Like most correspondence sites, the International Correspondence Chess Federation allows engine use. Players at the top of the correspondence heap don’t just play whatever their engine’s top choice happens to be. Their play is a combination of engines guided by humans.
How do you get to the top in a world full of engine users? American Correspondence GM Jon Edwards, who won the 2020-2022 World Correspondence title, said he did it by emulating former over the board World Champion Tigran Petrosian who was known for his super-boring defensive play. In his games, Edwards tries to create a small advantage then keep milking it and, hopefully, making it grow.
You can read the story about how he won the title and play over one of his annitated games in the Chess Life article HERE.
Still, the majority of correspondence games end in a draw because it is nearly impossible to beat an engine. Within the last year I wanted to test an engine and even I was able to draw two games against a player who is an IM in both OTB and correspondence play. Out of the 136 games played in the 32nd World Correspondence Championship, 119 were draws (87.5%). Even then a few of the decisive games were due to human, not engine, errors. In spite of all the draws, Edwards doesn’t think correspondence chess is dead. Some players still enjoy the challenge of trying to outwit an engine I guess.
ChessBase had a 2022 interview with Edwards in which he revealed his secrets of how he won the World Championship. Here is a brief synopsis of how he did it.
Hardware makes a difference...the better the hardware, the better the engine output. Edwards said you can run analysis on chess.com or lichess, or your own specialized equipment if you’re computer savvy enough.
One thing he made clear was not to use the engine’s default parameters. The number of threads and the hash table size need to be as large as your computer can handle. The same goes for endgame tablebases...use the largest your computer can handle. Endgame tablebases are important because sometimes they will help the engine make a better middlegame evaluation.
Edwards advised that you spend more time on games where engine evaluations are in your favor. In those games search for dozens (yes, dozens!) of lines to see if the evaluation improves. And, don’t waste any time looking at positions that are 0.00.
Of course, there is more. For example having a large database of top level correspondence games and conducting a search for opening novelties is an important factor.
It’s clear that if you wish to excel in correspondence chess these days you need not only good hardware and software, but a lot of patience and you need to put in a lot of time.
My laptop is pretty old and it’s held together with a couple of large clamps and duct tape, but at least it has a SSD. I ran a 4-minute tournament and invited my top 8 engines to play and was shocked by the result! Berserk won.
Here is one of the more interesting games. The first thing I noticed was that a full tactical analysis with PlentyChess returned an Accuracy Rating of 100% for white and 70% for black. Analysis with Berserk was a different story. White still scored 100%, but black’s dropped to 53%! Apparently Berserk found more improvements for black than PlentyChess did.
[Event "Blitz G/4 min"]
[Site ""]
[Date "2025.??.??"]
[Round "?"]
[White "Dragon by Komodo"]
[Black "Deep Fritz 14"]
[Result "1-0"]
[ECO "C65"]
[Annotator "Multiple engines"]
[PlyCount "53"]
[EventDate "2025.11.17"]
[EventType "tourn"]
[SourceTitle "Fritz Engine Tournament"]
[Source "Doe"]
{C65: Ruy Lopez: Berlin Defense (3...Nf6), unusual lines} 1. e4 e5 2. Nf3 Nc6
3. Bb5 Nf6 {The Berlin Defense has long had a reputation for being solid and
drawish. It was played in the late 1800s and early 1900s by Emanuel Lasker and
others. In modern times it was a favorite of Arthur Bisguier and it has been
used by Vladimir Kramnik, Alexei Shirov, Veselin Topalov, Hikaru Nakamura,
Gary Kasparov, Magnus Carlsen and Viswanathan Anand...all players with agressive styles.} 4. O-O {This and 4.d3 are the two main choices.} Bd6 {
Usual are 4...Nxe4 and 4...Bc5. The text is highly unusual and is inferior as
it greatly hinders black's development. To be fair I let Deep Fritz cogitate
on the position after 4.O-O and it actually liked 4...Bd6 fir a bit, but then
settled on the better 4...Be7 or 4...Bc5} 5. d4 Qe7 6. Re1 O-O 7. c3 (7. Bg5 h6
8. Bh4 a6 9. Ba4 b5 10. Bb3 Na5 11. Bg3 Nc6 12. Bd5 {White stands very well.
Zilka,S (2602)-Taran,S (1932) Braila ROU 2022}) 7... Re8 8. Bg5 a6 9. Bd3 h6
10. Bh4 Rb8 11. Nbd2 {[%mdl 32]} b5 12. Nf1 Qd8 13. Ne3 {Strategically white
has a won position thanks to black's atrocious 4th move, but black's next move
is a serious blunder, especially coming from an engine. Retreating the B to e7
was the best try. However, again to be fair, Deep Fritz was allowed to examine
this position for a few minutes, but still lked 13...g5 giving white only
about a half P advantage. Stockfish knows better. White's advantage is almost
2.5 Ps.} g5 14. Nxg5 {[%mdl 512] Deep Fritz still refuses to change its
evaluation of the position. Stockfish thinks white's advantage is 3-plus Ps.}
hxg5 15. Bxg5 Be7 {After this Deep Fritz finally realizes that white os better.
..by a P and a half.} 16. Nf5 Nh7 17. Bh6 exd4 18. Qg4+ Bg5 19. f4 dxc3 20.
bxc3 d6 21. Qg3 Kh8 22. Bxg5 Nxg5 23. fxg5 Ne5 (23... Re6 24. e5 Nxe5 25. Qh4+
Kg8 26. Re3 {This R lift brings about a quick decision.} Rg6 27. Nh6+ Kf8 28.
Nxf7 Kxf7 29. Qh7+ Ke6 30. Qxg6+ Kd7 31. Rxe5 dxe5 32. Bf5+ Ke7 33. Qf6+ Ke8
34. Bg6+ Kd7 35. Rd1#) 24. Qh4+ Kg8 25. Re3 Ng6 26. Qh6 Bxf5 27. exf5 {Black
resigned.} (27. exf5 Rxe3 28. fxg6 Qe8 29. Qh7+ Kf8 30. Rf1 {is hopeless for
black.}) 1-0
No comments:
Post a Comment