Best of Chess

[Peter Berger]

Regarding the bazillions of engine games you speak of, that will be an easy one provided the engines are within a +/- margin of 50 elo points. If you play Stockfish against an engine 300 elo less it will outsearch the lesser engine frequently and what you get is not [Strategic Insight] the thing you want but [Tactical Insight].

For human games, use (write) a tool that analysis PGN games first with scores inside, then treat them as comp-comp games.

Regarding computer games I don't think you have to be +that+ strict rating-wise (unless you start to pit engines against each other with a ridiculous difference of strength). My idea is to balance it with the number of moves requested where the engines basically disagree in principle/heavily. I tried with my Stockfish-Crafty games at long time control and looked at the games where both thought they were better for five consecutive moves. The resulting games look pretty reasonable for better "strategic insight".

This with the caveat that strategy and tactics is human terminology anyway, and we often don't know the exact reasons why the engines disagree.

When it is about human games, I am at a complete loss to understand your idea. Let's say you analyze a human game with some engine - you get a file with evals added ( I suppose pretty symmetrical ones) - and what to with the data then?? I don't see how you could ever spot some "strategic insight" this way.

[Rebel]

I already have a working tool for comp games, here is a sample.

[pgn][Event "Best of Chess"]
[Site "Evaluation Strategic Insight"]
[Date "2025.07.19"]
[White "CSTal-2.1-EAS"]
[Black "Igel-3.30"]
[Result "1-0"]
[PlyCount "74"]
[StrategicInsight "2155"]

{26=1000} {27=1155} 1. d4 {book} d5 {book} 2. Bg5 {book} f6 {book} 3. Bd2 {book} Nc6 {book} 4. e3 {book} e5 {book} 5. Bb5 {book} exd4 {book} 6. exd4 {book} Be6 {book} 7. Nc3 {+0.46/14 0.60s} Qd7 {+0.14/17 0.51s} 8. Nge2 {+0.27/14 0.34s} a6 {+0.28/18 0.48s} 9. Ba4 {+0.19/15 0.35s} Bf7 {+0.42/17 0.35s} 10. Nf4 {+0.29/14 0.34s} Nge7 {+0.35/17 0.45s} 11. O-O {+1.00/14 0.36s} b5 {+0.22/19 0.41s} 12. Bb3 {+0.84/14 0.33s} Nxd4 {+0.07/17 0.40s} 13. Re1 {+0.72/14 0.39s} O-O-O {+1.14/17 0.40s} 14. Be3 {+0.35/16 0.63s} Ndf5 {+1.10/17 0.42s} 15. a4 {+0.46/12 0.19s} b4 {+1.12/16 0.38s} 16. Qe2 {+1.22/13 0.32s} Qc6 {+0.18/17 0.36s} 17. Nb5 {+1.34/14 0.41s} Kb7 {+0.07/16 0.35s} 18. Nd3 {+1.98/13 0.26s} Nxe3 {+1.49/16 0.35s} 19. Nd4 {+1.07/14 0.31s} Qd6 {+0.57/17 0.33s} 20. Qxe3 {+0.96/13 0.19s} Ng6 {+0.83/17 0.34s} 21. Nf5 {+0.77/13 0.67s} Qb6 {+1.89/17 0.34s} 22. Qf3 {+0.40/13 0.48s} c5 {+1.34/17 0.35s} 23. a5 {+1.20/12 0.18s} Qc7 {+1.83/18 0.30s} 24. c4 {+0.11/14 0.28s} bxc3 {+0.90/16 0.30s} 25. bxc3 {+1.27/14 0.18s} c4 {+0.53/17 0.28s} 26. Reb1 {+2.00/14 0.26s} cxb3 {+1.38/16 0.27s} 27. Rxb3+ {+2.31/15 0.14s} Ka8 {+0.64/17 0.27s} 28. Rb6 {+1.93/15 0.19s} Qc8 {+0.95/18 0.26s} 29. Rab1 {+1.72/15 0.49s} Rd7 {+0.91/16 0.24s} 30. Qe2 {+1.72/13 0.35s} Be7 {+0.72/16 0.24s} 31. Nb2 {+2.40/12 0.27s} Ka7 {+1.91/14 0.23s} 32. Na4 {+3.97/13 0.14s} Rc7 {+3.21/13 0.21s} 33. Qe3 {+9.52/14 0.15s} Ka8 {-2.35/16 0.19s} 34. Nd6 {+10.10/17 0.21s} Ra7 {-7.67/13 0.19s} 35. R6b3 {+M29/17 0.16s} Bxd6 {-14.43/14 0.17s} 36. Nb6+ {+M21/26 0.051s} Kb7 {-15.86/17 0.17s} 37. Nxc8+ {+M19/30 0.067s} Kxc8 {-16.23/17 0.14s, White wins by adjudication} 1-0
[/pgn]

PGN tag : [StrategicInsight "2155"]

Note, it lists at the start the critical move or moves (labeled as Strategic Insight), move 26 and move 27 in the game rewarded with a bonus.

26=1000
27=1155

If we look at the moves in detail -

26. Reb1 {+2.00/14 0.26s} cxb3 {+1.38/16 0.27s}
27. Rxb3+ {+2.31/15 0.14s} Ka8 {+0.64/17 0.27s}
28. Rb6 {+1.93/15 0.19s} Qc8 {+0.95/18 0.26s}
29. Rab1 {+1.72/15 0.49s} Rd7 {+0.91/16 0.24s}
30. Qe2 {+1.72/13 0.35s} Be7 {+0.72/16 0.24s}
31. Nb2 {+2.40/12 0.27s} Ka7 {+1.91/14 0.23s}
32. Na4 {+3.97/13 0.14s} Rc7 {+3.21/13 0.21s} ???
33. Qe3 {+9.52/14 0.15s} Ka8 {-2.35/16 0.19s} finally...

We see that black has no idea what is coming to him until 6 moves later.

That's the idea, extract those games into a new PGN.

[Peter Berger]

I am still not quite sure, that I have understood completely what you can already do and what you can't. I think I understand how this works with a PGN that has evals for both sides though.

Let's make the challenge one natural level more difficult. The following game is most clearly an excellent example of superior strategical insight. We can see the evals of Lc0 v0.32.0 in the PGN, we can't see the Crafty ones. But I can tell you about them - Crafty thought that it was better for several consecutive moves, and it even offered a draw as late as move 28 (because of several consecutive moves with 0.00 eval) - it resigned only a very few moves later.
We can rule out purely tactical reasons here as my notebook doesn't have a GPU - also I could see lc0s mainlines live - there were no tactical reasons for its initial assessment.

What would be your current idea to show that we see an example of superior strategic insight here?

Peter

[Event "Lang 120min+10sek"]
[Site "Berlin"]
[Date "2025.10.11"]
[Round "?"]
[White "Lc0 v0.32.0"]
[Black "Crafty 25.6"]
[Result "1-0"]
[ECO "E04"]
[PlyCount "71"]
[TimeControl "7200+10"]

{4096MB, DESKTOP-8OCGGEO} 1. d4 {[%eval 9,6] [%wdl 201,642,157] [%emt 0:07:20]} Nf6 {[%emt 0:00:07]} 2. c4 {[%eval 14,6] [%wdl 214,636,150] [%emt 0:01:48]} e6 {[%emt 0:00:06] (c6)} 3. g3 {[%eval 19,7] [%wdl 222,643,135] [%emt 0:07:01]} d5 {[%emt 0:00:07] (Lb4+)} 4. Bg2 {[%eval 19,8] [%wdl 213,661,126] [%emt 0:04:40]} dxc4 {[%emt 0:00:07] (Lb4+)} 5. Nf3 {[%eval 22,10] [%wdl 218,667,115] [%emt 0:01:36]} Nc6 {[%emt 0:00:06] (c5)} 6. Qa4 {[%eval 43,12] [%wdl 298,600,102] [%emt 0:01:44]} Bb4+ {[%emt 0:00:07]} 7. Bd2 {[%eval 35,16] [%wdl 285,589,126] [%emt 0:01:27]} Bd6 {[%emt 0:00:06] (Sd5)} 8. Na3 {[%eval 40,9] [%wdl 291,601,108] [%emt 0:04:05]} Bxa3 {[%emt 0:00:07] (Se4)} 9. Qxa3 {[%eval 49,8] [%wdl 327,566,107] [%emt 0:01:26]} Ne4 {[%emt 0:00:07] (Dd6)} 10. Bf4 {[%eval 41,8] [%wdl 300,586,114] [%emt 0:02:56]} f6 {[%emt 0:04:46] (Sd6)} 11. O-O {[%eval 63,8] [%wdl 374,535,91] [%emt 0:02:36]} Qe7 {[%emt 0:06:07] (Sd6)} 12. Qa4 {[%eval 56,10] [%wdl 355,542,103] [%emt 0:00:13]} c3 {[%emt 0:02:48]} 13. Qc2 {[%eval 50,11] [%wdl 327,573,100] [%emt 0:04:20]} cxb2 {[%emt 0:02:42]} 14. Qxb2 {[%eval 53,12] [%wdl 330,580,90] [%emt 0:03:03]} Nd6 {[%emt 0:00:31] (0-0)} 15. e4 {[%eval 57,9] [%wdl 346,569,85] [%emt 0:02:00]} O-O {[%emt 0:03:18]} 16. e5 {[%eval 45,11] [%wdl 318,567,115] [%emt 0:01:46]} Nc4 {[%emt 0:01:54] (fxe5)} 17. exf6 {[%eval 84,8] [%wdl 446,483,71] [%emt 0:01:42]} gxf6 {[%emt 0:00:39]} 18. Qc1 {[%eval 73,8] [%wdl 415,490,95] [%emt 0:03:44]} Nb6 {[%emt 0:02:28]} 19. Re1 {[%eval 69,9] [%wdl 406,485,109] [%emt 0:00:41]} Rd8 {[%emt 0:05:38] (Sd5)} 20. g4 {[%eval 82,11] [%wdl 451,434,115] [%emt 0:02:54]} Nd5 {[%emt 0:06:36] (Sxd4)} 21. Bg3 {[%eval 119,8] [%wdl 560,380,60] [%emt 0:01:36]} a5 {[%emt 0:02:14]} 22. g5 {[%eval 115,9] [%wdl 543,380,77] [%emt 0:01:36]} fxg5 {[%emt 0:02:21] (f5)} 23. Nxg5 {[%eval 115,10] [%wdl 544,377,79] [%emt 0:02:08]} Nxd4 {[%emt 0:03:07]} 24. Be5 {[%eval 112,14] [%wdl 533,381,86] [%emt 0:00:45]} h6 {[%emt 0:04:20] (Sf5)} 25. Ne4 {[%eval 117,10] [%wdl 552,381,67] [%emt 0:02:32]} Nc6 {[%emt 0:04:51]} 26. Bg3 {[%eval 118,11] [%wdl 557,383,60] [%emt 0:01:02]} Qg7 {[%emt 0:04:15] (Df8)} 27. Kh1 {[%eval 184,9] [%wdl 724,239,37] [%emt 0:03:10]} Kh8 {[%emt 0:01:25]} 28. Rg1 {[%eval 193,10] [%wdl 740,224,36] [%emt 0:01:42]} Rg8 {[%emt 0:01:24] R} 29. Qd2 {[%eval 201,10] [%wdl 761,208,31] [%emt 0:01:37]} h5 {[%emt 0:08:39] (Df8)} 30. Rae1 {[%eval 684,6] [%wdl 967,26,7] [%emt 0:05:16]} Rb8 {[%emt 0:02:39] (Ld7)} 31. h3 {[%eval 1008,6] [%wdl 978,18,4] [%emt 0:04:39]} b6 {[%emt 0:02:23] (Ld7)} 32. Rc1 {[%eval 681,6] [%wdl 966,28,6] [%emt 0:02:08]} h4 {[%emt 0:01:22]} 33. Bh2 {[%eval 822,7] [%wdl 973,22,5] [%emt 0:01:02]} Nde7 {[%emt 0:01:16] (Lb7)} 34. Bf3 {[%eval 1160,7] [%wdl 983,11,6] [%emt 0:02:18]} Bb7 {[%emt 0:04:04] (Df8)} 35. Rxg7 {[%eval 4393,6] [%wdl 996,3,1] [%emt 0:01:53]} Rxg7 {[%emt 0:00:35]} 36. Qh6+ {[%eval 3312,6] [%wdl 994,6,0] [%emt 0:03:42]} 1-0

[Ajedrecista]

Hello:

I am still not quite sure, that I have understood completely what you can already do and what you can't. I think I understand how this works with a PGN that has evals for both sides though.

Let's make the challenge one natural level more difficult. The following game is most clearly an excellent example of superior strategical insight. We can see the evals of Lc0 v0.32.0 in the PGN, we can't see the Crafty ones. But I can tell you about them - Crafty thought that it was better for several consecutive moves, and it even offered a draw as late as move 28 (because of several consecutive moves with 0.00 eval) - it resigned only a very few moves later.
We can rule out purely tactical reasons here as my notebook doesn't have a GPU - also I could see lc0s mainlines live - there were no tactical reasons for its initial assessment.

What would be your current idea to show that we see an example of superior strategic insight here?

Peter

[...]

Just enclosing the game between PGN tags to know what is going on:

[pgn][Event "Lang 120min+10sek"]
[Site "Berlin"]
[Date "2025.10.11"]
[Round "?"]
[White "Lc0 v0.32.0"]
[Black "Crafty 25.6"]
[Result "1-0"]
[ECO "E04"]
[PlyCount "71"]
[TimeControl "7200+10"]

{4096MB, DESKTOP-8OCGGEO} 1. d4 {[%eval 9,6] [%wdl 201,642,157] [%emt 0:07:20]} Nf6 {[%emt 0:00:07]} 2. c4 {[%eval 14,6] [%wdl 214,636,150] [%emt 0:01:48]} e6 {[%emt 0:00:06] (c6)} 3. g3 {[%eval 19,7] [%wdl 222,643,135] [%emt 0:07:01]} d5 {[%emt 0:00:07] (Lb4+)} 4. Bg2 {[%eval 19,8] [%wdl 213,661,126] [%emt 0:04:40]} dxc4 {[%emt 0:00:07] (Lb4+)} 5. Nf3 {[%eval 22,10] [%wdl 218,667,115] [%emt 0:01:36]} Nc6 {[%emt 0:00:06] (c5)} 6. Qa4 {[%eval 43,12] [%wdl 298,600,102] [%emt 0:01:44]} Bb4+ {[%emt 0:00:07]} 7. Bd2 {[%eval 35,16] [%wdl 285,589,126] [%emt 0:01:27]} Bd6 {[%emt 0:00:06] (Sd5)} 8. Na3 {[%eval 40,9] [%wdl 291,601,108] [%emt 0:04:05]} Bxa3 {[%emt 0:00:07] (Se4)} 9. Qxa3 {[%eval 49,8] [%wdl 327,566,107] [%emt 0:01:26]} Ne4 {[%emt 0:00:07] (Dd6)} 10. Bf4 {[%eval 41,8] [%wdl 300,586,114] [%emt 0:02:56]} f6 {[%emt 0:04:46] (Sd6)} 11. O-O {[%eval 63,8] [%wdl 374,535,91] [%emt 0:02:36]} Qe7 {[%emt 0:06:07] (Sd6)} 12. Qa4 {[%eval 56,10] [%wdl 355,542,103] [%emt 0:00:13]} c3 {[%emt 0:02:48]} 13. Qc2 {[%eval 50,11] [%wdl 327,573,100] [%emt 0:04:20]} cxb2 {[%emt 0:02:42]} 14. Qxb2 {[%eval 53,12] [%wdl 330,580,90] [%emt 0:03:03]} Nd6 {[%emt 0:00:31] (0-0)} 15. e4 {[%eval 57,9] [%wdl 346,569,85] [%emt 0:02:00]} O-O {[%emt 0:03:18]} 16. e5 {[%eval 45,11] [%wdl 318,567,115] [%emt 0:01:46]} Nc4 {[%emt 0:01:54] (fxe5)} 17. exf6 {[%eval 84,8] [%wdl 446,483,71] [%emt 0:01:42]} gxf6 {[%emt 0:00:39]} 18. Qc1 {[%eval 73,8] [%wdl 415,490,95] [%emt 0:03:44]} Nb6 {[%emt 0:02:28]} 19. Re1 {[%eval 69,9] [%wdl 406,485,109] [%emt 0:00:41]} Rd8 {[%emt 0:05:38] (Sd5)} 20. g4 {[%eval 82,11] [%wdl 451,434,115] [%emt 0:02:54]} Nd5 {[%emt 0:06:36] (Sxd4)} 21. Bg3 {[%eval 119,8] [%wdl 560,380,60] [%emt 0:01:36]} a5 {[%emt 0:02:14]} 22. g5 {[%eval 115,9] [%wdl 543,380,77] [%emt 0:01:36]} fxg5 {[%emt 0:02:21] (f5)} 23. Nxg5 {[%eval 115,10] [%wdl 544,377,79] [%emt 0:02:08]} Nxd4 {[%emt 0:03:07]} 24. Be5 {[%eval 112,14] [%wdl 533,381,86] [%emt 0:00:45]} h6 {[%emt 0:04:20] (Sf5)} 25. Ne4 {[%eval 117,10] [%wdl 552,381,67] [%emt 0:02:32]} Nc6 {[%emt 0:04:51]} 26. Bg3 {[%eval 118,11] [%wdl 557,383,60] [%emt 0:01:02]} Qg7 {[%emt 0:04:15] (Df8)} 27. Kh1 {[%eval 184,9] [%wdl 724,239,37] [%emt 0:03:10]} Kh8 {[%emt 0:01:25]} 28. Rg1 {[%eval 193,10] [%wdl 740,224,36] [%emt 0:01:42]} Rg8 {[%emt 0:01:24] R} 29. Qd2 {[%eval 201,10] [%wdl 761,208,31] [%emt 0:01:37]} h5 {[%emt 0:08:39] (Df8)} 30. Rae1 {[%eval 684,6] [%wdl 967,26,7] [%emt 0:05:16]} Rb8 {[%emt 0:02:39] (Ld7)} 31. h3 {[%eval 1008,6] [%wdl 978,18,4] [%emt 0:04:39]} b6 {[%emt 0:02:23] (Ld7)} 32. Rc1 {[%eval 681,6] [%wdl 966,28,6] [%emt 0:02:08]} h4 {[%emt 0:01:22]} 33. Bh2 {[%eval 822,7] [%wdl 973,22,5] [%emt 0:01:02]} Nde7 {[%emt 0:01:16] (Lb7)} 34. Bf3 {[%eval 1160,7] [%wdl 983,11,6] [%emt 0:02:18]} Bb7 {[%emt 0:04:04] (Df8)} 35. Rxg7 {[%eval 4393,6] [%wdl 996,3,1] [%emt 0:01:53]} Rxg7 {[%emt 0:00:35]} 36. Qh6+ {[%eval 3312,6] [%wdl 994,6,0] [%emt 0:03:42]} 1-0[/pgn]

Black king safety is horrible. I expected 34.- ..., Qf8 after 34.- Bf3, keeping an eye on h6 square for the mating threat Qh6# in Epaulette-like style (not 34.- ..., Qf7; 35.- Qh6+, Qh7; 36.- Qxh7+, Kxh7; 37.- Nf6+, forking the king and rook) and having a hard time defending, surely losing. Therefore, 34.- ..., Bb7 surprised my fairly low level experience, thinking it was directly throwing in the towel and making things easier for white, gifting a queen for a rook; only to learn later that SF thinks that this position is clearly lost for black regardless of what plays (34.- ..., Bb7; 34.- ..., Ng6; 34.- ..., Qf8; etc.), with evals over +7.

I was not aware of the long TC of the game.

Regards from Spain.

Ajedrecista.

[Peter Berger]

I am not used to people really looking at the games I post anymore.

I initially thought I should explain 34. ...Bb7 myself - you found out for yourself that it didn't matter though. Crafty is not that weak, at this point in the game it was well aware of the fact that it is toast.

There are two amazing moves in this games IMHO: 20. g4, that shows amazing strategic insight for miserable me, I also really liked 31. h3 ( though at this point in the game there are alternatives of course).

Peter