Correspondence ICCF Champion Vs Top Engine ?

[Chessqueen]

I always wondered this if the current ICCF champion 2017-2019 Russia Andrey Leonidovich Kochemasov, can beat the top chess engine, but here are some good point of views.

Uri Blass wrote: ↑Mon May 11, 2020 2:28 am
I doubt if a strong correspondence player+top engine can score heavily against the same engine unassisted.

Maybe it was the case some years ago but after engines became stronger I doubt if today humans can find a way to beat top engines at correspondence time control in 30% of the cases(what you need to get 65%)

I am sure strong correspondence player can avoid losing but I am not sure if they can do more than it against unassisted stockfish or lc0 with no opening book.

It may be interesting to see what is the percentage of draws between lc0 and stockfish at time control of one hour per move when both use no book and a good hardware like hardware in the last TCEC final(stockfish with many cores is not deterministic so the games are not going to be the same).

Maybe I am wrong but I guess we will get at least 90 draws out of 100 games.


lkaufman Wrote:
I suppose that to a great extent the result of a match between a top engine and a correspondence player using that engine depends on whether the engine has an opening book and if so how good it is. Maybe my 65% figure is realistic only if no opening book (or a minimal or bad one).
Komodo rules!

[Cornfed]

I always wondered this if the current ICCF champion 2017-2019 Russia Andrey Leonidovich Kochemasov, can beat the top chess engine, but here are some good point of views.

Uri Blass wrote: ↑Mon May 11, 2020 2:28 am
I doubt if a strong correspondence player+top engine can score heavily against the same engine unassisted.

Maybe it was the case some years ago but after engines became stronger I doubt if today humans can find a way to beat top engines at correspondence time control in 30% of the cases(what you need to get 65%)

I am sure strong correspondence player can avoid losing but I am not sure if they can do more than it against unassisted stockfish or lc0 with no opening book.

It may be interesting to see what is the percentage of draws between lc0 and stockfish at time control of one hour per move when both use no book and a good hardware like hardware in the last TCEC final(stockfish with many cores is not deterministic so the games are not going to be the same).

Maybe I am wrong but I guess we will get at least 90 draws out of 100 games.


lkaufman Wrote:
I suppose that to a great extent the result of a match between a top engine and a correspondence player using that engine depends on whether the engine has an opening book and if so how good it is. Maybe my 65% figure is realistic only if no opening book (or a minimal or bad one).
Komodo rules!

Larry must be right...higher than 65% - likely MUCH higher in a match. Single games are too randomizing anyway, even the short-lived 'knock out' formulas in regular chess produced World Champions that would likely never have won the title by traditional means.

Most ever so called "Correspondence" 'top player' these days got their due to their use of engines anyway...

[jr66]

Well, it is sure the biggest advantage in favor of CC player is databases and also some others factors which are less determinant year after year ( for example, i improved often SF 9 play but it is much more difficult now )
If you consider several games with openings choices open with no Book for engine alone, i think however the ICCF champion will win the match ???

[Peter Berger]

If you consider several games with openings choices open with no Book for engine alone, i think however the ICCF champion will win the match ???

I would certainly hope so (else correspondence chess were +completely+ futile).
It is probably a matter of patience and research mostly these days. From time to time you come across opening lines were you spot that engines have a problem, ( usually with the help of another engine or high-level human analysis). These cases have become rare, but they still exist.
But even then one may wonder if this ICCF champion ability couldn’t be automated itself – mostly this is just clever database management, isn’t it?
I think I could be an ICCF GM if I cared enough and had the time and ressources necessary. But what a boring achievement.
OTOH over the board chess at a sub-GM-level has actually become more interesting. You can always check where you went wrong and given that you are determined enough improvement has become easier. IMHO – YMMV.

[Cornfed]

I would certainly hope so (else correspondence chess were +completely+ futile).

Many have been saying that for years now...

It is probably a matter of patience and research mostly these days.

P...as in 'patience', not so much 'playing'. Games should be for playing.

YMMV...

[jr66]

In such match, choice of openings would be important of course but it is not only a clever databases management !
Even today, engines have horizon effect and a good CC player can see when an evaluation is wrong but, unfortunatly, it do not mean that he will find a better move which win and draws become painful Indeed !!!

[Milos]

Well, it is sure the biggest advantage in favor of CC player is databases and also some others factors which are less determinant year after year ( for example, i improved often SF 9 play but it is much more difficult now )
If you consider several games with openings choices open with no Book for engine alone, i think however the ICCF champion will win the match ???

No human can win even a single game out of hundred against unassisted SFdev even on a very moderate hardware. Bear in mind SF would have 120 hours of thinking time per move at least. That is on average depth over 70 from a starting position. It would be simply 100 draws in a 100 game match.
And databases are so overrated. In 120 hours per move there is virtually 0 chance engine would make a subpar move in an opening. Chances are much higher for a human though.

[jr66]

Well, it is sure the biggest advantage in favor of CC player is databases and also some others factors which are less determinant year after year ( for example, i improved often SF 9 play but it is much more difficult now )
If you consider several games with openings choices open with no Book for engine alone, i think however the ICCF champion will win the match ???

No human can win even a single game out of hundred against unassisted SFdev even on a very moderate hardware. Bear in mind SF would have 120 hours of thinking time per move at least. That is on average depth over 70 from a starting position. It would be simply 100 draws in a 100 game match.
And databases are so overrated. In 120 hours per move there is virtually 0 chance engine would make a subpar move in an opening. Chances are much higher for a human though.

You like to break computers ?
Because i really don't use so much engines each day when i play !
You have others ways than wait the first move after 120h ???
The correspondance time control is not adapted for a such match of course

[chrisw]

Time is not the issue in human vs comp games, whereas it is very much the issue in human human, or human comp.

For human human equal time equalises the amount of thought and the amount of pressure/stress.
For comp comp equal time equalises the look-ahead capability.

But what does more time do for human? Human performance is asymptotic with time, there gets to be a point where more and more is not useful. Less time for a human creates stress and mistakes. There’s probably a sweet spot, you can call it enough time. None of that applies to a computer. All a strong engine needs is enough time to find moves that outperform the strongest humans operating with enough human time. Quite probably that’s only a few seconds per move for Stockfish.
It really makes no sense to talk about Elo versus humans when the human is deliberately placed under stress and is going to makes mistakes that he otherwise would not make.

[towforce]

Does anybody have any information about the draw ratio in computer v computer games at long time controls (24 hours per move or more)?

Maybe Dann has done some experiments in the course of heating his house?

I would think that at 24h per move, you're going to see draw ratios of > 98% with top quality and roughly equal strength computers. If so, then I would expect correspondence GMs to find it very difficult to beat the computers at this time control as well: IMO, the correspondence GM would see less than the computer, even with computer assistance!