time odds

[james uselton]

Rybka has 20 min for 40 moves and the GM has 2 1/2 hrs for 40 moves.
What is the outcome in a six game match?

[Ovyron]

Something close to 5.5-0.5 for Rybka, unless you're talking about a SuperGM.

[Dr.Wael Deeb]

6-0 for if we're talking about an average GM and 5-1 for Rybka if we're talking about a super GM....

[gerold]

3-3 if you use stonewall expert.

[shiv]

While I appreciate the strength of Rybka, I do not believe the score will be so lopsided. First of all, it depends on processor hardware. If the hardware is really good, then the time odd are insignificant.

For top players like Kramnik, Kasparov who have played against Fritz etc, I doubt the score will favor Rybka by such a margin. Rybka may be stronger than other engines, but it has not shown that it is better at beating the best humans (compared to Fritz) in a regular match.

An interesting insight in the last Kramnik vs Fritz match was that he could draw at will (except for the mate in 1 mistake). If the best humans similarly play solidly against Rybka, there are good chances of a competitive score.

In summary, Rybka would be the strong favorite against anyone, but the score depends on the strenght of the player and the hardware.

[bob]

Rybka has 20 min for 40 moves and the GM has 2 1/2 hrs for 40 moves.
What is the outcome in a six game match?

That probably doesn't turn into very much in terms of odds. "pondering" changes how things work significantly.

[bob]

6-0 for if we're talking about an average GM and 5-1 for Rybka if we're talking about a super GM....

I would be surprised at 6 0 at any time control against a strong GM. Draws are certainly going to happen, and probably a loss somewhere along the way.

[Dr.Wael Deeb]

6-0 for if we're talking about an average GM and 5-1 for Rybka if we're talking about a super GM....

I would be surprised at 6 0 at any time control against a strong GM. Draws are certainly going to happen, and probably a loss somewhere along the way.

I hope to live the day and see with our own eyes....

[Thomas Mayer]

Rybka has 20 min for 40 moves and the GM has 2 1/2 hrs for 40 moves.
What is the outcome in a six game match?

As Bob already pointed out you have to take care of pondering. With pondering switched on the time odd won't be too big because it can use the 2 1/2 hrs of the opponent as well.

So let's start with ponder off which should reduce it's strength in comp-comp games something about 30-40 Elos... (All given values are VERY rough estimates here) -> we have no clear values for the play against humans without ponder. I believe myself the loss in strength is bigger then because psychologically an engine that moves because of ponder hits for several moves instantly might be a big problem for a human -> especially when you have some problems on the board and take a longer time it happens quite often that you ran in the ponder hits of the engine and get always fast responses at a moment where you definitely don't want to have them.

A good estimate about doubling time in Elo seems to be 60. (Comp-Comp games !) --> 150 min / 2=75 min (first 60 Elo) 75/2 = 37,5 (2nd 60 Elo) 37,5/2~20 (3rd 60 Elo) = 180 Elo + 40 Elo (Ponder) = 220 Elo. So my guess is that with ponder off Rybka should be about 220 Elo weaker then it's full version... 3000 - 220=2780, so still a very very tough opponent for every GM. Also with 20 min on his clock Rybka would still move very fast which doesn't help the GM. Also I believe that in comp-human games time odds don't have the same impact on strength then in comp-comp games. Some of Bobs observations go into that direction, once he mentioned that the GMs playing against Crafty do not really feel different playing against a double as fast computer. The reason is simple: IMO when an engine reaches depths about 13-15 ply in middle game it's strength in tactics is already above every human and there isn't too much change in it's strategical play at all. My own experience is the same: My engine plays from time to time in the main hall of playchess. It seems it doesn't matter which hardware I use, my results are always more or less the same.

Well, what else can influence the outcome ?

Shape of the GM, how much can he concentrate on the games, how much pressure does he have. How much interest does he have in a good result. (I believe that some of the human-comp match did also look so bad for the human side because the GMs weren't too much interested in the outcome) How good does his style fit to play against comps.

I still believe that a fully concentrated 2600+ GM who has his head free and can concentrate on the games without overlooking easy and not so easy tactics can compete with even the best programs on powerfull hardware. The problem seems to be the concentration, to withstand a full game or even a full match of several games against the full tactical power of todays beasts seems to be a nearly impossible job. Therefor the engines achieve their amazing results against humans - my reason for this is because the humans can't keep their concentration always on 110% and make mistakes on the board they would never do when just be confrontated by a single position.

So with all this in mind it is nearly impossible to guess any result. I believe everything from a 5-1 for the GM up to a 0-6 is possible but the advantages clearly lie on the computer side.
It might change a bit when you use Pablo Restrepo style in your games, a GM might play that even better then Pablo does. Nearly impossible to win such games except on time (which I wouldn't count) but a GM should achieve that way at least a 2-4, but of course he, the sponsors & the audience wouldn't like the games.

Greets, Thomas