Kasparov vs. Lomonosov

Through the years the internet communication of the Tata Steel tournament in Wijk aan Zee has gradually been modernized, by which we have gained a few things and lost some.

There is expert live commentary on the website now, which is undoubtedly a big gain. There is an abundance of short videos in which the players deliver their sound-bites; some of them are quite good at it. The tournament is all over the social networks and those who are not easily tired of endless scrolling can find some interesting tidbits of information there, as you will see later.

Lost are the round reports with technical analysis of the games, and the broadcasts of the sessions – which had already been cancelled many years ago – where the winner of a beautiful game would explain it to the journalists on an old-fashioned big wooden demonstration board. 

In fact this year the staff of the tournament, enthusiastic chess players all, had strict instructions from higher powers at the Tata Company to minimize the technical chess content. Any player interviewed for a video who would threaten to mention an interesting line should ruthlessly be cut off. As Bent Larsen used to say and write: “Long variation, wrong variation,” but here the interdiction applied also to short variations.

Press officer Tom Bottema told me that the idea was that internet watchers who were interested in the analytical truth of a game could find that with their computers. Maybe that’s often true, but not always and anyway, it’s not the same as receiving wisdom from the horse’s mouth.

In 1996 Vassily Ivanchuk won the Wijk aan Zee tournament. That was still the time when a winner of a great game came to the big demonstration board in the pressroom. The sessions with Ivanchuk were unforgettable.

He would show a variation and while he was doing that, often a new insight would dawn on him. At such moments, he would be silent for many minutes, while he was watching the ceiling. Not only Ivanchuk, but also the journalists were dead silent, realizing that the Big Think should not be disturbed. 

After the long silent think, there would come a move, and when Ivanchuk would execute this move on the demonstration board, it was as if a priest was leading us in a religious ceremony. 

If a chess engine were to spit out the same move, would it be the same? 

Computers lack the sacral element, but what about their technical proficiency? If you say “According to the computer...” properly, it should be mentioned what engine it was, on what kind of computer it was running and most importantly, how much time the engine was given to come to its verdict. 

The “live games” section of the Tata tournament website had the formidable Stockfish engine running. It would evaluate a position, but then another move would be made by one of the players, and then the previous tentative evaluation would not be amended, but stand for all time in the great junkyard of the internet archives. Also Stockfish may have been handicapped because it had to comment on fourteen games simultaneously.

This is Ivanchuk-Jobava, from the first round of Tata, after White’s 42nd move. Black can draw by 42...gxf5 43.gxf5 Rf4 44.f6 Rf3, but he played 42...Ra4 and after 43.Rxa4 he resigned immediately.

At this point the tournament’s Stockfish gave the evaluation 0.00, which gave rise to a flurry of excited human comments. Had Jobava resigned in a drawn position? 

He had not and Ivanchuk gave an explanation that was much more understandable to humans than the tree of computer variations. After 44...bxa4, White would stop the a-pawn with his king and by playing e5-e6 he would force a free passed f-pawn that, together with his c-pawn, would guarantee that one of them would promote. “Simple for a good grandmaster,” Ivanchuk said, by which I think he meant that it was normal that Jobava would have resigned immediately. 

No so simple though for a strong engine working on my budget computer.

I fed the position to my Stockfish. After 43.Rxa4, it took about two minutes before it came to a significant positive evaluation for White, about +2.00. Three minutes later it gave a realistic evaluation, +8.00, which stands for an overwhelming advantage. 

Later there was this:

Aronian-Jobava, 9th round, after Black’s 18th move.

Obviously White has a very strong attack in which all his pieces can take part. It didn’t take long for Aronian to force resignation. But after the game, Tom Bottema told me that at this point the tournament website’s Stockfish had given an evaluation of -1.5 to the position, which normally can be interpreted as “Black is winning.” I suppose Stockfish came to its senses quickly. 

There are limits to the performance of even the best engines, especially when they are given very little time, but as we all know, much more common are the cases where they reign supreme and humans are dumbfounded. Here is an interesting example:

Giri-Ivanchuk, 7th round, after Black’s 89th move.

The engines had evaluated the game for a long time as drawish and had become jubilant about White’s prospects only after Black’s 85...Qd2. Don’t ask me why that was such a bad move.

To the diagrammed position my Stockfish gives an evaluation of +30.55 and while the game was in progress, apparently an evaluation of +85 was given, which makes no real difference.

But when Giri, instead of the winning move 90.Qd6+, played 90.Qd4, the evaluations (then and now) dropped to around +1.

A very interesting case, to which Kasparov devoted three tweets: 

“90.Qd6 Kg8 91.Qe5!! looks ridiculous to a human, but it’s the only winning move. 91...Qc1+ 92.Qf4 Qc2 93.Qh4+ is the fastest win. Amazing.”

And also:

“In pre-computer era the Giri-Ivanchuk queen endgame would have been assumed obviously drawn with no analysis needed. But 90.Qd6+ wins. 

And this:

“The winning moves are pure computer. Impossible to imagine zugzwang theme in these positions. White queen goes to f4 3 times & e5 twice.” 

I could not figure out where the zugzwang theme would occur, but anyway, the case seemed settled: 90. Qd6+ would have been winning and with Giri’s 90.Qd4 he gave away the win. But maybe he didn’t. 

I found the Kasparov quotes on a website that I often visit:http://streathambrixtonchess.blogspot.nl/2015/01/revelation.html. In that article there is, in passing, also a link given to another tweet, one by the Dutch IM Twan Burg (called “this chap” on the Streatham & Brixton website):

“90.Qd4 still mate in 50!" 

Could Burg be right, in contradiction to Kasparov, Stockfish and all the other engines that I consulted? At first the round number 50 suggested to me that he might be joking, but later I found a subsequent tweet by him: 

“92.Kf3! 48 moves left until mate.”

It dawned on me that Burg, who is a strong correspondence player, would probably have access to the Lomonosov 7-man tablebases in which case his statements wouldn’t be speculative, but simply true.

Access to Lomonosov comes with the Aquarium software, which I don’t possess. I played with the thought of downloading the free program FinalGen, which would be able to analyze this queen endgame to perfection. Reportedly, FinalGen analyses in general take a lot of time, but in this case, with only seven pieces on the board, that would probably not be prohibitive.

It would be interesting to see who was right, Kasparov and his engines or Burg, possibly supported by Lomonosov.

Click to play through the games...

Ivanchuk-Jobava, Tata Steel 2015

Aronian-Jobava, Tata Steel 2015

Giri-Ivanchuk, Tata Steel 2015