Best of Chess

[Rebel]

Best of Chess extracts the most spectacular games from a PGN database. Each game is evaluated on 3 features.

. King Attack bonus.
. Material Sacrifice bonus.
. Length of the game, the less number of moves a game is won the higher the bonus.

If a game exceeds the total bonus of 20.000 points the game is stored in a new PGN file called Best-of-Chess.pgn.

.....

https://rebel7775.wixsite.com/rebel/about-1

Enjoy the games, here is one.

[pgn][Event "Best of Chess"]
[Site "King-Attacks, Sacrifices, Short Games"]
[Date "2025.08.14"]
[Round "?"]
[White "Igel-3.30"]
[Black "Rebel-Extreme-1.1"]
[Result "0-1"]
[PlyCount "41"]
[King "11712"]
[Short "6600"]
[Sac "11000"]
[Total "29312"]

1. e4 e5 2. Nf3 Nc6 3. Bc4 Nf6 4. Ng5 Bc5 5. Nxf7 Bxf2+ 6. Kf1 Qe7 7. Nxh8
d5 8. exd5 Nd4 9. Kxf2 Bg4 10. Qf1 Ne4+ 11. Ke3 Qg5+ 12. Kxe4 Bf5+ 13. Kxe5
Qf6+ 14. Kf4 Qh4+ 15. Ke3 Nxc2+ 16. Ke2 Qxc4+ 17. Kf3 Bd3 18. Qf2 g5 19.
Ng6 hxg6 20. Qh4 Qxh4 21. g4 0-1[/pgn]

[pohl4711]

I tried it (as usual) on my UHO-Top15 Ratinglist gamebase, which always contains 120.000 games. Only 42 games were found using the tool with the default margins. This is a hit-ratio of only 0.035%...
So, IMHO, it would be helpful to publish at least 3 sets of margin-numbers: For human games, engine games, very high level engine games.
Or make the tool itself able to distinguish between human games and enginegames and choose a set of margin-numbers for each case automatically. Like my EAS-tool, making the margins for short-wins automatically each time from the input-database (from the average length of all won games (human games are shorter than adjucated enginegames and these are shorter than non-adjucated enginegames, so distinguish these games is easy)).

[Rebel]

The tool is called Best-of-Chess, quite pretentious, so yes, the stake is very high to qualify a PGN output.

Try the Rebel-Extreme PGN included in the download, you already get 377 games.

Try your archive files with the likes of the Patricia's, the Rebel's and Cstal and I estimate you will get 1500-2000 games.

[Rebel]

Code: Select all

CCRL-Blitz gives : 1562 games.
CCRL-40/40       :  944 games.

https://rebel7775.wixsite.com/rebel/downloads

[Peter Berger]

Best of Chess extracts the most spectacular games from a PGN database. Each game is evaluated on 3 features.

I am not too happy with the current design of the feature "beauty" or "best". It is probably heavily influenced by Stefan Pohl's design of EAS.

I think you could use the advantages of your implementation to better use. I'd personally suspect beauty in cases where the two programs heavily disagree on evaluation for several moves, so that one engine shows (and proves) something that comes as a complete surprise to its opponent.

I have checked this idea manually and heuristically (with of course very/too limitted data) on my own computers - and spotted some truely amazing Stockfish games - where the reason might be strategic or tactical - the games amazing nevertheless.

Maybe this could be an interesting idea to pursue.

Peter

[Rebel]

Best of Chess extracts the most spectacular games from a PGN database. Each game is evaluated on 3 features.

I am not too happy with the current design of the feature "beauty" or "best". It is probably heavily influenced by Stefan Pohl's design of EAS.

I think you could use the advantages of your implementation to better use. I'd personally suspect beauty in cases where the two programs heavily disagree on evaluation for several moves, so that one engine shows (and proves) something that comes as a complete surprise to its opponent.

I have checked this idea manually and heuristically (with of course very/too limitted data) on my own computers - and spotted some truely amazing Stockfish games - where the reason might be strategic or tactical - the games amazing nevertheless.

Maybe this could be an interesting idea to pursue.

Peter

You make an important point.

I have often seen the style engines of Cstal and Rebel (and likely Patricia as well) seeing victory 3-4 moves earlier than equal or stronger opponents. Rebel giving a score of 5.xx or more while the opponent is still positive about itself happens frequently. How would you label that as a PGN tag, [Strategic-Insight] ?

The other side of the medal (which I did not research yet), style engines must be often wrong as well, they easily sac material (a pawn is nothing) and being wrong, cases like that should be taken into consideration in case of a new tool.

[Chacal X]

Best of Chess extracts the most spectacular games from a PGN database. Each game is evaluated on 3 features.

I am not too happy with the current design of the feature "beauty" or "best". It is probably heavily influenced by Stefan Pohl's design of EAS.

I think you could use the advantages of your implementation to better use. I'd personally suspect beauty in cases where the two programs heavily disagree on evaluation for several moves, so that one engine shows (and proves) something that comes as a complete surprise to its opponent.

I have checked this idea manually and heuristically (with of course very/too limitted data) on my own computers - and spotted some truely amazing Stockfish games - where the reason might be strategic or tactical - the games amazing nevertheless.

Maybe this could be an interesting idea to pursue.

Peter

You make an important point.

I have often seen the style engines of Cstal and Rebel (and likely Patricia as well) seeing victory 3-4 moves earlier than equal or stronger opponents. Rebel giving a score of 5.xx or more while the opponent is still positive about itself happens frequently. How would you label that as a PGN tag, [Strategic-Insight] ?

The other side of the medal (which I did not research yet), style engines must be often wrong as well, they easily sac material (a pawn is nothing) and being wrong, cases like that should be taken into consideration in case of a new tool.

Good evening,
In which folder should I put this .pgn game file to import into my book? I've already put it in almost every folder and none of them worked.
If you could guide me, I'd be very grateful!

[Rebel]

Good evening,
In which folder should I put this .pgn game file to import into my book? I've already put it in almost every folder and none of them worked.
If you could guide me, I'd be very grateful!

I don't understand your question, what do you mean with this .pgn game file ?

[Peter Berger]

I have often seen the style engines of Cstal and Rebel (and likely Patricia as well) seeing victory 3-4 moves earlier than equal or stronger opponents. Rebel giving a score of 5.xx or more while the opponent is still positive about itself happens frequently. How would you label that as a PGN tag, [Strategic-Insight] ?

The other side of the medal (which I did not research yet), style engines must be often wrong as well, they easily sac material (a pawn is nothing) and being wrong, cases like that should be taken into consideration in case of a new tool.

[Strategic-Insight] makes sense to me.

I am not completely sure if you understand what I am really up to though, so please forgive me for elaborating a Little further.

I agree by design there have to be several games where a "style engine" sacs material to no avail, the other one just takes it - and slowly converts. These games may provide interesting data for its developper, but they are probably not too interesting to watch for a spectator.
In a similar way +most+ games where one engine gets/finds something "random", the other one would too, just a tiny bit later, tend to not be too exciting to watch, at least for chessplayers way weaker than Magnus Carlsen.

Beauty and excitement probably need some element of surprise IMHO. When one engine realizes something important like 5 moves earlier than its appropiately strong opponent, things tend to get fascinating to watch. Sacs like Rebel's or Patricia's are one way to reach this goal, but there are others. I have spotted some truely amazing zugzwang and "positional" games by Stockfish on my computers, that I enjoyed watching very much. So, a future version of your tool might provide a way to find games with different kinds of beauty, too.

Peter

[Chacal X]

I have often seen the style engines of Cstal and Rebel (and likely Patricia as well) seeing victory 3-4 moves earlier than equal or stronger opponents. Rebel giving a score of 5.xx or more while the opponent is still positive about itself happens frequently. How would you label that as a PGN tag, [Strategic-Insight] ?

The other side of the medal (which I did not research yet), style engines must be often wrong as well, they easily sac material (a pawn is nothing) and being wrong, cases like that should be taken into consideration in case of a new tool.

[Strategic-Insight] makes sense to me.

I am not completely sure if you understand what I am really up to though, so please forgive me for elaborating a Little further.

I agree by design there have to be several games where a "style engine" sacs material to no avail, the other one just takes it - and slowly converts. These games may provide interesting data for its developper, but they are probably not too interesting to watch for a spectator.
In a similar way +most+ games where one engine gets/finds something "random", the other one would too, just a tiny bit later, tend to not be too exciting to watch, at least for chessplayers way weaker than Magnus Carlsen.

Beauty and excitement probably need some element of surprise IMHO. When one engine realizes something important like 5 moves earlier than its appropiately strong opponent, things tend to get fascinating to watch. Sacs like Rebel's or Patricia's are one way to reach this goal, but there are others. I have spotted some truely amazing zugzwang and "positional" games by Stockfish on my computers, that I enjoyed watching very much. So, a future version of your tool might provide a way to find games with different kinds of beauty, too.

Peter

Good morning Peter,
What I meant is that I don't know how to download games that are compiled in PGN into my book. When I try to import these games into the book, it can't find them. I want to know where to download these games that come in PGN and how to import them into my book.