Chess Computers are they IM? or GM? or Super GM?

[ml]

The games against humans the SSDF list was originally based on were played more than 15 years ago, and the few games by "masters" were players rated nowhere near today's top GMs or programs.

True. At that time, GMs could regularly defeat the top programs and chess machines. It would have made no sense to use GMs to calibrate computer ratings.

Then 7 1/2 years ago the SSDF concluded program ratings were inflated, based on a variety of human-computer games by various programs on a variety of different hardware between 1997-2000, and dropped all program ratings by 100 points.

And the SSDF pointed out that this was only an ad hoc measure to compensate for inflation at the top of the list. It likely resulted in an overcompensation at the lower end of the list. Now it appears that another correction may be required at the top.

So while what you said is true, and the SSDF has done their best to keep the list calibrated to humans, it is also very unclear how closely present day SSDF ratings track present day FIDE ratings.

We do have a few official human vs. computer games involving GMs and IMs, though not nearly enough yet. The statistical error of this small dataset of games should be an indicator of the extent of this lack of clarity you refer to. But as computer ratings continue to increase, we will eventually get to a point where the large rating differences between the top GMs and top engines will preclude human ratings, including FIDE, from being used as a standard of reference for computer ratings.

[Uri Blass]

If a Chess Computer had the following:

1) one of the top five SSDF computer chess engines;

2) single Intel Xeon Quad core; and

3) eight gigs of RAM.

At Long Time Controls, would it be an IM or GM or Super GM?

At Long Time Controls, what would be the ELO range?

Thank you and comments are welcome

I guess that the elo range is going to be 2800-3100

Computers already achieved
performance of more than 2700 at tournament time control some years ago and computers clearly improved from that time.


Junior and Fritz and Rebel drew match against super grandmasters in
2002-2003
see
http://www.chessbase.com/newsdetail.asp?newsid=1229

Uri

[Dr.Ex]

If a Chess Computer had the following:

1) one of the top five SSDF computer chess engines;

2) single Intel Xeon Quad core; and

3) eight gigs of RAM.

At Long Time Controls, would it be an IM or GM or Super GM?

At Long Time Controls, what would be the ELO range?

Thank you and comments are welcome

I guess that the elo range is going to be 2800-3100

I would say it depends on the rating average of the human opponents.

I'd guess this Chess Computer would not score more than 90% against 2400 average rated opponents -> <2800

My wild guess is 75% against 2750 average rated opponents -> 2950

[S.Taylor]

So... the original question still remains unanswered !

Definitely a super grandmaster is my answer....

The only thing lacking, according to those who believe like you do, (and I think I do too), is that computers cannot yet play some of the wierd and wonderful games such as the famous Spaasky vs Larsen Belgrade 1971, from Spaasky's side (Black). Like Ng4, h5, h4(maybe only this), and some other immortal games (I mean which are also the strongest moves, and without faults).
This is besides the other known weaknesses in present day computer chess progress.

I also wonder how it is possible to rate computers so highly, when some chess players who are not even at the top, know how to subdue computers by exploiting computer weaknesses.

[Ovyron]

Probably the high rating range of the engines would only be possible if the humans don't know they're playing a computer, as knowing it may cause them to play anti-computer chess and that distorts the ratings (for example, some anti-computer expert may have a much stronger rating against computers than a Super GM, and that would make the computer underrated.)

[Marc MP]

If a Chess Computer had the following:

1) one of the top five SSDF computer chess engines;

2) single Intel Xeon Quad core; and

3) eight gigs of RAM.

At Long Time Controls, would it be an IM or GM or Super GM?

At Long Time Controls, what would be the ELO range?

Thank you and comments are welcome

I'm quite sure that OTB, comps beat the humans with an appropriate opening repertoire (see last Kramnik-Fritz match).

By correspondence, I would be curious to compare say, a correspondence GM with some computer help (database plus single cpu engine) against one monster machine running a top program (just as you mentioned).

[ozziejoe]

you can get some idea by looking at the chess engine rating on ICC (which presumably play against alot of humans.

.e.g, linuxknight runs rybka 2.3.2a mp on a core 2 quad....its bullet rating is 2850, blitz rating is 3369, and standard rating is 2616 (127 wins, 7 losses, 1 draw).


That standard rating seems a bit low. I seem to recall crafty having a much higher rating than that at standard, but their are no crafty's running standard now

[bigo]

This question has been so completlely answered that to come to any other conclusion other then that today's programs are Super Grandmaster would demonstrate human bias and prejudice.

[BBauer]

Hi,
to answer your question:

Computers are no masters, they are computers.
Often they don't understand the opening.
Often they don't understand the middle game.
Often they don't understand the end game.

Here an example for the middle game:
[d]r1b1N2k/1pBn2p1/p3Q2p/5n2/8/2q5/2PRB1PP/7K w KQkq -


Here an example for tne endgame:
[d]8/4k3/5p2/6p1/4P1Bp/5PP1/1r5P/6K1 b - - 0 1

Masters and grandmasters understand the game, but
a fool with a tool is still a fool.

kind regards
Bernhard

[Ovyron]

Computers are no masters, they are computers.
Often they don't understand the opening.
Often they don't understand the middle game.
Often they don't understand the end game.

Define "Often."

Until it's defined I could say that:

Grandmasters often don't understand the opening.
Grandmasters often don't understand the middle game.
Grandmasters often don't understand the end game.

And that "often" is when computers that understand it are able to defeat them, that's why computers "often" beat the masters.

After a deeper look, the human may comprehend that it's him who doesn't understand a position, not the computer (Just look at one of these 6men tablebase wins that only have one winning move that the humans can't understand why it's the only move that wins.)