Can the top players beat the best computers anymore?

Uh...these statements mean exactly the same thing.

no no no, calculate one move per second not play one move per second.

If they can calculate as fast as I can, then we'll see who knows more.

Alternatively, let me calculate tens of millions of positions before each move.  I think the first idea is more practical in terms of time needed

@bondocel, your analogy is an excellent one (although I suspect you may not realise why), since very recently human-powered flight using a machine with flapping wings was achieved for the first time (it went substantially further than the Wright brothers' first flight, from which I draw no conclusions). Similarly, draws have been achieved by humans against the strongest computers, with correspondence speeds being the (relatively) easiest at which to achieve this.

Just to clarify the output log here is an example from the game:

22... Qf7 { Ra8 -29 13 : Qf7 -21 13 : Qe8 -17 13 : Ra7 -14 13 }

On move 22 Black played Qf7.

The engine preferred 22...Ra8, scoring it -29 at depth = 13 and 22...Qf7 was 2nd choice with a score of -21 at depth = 13.  The played move was 2nd choice (in bold). 

So the output log shows moves analysed, score & depth at which the analysis for that ply ended.

Obviously where a move is outside of the top 4 engine choices, you can see because none of the output for that move number are in bold.

Apart from the actual match up % rates, it's interesting to see how close the played moves are to Rybka's choices when the position is fairly balanced & there are several candidate moves which have little between them in terms of engine evaluation score.

----------------------------------------------

http://www.chess.com/votechess/game.html?id=15862

Deep Rybka 3 x64 Hash:1024 Time:45s Depth:12-22ply

AMD Phenom x 4 2.30Ghz 4GB DDR2 RAM

Database Used: Batch Analyzer

[Event "Alliance versus Computer"]
[Site "Chess.com"]
[Date "2010.04.26"]
[White "The Chess.com Alliance"]
[Black "Engines"]
[Result "1/2-1/2"]
[Analysis "Deep Rybka 3 x64 Hash:1024 Time:45s Depth:12-22ply"]

{ Book Moves: }

1. d4 f5 2. g3 Nf6 3. Bg2 d5 4. Nf3 e6 5. c4 c6 6. 0-0 Bd6 7. Nc3 0-0 8. c5 Bc7
9. b4

{ Analyzed: }

9... a5 { Nbd7 19 13 : Ne4 19 13 : a5 27 13 : b6 31 13 }

10. b5 { b5 19 13 : bxa5 10 13 : Qb3 7 13 : Qa4 -1 13 }
10... Ne4 { Ne4 9 14 : Qe8 51 13 : b6 62 13 : Nbd7 66 13 }

11. Nxe4 { Qc2 14 15 : Qd3 14 15 : Qe1 14 15 : Nxe4 1 15 }
11... fxe4 { fxe4 1 14 : dxe4 59 14 : cxb5 262 14 : Qe8 342 14 }

12. b6 { b6 1 14 : Nd2 -20 14 : Bg5 -37 13 : Ne1 -51 13 }
12... Bxb6 { Bxb6 1 14 : Bd6 65 14 : exf3 89 14 : Bxg3 89 13 }

13. Bg5 { Bg5 1 14 : cxb6 -10 14 : Ne5 -36 14 : Ne1 -47 13 }
13... Qe8 { Qe8 2 15 : Rf6 136 15 : exf3 173 14 : Qc7 256 14 }

14. cxb6 { cxb6 2 14 : Ne5 -12 14 : Nd2 -54 14 : Ne1 -62 13 }
14... exf3 { exf3 1 15 : Qg6 49 15 : h6 54 14 : Qh5 63 14 }

15. Bxf3 { Bxf3 1 14 : exf3 -7 14 : Bh1 -116 13 : Bf4 -267 13 }
15... Nd7 { Nd7 1 14 : Ra6 17 14 : e5 23 14 : Rf7 46 14 }

16. Rb1 { Rb1 1 14 : e4 -13 14 : Rc1 -13 14 : Qb3 -15 14 }
16... Ra6 { Ra6 -2 14 : h6 6 13 : e5 9 13 : Qg6 12 13 }

17. Bf4 { Bf4 -14 14 : Rb3 -14 13 : Bd2 -15 13 : Bg2 -17 13 }
17... Nxb6 { Nxb6 -8 14 : Rxb6 0 14 : e5 19 14 : a4 56 14 }

18. e4 { Qd3 -8 13 : Bd6 -15 13 : Qc2 -15 13 : h3 -24 13 }
18... Qf7 { Nc4 -23 14 : Qe7 -3 13 : Qd7 -2 13 : Qd8 0 13 }

19. Bh5 { Bh5 6 14 : Re1 -9 14 : Be2 -9 13 : h4 -12 13 }
19... Qd7 { Qd7 -5 15 : Qe7 7 14 : g6 11 14 : Qf6 108 14 }

20. Be2 { Be2 -5 14 : Re1 -12 14 : Qc2 -12 13 : a4 -14 13 }
20... Nc4 { Nc4 -5 14 : Qd8 142 14 : e5 149 14 : Rxf4 192 13 }

21. Bd3 { Bd3 -5 14 : Rc1 -13 14 : Re1 -12 13 : Qc2 -15 13 }
21... b5 { b5 -5 13 : h6 0 13 : Ra8 0 13 : Ra7 1 13 }

22. e5 { Re1 -8 13 : Qe2 -9 13 : Qh5 -12 13 : Kg2 -15 13 }
22... Qf7 { Ra8 -29 13 : Qf7 -21 13 : Qe8 -17 13 : Ra7 -14 13 }

23. Qg4 { Qg4 -21 14 : Qc2 -28 14 : Rb3 -46 14 : Re1 -47 14 }
23... Ra7 { Bd7 -26 14 : Kh8 -26 13 : Ra7 -21 13 : Ra8 -20 13 }

24. Qh4 { Qh4 -30 13 : Rb3 -30 13 : Qh3 -33 13 : Rbc1 -41 13 }
24... g6 { g6 -33 13 : h6 0 13 : g5 164 13 : h5 229 13 }

25. Bh6 { Rb3 -33 14 : Bh6 -35 14 : Qg4 -49 14 : Bc2 -48 13 }
25... Re8 { Re8 -35 14 : b4 30 14 : Qf3 36 14 : Rb7 41 13 }

26. Qg4 { Rfe1 -18 14 : Be2 -26 14 : Qg4 -35 13 : Rfc1 -41 13 }
26... Ba6 { b4 -18 14 : Ba6 -18 14 : Kh8 7 14 : Na3 21 13 }

27. Rfe1 { Rfe1 -18 14 : Rfd1 -37 13 : h4 -47 13 : Rfc1 -49 13 }
27... b4 { b4 -18 13 : Kh8 -3 13 : Rc8 0 13 : Raa8 13 13 }

28. Rb3 { Rb3 -17 14 : Rbc1 -83 14 : Qe2 -87 14 : Bc2 -91 13 }
28... a4 { a4 -17 13 : Kh8 0 13 : Raa8 17 13 : Rb8 44 13 }

29. Rxb4 { Rxb4 -14 15 : Rbb1 -108 15 : Bxc4 -163 15 : Bxg6 -229 14 }
29... Nxe5 { Nxe5 -14 14 : Na3 53 14 : Nd2 32 13 : Bb5 34 13 }

30. Rxe5 { Rxe5 -14 13 : dxe5 -44 13 : Bxg6 -354 13 : Qe2 -518 13 }
30... Bxd3 { Bxd3 -14 13 : c5 153 13 : Raa8 395 13 : a3 406 13 }

31. Qd1 { Qd1 -15 14 : Re1 -16 13 : Kg2 -25 13 : h4 -27 13 }
31... Be4 { Be4 -15 15 : Bb5 0 15 : Bc4 -1 14 : Bf5 0 14 }

32. Bf4 { Bf4 -15 15 : Bc1 -29 15 : a3 -28 14 : Bd2 -41 14 }
32... Rb7 { Rb7 -15 14 : Qe7 2 14 : Qf8 7 14 : Ra5 8 14 }

33. Qxa4 { Qxa4 -20 15 : Rxb7 -23 15 : Qe1 -56 15 : Qd2 -59 15 }
33... Rxb4 { Rxb4 -20 14 : h6 0 14 : Ree7 0 14 : Qe7 2 14 }

34. Qxb4 { Qxb4 -20 12 : Qd1 -963 12 : Qa6 -1426 12 : Qa3 -1390 11 }
34... Qa7 { Qa7 -20 13 : h6 0 13 : Qe7 0 13 : h5 10 13 }

35. Qc3 { Qc3 -20 13 : f3 -49 13 : Rxe4 -87 13 : Qd2 -91 13 }
35... Qxa2 { Qxa2 -22 14 : Qa4 0 14 : Qb7 0 13 : Qb6 0 13 }

36. f3 { f3 -22 13 : Kf1 -70 13 : Bd2 -84 13 : Qxc6 -95 13 }
36... Qb1+ { Qb1+ -22 14 : Qc2 -14 14 : Rb8 0 14 : Bf5 43 13 }

37. Kg2 { Kf2 -19 15 : Kg2 -19 15 : Bc1 -29 15 : Qc1 -97 15 }
37... Qa2+ { Bd3 -19 15 : Qd3 -11 15 : Qa2+ 0 15 : Qc2+ 0 15 }

38. Kh3 { Kh3 0 13 : Bd2 0 13 : Kg1 -22 13 : Qd2 -22 13 }
38... Qc2 { Qc2 0 14 : Bb1 42 14 : Bc2 49 14 : Bf5+ 104 14 }


1/2-1/2

{ Game Summary }

{ White: The Chess.com Alliance }
{ Top 1 Match: 23/29 ( 79.3% )
{ Top 2 Match: 25/29 ( 86.2% )
{ Top 3 Match: 26/29 ( 89.7% )
{ Top 4 Match: 27/29 ( 93.1% )

{ Black: Engines }
{ Top 1 Match: 24/30 ( 80.0% )
{ Top 2 Match: 26/30 ( 86.7% )
{ Top 3 Match: 29/30 ( 96.7% )
{ Top 4 Match: 29/30 ( 96.7% )

{ All Players }
{ Top 1 Match: 47/59 ( 79.7% )
{ Top 2 Match: 51/59 ( 86.4% )
{ Top 3 Match: 55/59 ( 93.2% )
{ Top 4 Match: 56/59 ( 94.9% )

Just a quick glance through the game does raise some interesting points.

White chose the 1st choice move in many instances when there were several candidates all within a few centipawns of best move. eg:

16. Rb1

17. Bf4

19. Bh5

20. Be2

21. Bd3

23. Qg4

24. Qh4

27. Rfe1

31. Qd1

32. Bf4

&

38. Kh3

Where the draw was astutely agreed when Rybka shows the score as 0.00 after this move from White.

White dropped around 0.20 of a pawn twice I think.  Other than that a very accurate game indeed.

Svidler once said that he simply does not consider an evaluation of an engine which is + or - 0.50. He added that this is how a machine with no chess understanding evaluates an equal position.

I personally believe that no computer will ever be better than humans at chess. True, they may beat top players a lot, but even so, GMs don't always win. Not every player plays his best game every time he sits down at the board, and computers are not affected by things like psychology, sleep deprivation, emotions, or even real thought. All they do is cold calculation.

Which is why computers are most dominant with tactical games. I believe a deep positional player could beat any computer with enough experience. Strategy, an integral part of chess, involves planning of long-term objectives. It has little to do with calculation, which makes humans the better strategic opponent, IMO.

That's not to say a future Tal might not beat the computer, just that you have to find your opponent's weakness and make him suffer for it. In this case, the computer can only calculate. That is a limitation, and simultaneously its greatest strength and weakness.

Why not ... I'd like to see him try holding water back better than the Hoover Dam :).

Did you actually watch the video in the page I linked, notlescu? Your use of the word "gliding" suggests either you did not, or you have no idea what the word means. Of course, I too would be more impressed if they had managed to take off, but the video makes it entirely clear that height was maintained for a substantial period of time by human power (actually height increased for short periods). Somewhere in that article or another one it points out that the large majority of the lift (and all of the thrust, of course) achieved by the craft comes from the flapping rather than the aerofoil shape. The way in which birds (and bumblebees) create lift is a rather difficult fluid dynamics problem, which requires very precise analysis to understand. So to describe what was achieved as "gliding" either indicates poor understanding or simply troll-like behaviour.

Interestingly, it is very likely that the ancestors of birds were unable to take off from level ground. The progression is very likely to have been from gliding (as many current tree dwelling creatures), flapping to guide and extend gliding, launch from a height with the ability to maintain and gain a little height, and only much later to being able to take off without an initial drop. In terms of the behaviour, this craft could be argued to be at stage 3 of this progression, in a very limited (and impractical) way.

[In case there is any doubt, I don't think it's a practical means of transport! Not unless you need to only go a couple of hundred yards, have suitable space to take off and land and a powered winch, be exceptionally fit and lean, be wealthy enough to afford to own and maintain such a vehicle, be averse to all the alternative modes of locomotion, and probably a few other unlikely conditions I have omitted. Humans are indeed not well designed for powered flight. Or doing chess problems.]

Is there a difference in adding more ply to the calculation search? Yes! Human chess masters have bragged for decades that computers were incapable of long-term planning and strategy and that humans can see in an instant a pattern or position that could lead to a favorable game.

Well with the development of faster computer processors and better programming, computers can now calculate the same long-term plan or strategy in the same time that a human master can recognize it.

Sure computers cannot recognize all the positions instantly without a database, but with faster processors, the 21-ply calculation could be reduced to mere seconds.

We know that ratings increase with each additional win against equally-rated opponents. As long as the engine keeps winning, its rating will continue to rise. Even when the second-best opponent does not increase in rating, the winner's increase will only be slower but it will not stop as long as it does not stop playing.

You would be pleased to know that there is a thread started by NM Charles Galofre titled, "The future of computer chess!!-!?-?!," and introduces additional analysis for equal positions. It is a theoretical discussion on how computers can win even in equal positions, finding new ways to win even when the strongest engines are attempting to force draws.

If humans are already struggling to force draws with computers, it would be a steeper struggle against engines that will be programmed to deal with that strategy.

 Yes

Humans will always be able to win in my opinion. It most definetly will not be easy but we can. Chess is a game where every action, attack or defense can be undone. There is no perfect chess defense or attack. Therefore there is always a chance of victory even if it seems impossible. We would have to get very good at chess. Some even devote their lives to it and perhaps take classes.

BUT. It can be done.

Also this depends on whether the games are timed or untimed. Timed games yes humans will come to the point where it is impossible to win; only because our minds can't function as fast as a computer...

Untimed however is a totally different story. Computers have infinite patience. Chess players can leave go relax, eat coffee hang out and come back to be greeted by their computer. Still awaiting your move. This means we will have the upperhand. You can do one move a day. Sit and study to find the best possible move like in our tournaments here on chess.com. The score between humans and computers will always be tied.

Computers > Humans ~ Timed Chess.

Humans > Computers ~ Untimed Chess.

I hope my answer proves adequate. I think it does. ;D

Do you really believe that humans can still beat the best chess engines when the machines are allowed as much time as needed to find the best possible moves?

Nice! :) Well, the original dream was that the human uses the enrgy produced by his body to lift and fly. In this project, he only flaps the wings while gliding.

You're wasting your time. Most of chess.com members are unaware of sites like "google" or "wikipedia". But I guess that each one has the right to believe in whatever he wants. If someone, with plenty of information available, believes that the Earth is flat, who are we to convince them of the contrary?

IMHO that's not a very good argument. If you handicap a machine by removing one of the ways it's stronger than humans, then duh, of course you'd have a better chance of beating it. I mean, that's kind of like saying "An airplane can fly better than a human? Nah, just take away its wings and let's see it fly better than me." or "A car can attain a higher speed than me? Nah, let's just take away its engine and I can outrun it easily."

Also, the people here insisting that humans can beat current computer engines in a fair match need to get a Clue(TM). I hear they're sold in all major supermarkets, next to the Life aisles.

[double post deleted]

Gliding is a mode of unpowered flight where a craft loses altitude in order to maintain speed, not a form of powered flight where a craft maintains the same height and speed. You are free to be unimpressed by the achievement, but not to redefine the meaning of English words.

This is only true if the stronger computer wins all of the time. Even the best chess computers at present lose a few games and draw a lot more against their lower rated opposition. With an adequate number of games, a computer finds their true rating level and maintains it. In order for the ratings to rise, they need to get stronger, not merely to play more.