Magnus Carlsen Vs Houdini 3 Pro in Correspondence Chess - Who wins?

Hello.

I know that in blitz computers can completely obliterate any super-GM. Even in classical chess, computers are superior, though super-GM can sometime draw them. But what about in correspondence chess?

Let's say we make a match between Magnus Carlsen and Houdini 3 Pro, consisting of 10 games, with a time control of one week per move (and we will suppose that Carlsen stays focused at 100% during the entire match and meditates about the ongoing game at least 14 hours a day).
Houdini gets an opening database and the Nalimov endgame tablebases. While Carlsen isn't allowed to use anything but his brain (he probably wouldn't need any opening database anyway).
Who would win?

Originally posted by Marc Benford
Hello.

I know that in blitz computers can completely obliterate any super-GM. Even in classical chess, computers are superior, though super-GM can sometime draw them. But what about in correspondence chess?

Let's say we make a match between Magnus Carlsen and Houdini 3 Pro, consisting of 10 games, with a time control of one week per move (and we w ...[text shortened]... ything but his brain (he probably wouldn't need any opening database anyway).
Who would win?

Hello

Are you looking for a decent opponent for your recent purchase by any chance? Thread 152917 😉

Anyway I think you'll find he is busy right now preparing for a world championship.

What you describe kind of happens anyway on many correspondence sites as Houdini needs an operator to feed in the moves and carry out the menial admin tasks. Of course that is all well and good when the opponent is 100% Human and not allowed to consult reference material, however when they are not and do....

PS: Is there any reason why you repeat your questions on a number of different chess fora?

You are full or rubbish. Strong GM's can still beat the best computer.

Originally posted by plopzilla
You are full or rubbish. Strong GM's can still beat the best computer.

Can they?

After Adams v Hydra (2005) and Kramnik v Deep Fritz (2006) and the odds matches by Ehlvest and Benjamin in 2007 - all losses for the humans by the way - I can only find Nakamura's ICC Blitz wins against Rybka in 2008. It doesn't seem that there are too many out there willing to try.

In CC we might look at OTB/CC GMs first as possible material for a match given their proven correspondence experience. Looking at their ICCF profiles it is about ten years since Curt Hansen, Ulf Andersson or Johnny Hector made a correspondence move, yet two of the three were FIDE active this year. I don't know of any other GMs who were CC players in the modern era. Alright there was Mark Hebden but not since the 1980's I believe.

In a 2002 article David Levy had written off humans vs chess engines in the long run and advised a move over to GO, on the grounds that there were many more possibilities than in chess and that no algorithms had been developed from the best human players to allow a machine to make sense of the game. I don't know if GO engines have leapt forward in the 11 years since or whether that is still a game where the human can seek sanctuary.

Originally posted by Ragwort
Can they?

After Adams v Hydra (2005) and Kramnik v Deep Fritz (2006) and the odds matches by Ehlvest and Benjamin in 2007 - all losses for the humans by the way - I can only find Nakamura's ICC Blitz wins against Rybka in 2008. It doesn't seem that there are too many out there willing to try.

In CC we might look at OTB/CC GMs first as possible materi d in the 11 years since or whether that is still a game where the human can seek sanctuary.

On June 5, 2013, computer program Zen defeated Takuto Ooomote with a 3 stone handicap. Takuto Ooomote is a 9 dan on the Tygem server. The 19×19 game used Japanese rules with a time setting of 60 minutes plus 30 seconds byoyomi. They played at the 27th Annual Conference of The Japanese Society for Artificial Intelligence.[22]

From Wiki. Sounds like Go players better worry too🙂 3 stone handicap against a 9 dan amateur player is pretty good! So it would not win yet with no handicap in a Go tourney but they would do pretty good against amateurs.

Originally posted by sonhouse
On June 5, 2013, computer program Zen defeated Takuto Ooomote with a 3 stone handicap. Takuto Ooomote is a 9 dan on the Tygem server. The 19×19 game used Japanese rules with a time setting of 60 minutes plus 30 seconds byoyomi. They played at the 27th Annual Conference of The Japanese Society for Artificial Intelligence.[22]

From Wiki. Sounds like Go pl ...[text shortened]... uld not win yet with no handicap in a Go tourney but they would do pretty good against amateurs.

I'm thinking the only reason Go has been safe is that programmers haven't spent as much time on it.

Originally posted by SwissGambit
I'm thinking the only reason Go has been safe is that programmers haven't spent as much time on it.

True, but I think this is partly recursive - the reason programmers haven't spent as much time on Go as chess is because chess is easier - from a computing point of view.

There are several reasons for this, the main one being the branching factor. In chess the average number of legal moves at any time is around 30. In Go there are 19x19 places to go, and you can go anywhere (except spaces already occupied by stones). This makes tree-based search intractable in Go.

Chess also lends itself quite nicely to mathematical evaluation - roughly, chess engines give +1 for a pawn, etc, if the position is equal in other ways. In Go this is less useful - all the pieces have the same value, so the thought process is more diagrammatic. This is much harder to encode algorithmically .. but not impossible. As has been mentioned Go computers are catching up.

On a personal level, I like the idea of Go but I'm hopeless at it. I tried playing a good human once, and after a while he said "Well, you've lost that". I had no idea why the game was even over. So I'll learn Go when I've mastered chess. :-)

Who would win?

Engines have been programmed by many programmers with the help and advice of more than one very very strong chess player. The programmers had months to do their research and fine tune the transistors. So a single human playing under tournament time limits against an engine is essentially playing against a roomful of players who had all the time they needed. Think of an engine as a proxy for a roomful of GMs & geeks compressed in time, and no one man is likely to be a match for them.

I think, therefore, the only fair match these days is a consultation of humans against an engine, not OTB with tourney time limits, but correspondence time limits. Make it a Turing test if you want: the consulting humans do not know whether they are playing against an engine or another team of consulting humans.

In this case, the constellation of human players would be significant: if you got just the right mix of deep strategists and tacticians and endgame boffins, I think they could out-fox the best engines in a match.