Can A GM And Rybka Beat Stockfish?

[Ozymandias]

I believe that with big prize money and match of 100 games the GM+computer can get more than 50%.

Finally the voice of reason. This is on the money IMO.

Problem being the money, because we'll never see something like this.

[Harvey Williamson]

In some games a computer will make an anti-positional move that the GM with computer help can exploit.

If they couldn't exploit them, at a time when 32-bit Pentium IV and Fritz still dominated, 6-men Nalimov were a novelty and opening books were laughable, what chance do they have today, against 64-bit MP multicore servers, running Komodo with 6-men syzygy and a good public book?

We are not talking about gm v comp. We are talking about advanced Chess when a strong player can have some influence on the play of a computer.

[Ozymandias]

I prefer to see same hardware for both sides and same books for both sides.

Note that PAL/CSS means that amateurs can easily cheat by registering with more than one entry and simply win against themselves.

Books at the time had lots of holes, GMs should've been benefited by this, but they weren't. Hardware was weak too, the winners had an AMD 3200+, a 2.8 MHz and a 1.6 MHz Pentium, Vladimir Dobrov (and his 2600+ colleague), didn't say what computers they were using, but they couldn't possibly be much worse. As for collusion, that's indeed been a concern, in swiss tournaments, but in this case it was a direct match between the two teams. Maybe 4 games weren't much, but enough to discourage GMs from participating in future events.

[Ozymandias]

In some games a computer will make an anti-positional move that the GM with computer help can exploit.

If they couldn't exploit them, at a time when 32-bit Pentium IV and Fritz still dominated, 6-men Nalimov were a novelty and opening books were laughable, what chance do they have today, against 64-bit MP multicore servers, running Komodo with 6-men syzygy and a good public book?

We are not talking about gm v comp. We are talking about advanced Chess when a strong player can have some influence on the play of a computer.

You said that the positional mistakes of an engine could be exploited by a computer-aided GM. If that didn't happen 10 years ago, when those mistakes where more glaring, how do you expect they will, now?

[Harvey Williamson]

In some games a computer will make an anti-positional move that the GM with computer help can exploit.

If they couldn't exploit them, at a time when 32-bit Pentium IV and Fritz still dominated, 6-men Nalimov were a novelty and opening books were laughable, what chance do they have today, against 64-bit MP multicore servers, running Komodo with 6-men syzygy and a good public book?

We are not talking about gm v comp. We are talking about advanced Chess when a strong player can have some influence on the play of a computer.

You said that the positional mistakes of an engine could be exploited by a computer-aided GM. If that didn't happen 10 years ago, when those mistakes where more glaring, how do you expect they will, now?

Look at some of the freestyle games of GM Jobava. A strong GM when he 1st played the PAL freestyle tournament he took down some strong engines on very good hardware aided only by an old version of Fritz on a slow laptop. He is still rated as one of the top freestyle players with the username, Oligarck. Look at many correspondence players they use computer software well. Uri is a GM and former Israeli Champion. I am a former British Champion and a Senior International Master. You don't win games at this level without steering engines a little.

[SuneF]

GM+SF will be a high-quality more selective engine than SF, so should easily be stronger at longer time controls.

In correspondence chess SF would obviously stand little chance against GM+SF.

And GM-level is probably not needed. Any decent human chess player has a lot more common sense than any chess engine. Just a bit of that to direct the search to promising lines should do the trick.

At shorter time controls letting the user decide when to play a move might already be enough to get a significant edge over SF on its own.

Yes better usage of the clock would be one area where the human-comp would be better.

Also AFAIK engines are still not good at recognizing fortresses much less create them, to hold draws.

Finally I'd never trust a null-moving engine in positions with lots of zugzwangs, so I'd turn that off for the computer that was helping the human when positions like that started appearing in the search.

It would take some practice and skill to make the most of these advantages though and it probably wouldn't even be significantly stronger but rather something like engine+delta.

[syzygy]

GM+SF will be a high-quality more selective engine than SF, so should easily be stronger at longer time controls.

In correspondence chess SF would obviously stand little chance against GM+SF.

And GM-level is probably not needed. Any decent human chess player has a lot more common sense than any chess engine. Just a bit of that to direct the search to promising lines should do the trick.

At shorter time controls letting the user decide when to play a move might already be enough to get a significant edge over SF on its own.

I am not sure that in correspondence game SF would stand a little chance
against SF+GM

remember that SF play better in correspondence games relative to tournament time control so my guess is that most of the games are going to be draws.

I agree. I do not mean that SF stands no chance in a particular game. I mean that there no doubt, at least not in my mind, that in correspondence chess GM+SF is a stronger entity than SF alone.

I can imagine that at a more regular time control GM+SF might mess up, especially if the GM gets too active. At short time control the GM maybe should only decide when SF makes a move and for the rest keep his hands off the computer.

[Ozymandias]

Look at some of the freestyle games of GM Jobava. A strong GM when he 1st played the PAL freestyle tournament he took down some strong engines on very good hardware aided only by an old version of Fritz on a slow laptop. He is still rated as one of the top freestyle players with the username, Oligarck. Look at many correspondence players they use computer software well. Uri is a GM and former Israeli Champion. I am a former British Champion and a Senior International Master. You don't win games at this level without steering engines a little.

I look at his results, and those don't compel me to change my POV.
As for correspondence players, I can only judge Uly (not Uri).

[Milos]

I believe that the GM is not going to change the machine's moves in most of the games but in some positions they are going to change it.

This is the key. In some games a computer will make an anti-positional move that the GM with computer help can exploit.

Problem is most of GMs (like 99.9%) have no clue about computer chess beside playing against engines. They don't understand how engines work, how search works (they don't even understand simple concept such as fail high and fail low, i.e. alpha-beta, null move and LMR), how is evaluation performed and especially have no clue how to use engine effectively to beat another engine. They might only recognize other engine weaknesses but that is far from enough to beat it.
Average player (like FIDE master or even weaker) who has deeper understanding of chess engine concepts (or even better has programmed an engine) would be far superior to any top GM in a game where both use the same engine.

[Milos]

At short time control the GM maybe should only decide when SF makes a move and for the rest keep his hands off the computer.

When to make the move??? You must be joking...
For the start what's the percentage of GMs that understand what fail high or fail low means?
You really think it's higher than 1%???
How the hell do you think they would be able to be better at time management than engines time manager with zero knowledge about search???
If you really think so, you are clueless, at least about engine time management.