer. WHY are the chess engines doing this?

[Chessqueen]

I was reading this article https://en.chessbase.com/post/why-chess-players-blunder and curiously I set up the position they were talking about - Carlsen-Aronian, 6th Norway Blitz, 2018 (8)

[d]8/6p1/8/7p/3kP1p1/5P1P/2K2P2/8 w - - 0 1 and all the engines I have pick 52.hxg4?? which is the move Carlsen used and it's completely losing as you can't stop the passer. WHY are the chess engines doing this?

If you take an engine rated close to Carlsen and Aronian it will NOT pick the Wrong move. Here is Dumb 2.1 rated around 2850 like Carlsen.
NOTE: In my Opinion this Dumb engine can easy beat Carlsen

[pgn][Event "Computer chess game"]
[Date "2024.04.16"]
[Round "?"]
[White "Dumb-2.1"]
[Black "Dumb-2.1"]
[Result "1/2-1/2"]
[BlackElo "2850"]
[Time "08:36:28"]
[WhiteElo "2850"]
[TimeControl "60+1"]
[SetUp "1"]
[FEN "8/6p1/8/7p/3kP1p1/5P1P/2K2P2/8 w - - 0 1"]
[Termination "unterminated"]
[PlyCount "4"]
[WhiteType "program"]
[BlackType "program"]

1. fxg4 {(f3xg4 h5xg4 h3xg4 Kd4xe4 g4-g5 Ke4-f5 Kc2-d3 Kf5xg5 Kd3-e4 Kg5-g4
f2-f3+ Kg4-g3 Ke4-f5 Kg3xf3 Kf5-g6 Kf3-g2 Kg6xg7) 0.00/34 12} hxg4 {(h5xg4
h3xg4 Kd4xe4 g4-g5 Ke4-f5 Kc2-d3 Kf5xg5 Kd3-e4 Kg5-g4 f2-f3+ Kg4-g3 f3-f4
g7-g6 f4-f5 g6xf5+ Ke4xf5) 0.00/36 12} 2. hxg4 {(h3xg4 Kd4xe4 g4-g5 Ke4-f5
Kc2-d3 Kf5xg5 Kd3-e4 Kg5-g4 f2-f3+ Kg4-g3 Ke4-f5 Kg3xf3 Kf5-g6 Kf3-g2
Kg6xg7) 0.00/35 15} Kxe4 {(Kd4xe4 g4-g5 Ke4-f5 Kc2-d3 Kf5xg5 Kd3-e4 Kg5-g4
f2-f3+ Kg4-h4 Ke4-f5 Kh4-g3 Kf5-g6 Kg3xf3 Kg6xg7) 0.00/40 2} 3. g5 {(g4-g5
Ke4-f5 Kc2-d3 Kf5xg5 Kd3-e4 Kg5-g4 f2-f3+ Kg4-g3 Ke4-f5 Kg3xf3 Kf5-g6
Kf3-g2 Kg6xg7) 0.00/47 1} Kf5 {(Ke4-f5 Kc2-d3 Kf5xg5 Kd3-e4 Kg5-g4 f2-f3+
Kg4-g3 Ke4-f5 Kg3xf3 Kf5-g6 Kf3-g2 Kg6xg7) 0.00/48 2} 4. Kd3 {(Kc2-d3
Kf5xg5 Kd3-e4 Kg5-g4 f2-f3+ Kg4-g3 Ke4-f5 Kg3xf3 Kf5-g6 Kf3-g2 Kg6xg7)
0.00/52 1} Kxg5 {(Kf5xg5 Kd3-e4 Kg5-g4 f2-f3+ Kg4-g3 Ke4-f5 Kg3xf3 Kf5-g6
Kf3-g2 Kg6xg7) 0.00/51 2} 5. Ke4 {(Kd3-e4 Kg5-g4 f2-f3+ Kg4-g3 Ke4-f5
Kg3xf3 Kf5-g6 Kf3-g2 Kg6xg7) 0.00/52 1} Kg4 {(Kg5-g4) -0.05/54 2} 6. f3+
{(f2-f3+) -0.05/56 2} Kg3 {(Kg4-g3 Ke4-f5 Kg3xf3 Kf5-g6 Kf3-g2 Kg6xg7)
0.00/58 2} 7. Kf5 {(Ke4-f5 Kg3xf3 Kf5-g6 Kf3-g2 Kg6xg7) 0.00/75 2} Kxf3
{(Kg3xf3 Kf5-g6 Kf3-g2 Kg6xg7) 0.00/100 0} 8. Kg6 {(Kf5-g6 Kf3-g2 Kg6xg7)
0.00/100 0} Kg2 {(Kf3-g2 Kg6xg7) 0.00/100 0} 9. Kxg7 {(Kg6xg7) 0.00/100 0
Insufficient material} 1/2-1/2[/pgn]

[reflectionofpower]

Dumb 2.1 is still Dumb and so are the Smart engines. 52...hg4?? is losing instantly as it allows the passer as it allows the winning move h4. All my engines pick h4 after the howler 52...hg4??

[Jouni]

No they pick drawing fxg4!

[Uri Blass]

Dumb 2.1 is still Dumb and so are the Smart engines. 52...hg4?? is losing instantly as it allows the passer as it allows the winning move h4. All my engines pick h4 after the howler 52...hg4??

... means black move and after the correct 52.fxg4 that engines suggest the move 52...hxg4 that is not losing.

Engines do not play the losing move 52.hxg4 that allow for black 52...h4

[reflectionofpower]

I think it's the Lonnie Brain engine v.1.0 because I just woke up and thought I found a sleeper. My mistake and apologies.

[Ajedrecista]

Hello:

52.- hxg4?? is a disgrace, being gentle. Huntsman sees a checkmate in 15 moves after 52.- hxg4??, h4:

Code: Select all

FEN: 8/6p1/8/8/3kP1Pp/5P2/2K2P2/8 w - - 0 2

Huntsman:
Found 145 tablebases
classical evaluation enabled
[...]
 57/31	03:42	 835.204.012	3.750.287	-M15	e4e5 Kd4xe5 Kc2d3 h4h3 Kd3c4 h3h2 f3f4+ Ke5d6 Kc4d4 h2h1Q f4f5 Qh1f3 g4g5 Qf3f4+ Kd4d3 Kd6d5 Kd3e2 Qf4xf5 g5g6 Qf5e4+ Ke2f1 Kd5e5 f2f3 Qe4c2 Kf1g1 Ke5f4 Kg1f1 Kf4xf3 Kf1e1 Qc2c1+

There were 34,369,634 TB hits to the full set of 3, 4 and 5-man Syzygy EGTB.

Regards from Spain.

Ajedrecista.

[MOBMAT]

[d]8/6p1/8/7p/3kP1p1/5P1P/2K2P2/8 w - - 0 1

fxg4 and e5 both draw. Everything else loses....

FEN: 8/6p1/8/7p/3kP1p1/5P1P/2K2P2/8 w - - 0 1

Houdidit with 6 man Nalimov EGTB...

Code: Select all

 55/59	01:41	 533,557k	5,257k	-M11	1.Kd1 gxh3 2.Ke2 h2 3.e5 Kxe5 4.Ke3 h1Q 5.f4+ Kd5 6.f3 Kc4 7.f5 Qg1+ 8.Ke2 Kd4 9.f6 Qh2+ 10.Kf1 Ke3 11.fxg7
 55/59	01:41	 533,557k	5,257k	-M12	1.Kd2 gxh3 2.e5 Kxe5 3.Ke3 h2 4.f4+ Kd5 5.f3 h1Q 6.f5 Qg2 7.f6 gxf6
 55/59	01:41	 533,557k	5,257k	-M15	1.hxg4 h4 2.e5 Kxe5 3.Kd3 h3 4.Ke3 h2 5.f4+ Kd5 6.f5 h1Q 7.Kf4 Kd4 8.f6 Qh2+ 9.Kf5 Qxf2+
 55/59	01:41	 533,557k	5,257k	-M20	1.h4 gxf3 2.Kb3 Kxe4 3.Kc2 Kf4 4.Kd2 g6 5.Ke1 Kg4 6.Kd2 Kxh4
 55/59	01:41	 533,557k	5,257k	 0.00	1.e5 Kxe5 2.fxg4 hxg4
 55/59	01:41	 533,557k	5,257k	 0.00	1.fxg4 hxg4 2.hxg4

So, where is the win for white???

[Antihelion]

I was reading this article https://en.chessbase.com/post/why-chess-players-blunder and curiously I set up the position they were talking about - Carlsen-Aronian, 6th Norway Blitz, 2018 (8)

[d]8/6p1/8/7p/3kP1p1/5P1P/2K2P2/8 w - - 0 1 and all the engines I have pick 52.hxg4?? which is the move Carlsen used and it's completely losing as you can't stop the passer. WHY are the chess engines doing this?

This type of position is extremely easy for engine to solve. Do you have any strong engines? Stockfish, for example, solves the position on depth 1.

Code: Select all

info depth 1 seldepth 5 multipv 1 score cp 11 nodes 25 nps 833 hashfull 0 tbhits 0 time 30 pv f3g4 h5g4
info depth 2 seldepth 4 multipv 1 score cp 11 nodes 46 nps 1243 hashfull 0 tbhits 0 time 37 pv f3g4 h5g4
info depth 3 seldepth 4 multipv 1 score cp 11 nodes 75 nps 1829 hashfull 0 tbhits 0 time 41 pv f3g4 h5g4
info depth 4 seldepth 4 multipv 1 score cp 11 nodes 95 nps 2317 hashfull 0 tbhits 0 time 41 pv f3g4 h5g4
info depth 5 seldepth 6 multipv 1 score cp -15 nodes 613 nps 7961 hashfull 0 tbhits 0 time 77 pv e4e5 d4e5 f3g4 h5g4
info depth 6 seldepth 6 multipv 1 score cp -15 nodes 787 nps 10089 hashfull 0 tbhits 0 time 78 pv f3g4 h5g4 h3g4 d4e4 g4g5
info depth 7 seldepth 9 multipv 1 score cp -14 nodes 892 nps 11150 hashfull 0 tbhits 0 time 80 pv f3g4 h5g4 h3g4 d4e4 g4g5 e4f5 c2d2
info depth 8 seldepth 9 multipv 1 score cp -12 nodes 1400 nps 15730 hashfull 0 tbhits 0 time 89 pv f3g4 h5g4 h3g4 d4e4 g4g5 g7g6 c2d2
info depth 9 seldepth 10 multipv 1 score cp -1 nodes 2055 nps 20969 hashfull 0 tbhits 0 time 98 pv f3g4 h5g4 h3g4 d4e4 g4g5 e4f3 c2d3 f3f2
info depth 10 seldepth 11 multipv 1 score cp 0 nodes 3011 nps 26412 hashfull 0 tbhits 0 time 114 pv f3g4 h5g4 h3g4 d4e4 g4g5 e4f5 c2d2 f5g5 d2d3 g5h5

[Antihelion]

If you take an engine rated close to Carlsen and Aronian it will NOT pick the Wrong move.

That's not how chess programs work at all.