Nakamura vs Stockfish, public match 8/23

[Rein Halbersma]

Note that a billionaire can always be number 1 in correspondence chess in case that he is interested in being number 1 simply by hiring the best advisors to help him.
The best advisors can be humans(50 years ago) or computers or a combination of them(today).

Joop van Oosterom is both a billionaire, the employer of former GM Jeroen Piket (who is currently his personal secretary, and used to have a lot of his matches sponsored by van Oosterom), but he is also the 1955 Dutch U20 Chess Champion. He appears to be a genuine chess lover and -at least 60 years ago- reasonbly high skilled at it, rather than someone who would buy a meaningless title.

[Uri Blass]

Note that a billionaire can always be number 1 in correspondence chess in case that he is interested in being number 1 simply by hiring the best advisors to help him.
The best advisors can be humans(50 years ago) or computers or a combination of them(today).

Joop van Oosterom is both a billionaire, the employer of former GM Jeroen Piket (who is currently his personal secretary, and used to have a lot of his matches sponsored by van Oosterom), but he is also the 1955 Dutch U20 Chess Champion. He appears to be a genuine chess lover and -at least 60 years ago- reasonbly high skilled at it, rather than someone who would buy a meaningless title.

My point is not about a specific person but only that the success of a billionaire in correspondence chess proves nothing
so it is not relevant for the discussion about the question if humans can be better than engines in correspondence chess.

A billionaire can have the best software+hardware in order to win but he can also have the best human people to help him in order to win(in case that humans can help).

[Laskos]

Note that a billionaire can always be number 1 in correspondence chess in case that he is interested in being number 1 simply by hiring the best advisors to help him.
The best advisors can be humans(50 years ago) or computers or a combination of them(today).

Joop van Oosterom is both a billionaire, the employer of former GM Jeroen Piket (who is currently his personal secretary, and used to have a lot of his matches sponsored by van Oosterom), but he is also the 1955 Dutch U20 Chess Champion. He appears to be a genuine chess lover and -at least 60 years ago- reasonbly high skilled at it, rather than someone who would buy a meaningless title.

My point is not about a specific person but only that the success of a billionaire in correspondence chess proves nothing
so it is not relevant for the discussion about the question if humans can be better than engines in correspondence chess.

A billionaire can have the best software+hardware in order to win but he can also have the best human people to help him in order to win(in case that humans can help).

To support the claims that ICCF ratings follow more the hardware use than human abilities in Chess, I re-post the latest ICCF rating list, plotted it here, and calculated some statistical quantities. The upper half in ratings is especially dubious, and its shape resembles that of the distribution of used hardware:


Skewness: -0.772
Kurtosis: 0.274

These are not values of particular human cognitive abilities, be it Normal or Weibull with a given shape parameter k > 0. No such Skewness and Kurtosis of Weibull for any k. The shape follows more some sort of income distribution (be it in hardware, software or other).

[lucasart]

I bet a monkey using Stockfish could be a GM in correspondance chess…

But let's not underestimate the contribution of the monkey in this pseudo-cyborg setup!

[sedicla]

It was a good effort by Nakamura but probably the last one on the man vs machine match. The only outcome of this match would be: computers are very strong, even if Nakamura had won. He was using computer assistance and pawn odds.
I think the commentator was kind of naive by not recognizing SF strength.

In my opinion this is the same as human running and car races. Usain Bolt can't beat cars but we still appreciate both races.

[Uri Blass]

I bet a monkey using Stockfish could be a GM in correspondance chess…

But let's not underestimate the contribution of the monkey in this pseudo-cyborg setup!

As a GM in correspondence chess
I have the same feeling but
it does not contradict my opinion that it is possible to have for many players elo(human+computer) > max(elo(computer),elo(human))
espacially when the computer is rybka with slow hardware and not stockfish.

I believe that when rybka3 and stockfish include the hardware for the match:

elo(stockfish)>elo(nakamura+rybka3)>max(elo(nakamura),elo(rybka3))

[Dr.Wael Deeb]

Nakamura could have drawn all four, but he prodded and pushed, looking for wins that were not there. Respect to Naka!

I don't think he had any chance to draw the match ...
I find it highly unlikely that in a perhaps historically significant match of man vs machine...he had chances to draw against a 3300 engine (possibly proving human can be equal to a machine), and simply threw them away.
If that's true, he's either incredibly conceited or a complete idiot...(both of which I doubt)

I followed the match, all 11 hours of it. Game 2 would have been drawn by either repetition or the 50 move rule had Naka chosen not to press. This was fairly obvious. The other loss was also petering out into a draw but Naka gambled and lost. If this was straight up chess, without assistance or odds, I'd agree that splitting the points is unlikely.

And we really don't know what the engine ratings are relative to humans. If Stockfish is really 500 elo stronger than Nakamura, we would expect Naka to score only 5 points out of 100 and I have to believe he would fare better than that.

Naka is not a "complete idiot" but his temperament is to fight and that was his undoing.

I cannot believe the praise Naka is getting for the games he HAD a computer to assist him. Who is to say if it was not Ryka that allowed him to draw one of the two game with all the pieces on the board. Then it was two pawn odds games ( not chess), come on people wake up. Make the games regular time controls at normal chess and he would be crushed. On more than one occasion Naka has stated this himself.

Words of wisdom unfortunately not finding their way to most people ears....
Dr.D

[rcmaddox]

Nakamura could have drawn all four, but he prodded and pushed, looking for wins that were not there. Respect to Naka!

I don't think he had any chance to draw the match ...
I find it highly unlikely that in a perhaps historically significant match of man vs machine...he had chances to draw against a 3300 engine (possibly proving human can be equal to a machine), and simply threw them away.
If that's true, he's either incredibly conceited or a complete idiot...(both of which I doubt)

I followed the match, all 11 hours of it. Game 2 would have been drawn by either repetition or the 50 move rule had Naka chosen not to press. This was fairly obvious. The other loss was also petering out into a draw but Naka gambled and lost. If this was straight up chess, without assistance or odds, I'd agree that splitting the points is unlikely.

And we really don't know what the engine ratings are relative to humans. If Stockfish is really 500 elo stronger than Nakamura, we would expect Naka to score only 5 points out of 100 and I have to believe he would fare better than that.

Naka is not a "complete idiot" but his temperament is to fight and that was his undoing.

I cannot believe the praise Naka is getting for the games he HAD a computer to assist him. Who is to say if it was not Ryka that allowed him to draw one of the two game with all the pieces on the board. Then it was two pawn odds games ( not chess), come on people wake up. Make the games regular time controls at normal chess and he would be crushed. On more than one occasion Naka has stated this himself.

Words of wisdom unfortunately not finding their way to most people ears....
Dr.D

I agree and have said as much in other threads.

What I want to learn is the objective disparity in elo between the engines and humans. We can all agree that engines will "crush" the GM's but that's not very precise. If the engines are 500 elo better than humans, let's see it demonstrated OTB. Maybe the disparity is less, maybe it's more, but it would be interesting to find out.

I realize it probably won't happen. No one enjoys a prolonged drubbing at the board. It would require money - a straight fee with bonuses for draws and wins.

[ernest]

So Nakamura did not play f5 in an attempt to win, which was my (patzer) impression from going through the moves.

Indeed, your recent Stockfish 190814 syz also prefers 115...f5

2b2r2/5r1k/1p1q1pp1/p1pPp2p/P1P1P2P/1PQ3R1/1K2B1P1/6R1 b - - 0 1
Analysis by Stockfish 190814 syz:
.................
115...f5 116.exf5 Bxf5 117.Rg5 e4 118.Ka3 Qf6 119.Qxf6 Rxf6 120.g4 hxg4
+/= (0.67 --) Depth: 39/56 00:19:51 2204mN, tb=342
115...f5 116.exf5
+/= (0.55 !) Depth: 39/58 00:34:17 3828mN, tb=773
115...f5 116.Bd3 fxe4 117.Bxe4 Bg4 118.Ka3 Rf4 119.Re3 Qf6 120.g3 Rf2 121.Rb1 Kg7 122.Rb2 Re8 123.Re1 Rf1 124.Qe3 Rxe1 125.Qxe1 Qd6 126.Qe3 Rf8 127.Rd2 Rf7 128.Kb2 Qf6 129.Qc3 Rd7 130.Ka2 Qd6 131.Qe3 Rf7 132.Ka3 Qf6 133.Kb2
+/= (0.53) Depth: 39/58 00:50:22 5641mN, tb=1129
115...f5 116.exf5 Bxf5
+/= (0.60 --) Depth: 40/58 01:02:13 6954mN, tb=1573

Some people have said that 116...Rxf5 was a mistake, better would have been 116...Bxf5
but your Stockfish 190814 syz finds both moves quite equivalent:

2b2r2/5r1k/1p1q2p1/p1pPpP1p/P1P4P/1PQ3R1/1K2B1P1/6R1 b - - 0 1
Analysis by Stockfish 190814 syz: (after depth 42)
1. +/= (0.61): 116...Bxf5 117.Ka3 Qf6 118.Rg5 e4 119.Qxf6 Rxf6 120.g4 hxg4 121.Bxg4 Bxg4 122.R5xg4 Re8 123.Re1 e3 124.Rg3 Re4 125.Rexe3 Rxh4 126.Kb2 Kg7 127.Kc3 Kf7 128.Rg5 Rf2 129.Rge5 Rh1 130.Re6 Rc1+ 131.Kd3 Rd1+ 132.Ke4 Rf5 133.Re2 Rf6 134.Rxf6+ Kxf6
2. +/= (0.61): 116...Rxf5 117.Re3 Rf4 118.Rxe5 Rxh4 119.Bd3 Kh6 120.Rge1 Qf6 121.g3 Rd4 122.Ka3 Rf7 123.Qc2 Kg7 124.Qb1 Rg4 125.R5e3 Qc3 126.Bc2 Qf6 127.Be4 Re7 128.Qc1 Qd4 129.Rd1 Qf6 130.Bc2 Rf7 131.Rde1

Also note that when Tord says:
SF was literally two moves away from breaking open the kingside by Re3 followed by g4, a maneuver it had slowly been preparing over the last dozen moves or so.
it is not so easy to verify that with Stockfish 190814 syz, unless perhaps going to higher depths:

2b2r2/5r1k/1p1q1pp1/p1pPp2p/P1P1P2P/1PQ3R1/1K2B1P1/6R1 b - - 0 1
Analysis by Stockfish 190814 syz: (after depth 38)
1. +/= (0.61): 115...f5 116.exf5 Bxf5 117.Ka3
2. +/- (0.78): 115...Rh8 116.Qe1 Rg7 117.Rf3 Qe7 118.Re3 Bd7 119.g4 hxg4 120.Bxg4 f5 121.Bf3 f4 122.Re2 Kg8 123.Rg5 Rhh7 124.Reg2 Qf6 125.Qg1 Be8 126.Qf2 Kf8 127.Rg1 Bd7 128.Ka2 Bc8 129.Kb1

[bob]

The titled-player commentators are calling out computer engine programmers saying that they have taken too much credit in their creations' strength over humans when it's mostly been the inexorable progress of hardware speed.

That's simply false, but what do those guys know about chess...

I suspect that there is more right about their statement than wrong. That is, I believe that well beyond 50% of the improvements over the past 20 years has been hardware related. I REALLY want to take my old Cray Blitz source, circa 1989, and run it on today's hardware and test it on my cluster to see where it comes in. It had no reductions or forward pruning other than null-move R=1. Only work I need to do is add basic xboard protocol support as it was not xboard compatible. But it used the same "seaboard" command as Crafty so that is done, and it inputs and outputs SAN moves. All I really need is time/otim and it should probably work on my cluster.

I remember running Cray Blitz (fortran-only code of course) on a Pentium 133mhz box and hit about 100 nodes per second. I got this old version to compile cleanly a couple of years ago but don't remember the speed on more modern hardware. 64 bit hardware should be even better...

Be an interesting test.