That's one way of expressing an interesting finding noted by IM and computer science professor Ken Regan. In brief, the finding is that players with a slight - even a very slight - advantage tend to blunder much less than those with a slight - even a very slight - disadvantage. Interesting, and possibly practical.
Entries in Ken Regan (15)
We've already reported on this event several times, but now that it's officially over we can put it to bed: Komodo 9.3 is the winner of the eighth season of the TCEC (Top Chess Engine Championship), winning with surprising ease against Stockfish (021115), 53.5-46.5. There were, as usual, a LOT of draw - 89 of them - which means that in terms of decisive games Komodo beat Stockfish 9-2.
Komodo's rating (and Stockfish's too, on TCEC) is above 3200. This is obviously well beyond the range of even the finest carbon-based chess players in the world, but how would Komodo fare against, say, a 32-piece tablebase? (I would say "against God", but divine omniscience extends beyond the ability to play "perfect" chess - God could choose among the equally "perfect" options those which would be hardest for Komodo to handle.) One might think that such a question is not only unanswerable with any specificity, but unanswerable period. Not being able to play perfectly ourselves, who knows?
I'm not sure that Ken Regan knows, but he has offered some of his musing on the topic, and he suspects that the rating of a perfect player may be around 3600. We'll see!
It was live a few days ago, but in case you didn't get the chance then and want to see it now, here you go: this link will take you straight to his spot at the conference.
Ken Regan, whose work has been mentioned and discussed on this blog many times (and in many other places as well), will be giving a TEDx talk live in a little more than half an hour from now, at 10:40 a.m. ET. The primary focus of Regan's work has been to develop a sophisticated algorithm to help detect chess cheaters, and as an added side benefit it can be used to evaluate players' performances in various interesting ways. (One surprising result is that contrary to popular belief, the FIDE ratings have not been inflating over the years but have been rather stable.)
Ken Regan needs no introduction to long-time readers of this blog. He is an International Master and an associate professor in computer science professor at SUNY Buffalo, and is best known nowadays within the chess community for his work on "intrinsic performance ratings" or IPRs. His work has been mentioned regularly on this blog, and he has also been a frequent commentator here and has guest posted once or twice. (He's very welcome to do so again, too!)
If you haven't noticed him or his work here or elsewhere, or would like to know a bit more about this interesting character, Howard Goldowsky has written a long and very worthwhile feature article profiling him for the June issue of Chess Life; happily, for those who don't subscribe to that magazine, that article is also available online. Do have a look!
Still another post by one of our favorite and most valued commenters, on the issue of cheating in chess and how to fairly and properly nab the bad guys.
I was away all day on a coaching trip, but IM and computer scientist Ken Regan very graciously agreed to fill in for me today. His comments on the round - especially on the game between Magnus Carlsen and Teimour Radjabov follow, and then below that I'll add a few comments tidying everything up. I hope everyone enjoys his commentary and his further remarks to Carlsen-Radjabov in the replayable games section, linked to below. Ken has been a real boon to both this blog and its blogger alike, and it's a privilege to have him help out. Thanks, Ken!
The scoreboard says four draws, so Levon Aronian and Magnus Carlsen maintain their 1.5-point lead on the rest of the field. Of course the concept of "joint leader" in this event is ultimately hollow, for only one can emerge as the challenger for the world title held by Viswanathan Anand.
What the scoreboard doesn't say is that Aronian could have had the lead, even a full-point lead, all to himself at this halfway point. Aronian verged on domination before Alexander Grischuk simplified enough to reach a draw, while the favorite Carlsen clipped several gates on his way to a nervous draw with White against Teimour Radjabov. At which gates could he have been taken for a tumble? That's the main question to engage us with our own chess boards---and computers and/or minds, pick'em---after the play is over.
While thinking of themes for this guest post even before the clocks started, I mused on the idea that in every window when you are following the action on ICC or Playchess or Chessdom or other sites, at least four separate games are going on. They are:
1. the game being played by the human White, 2. the game being played by the human Black, 3. the game being played by the computer in the analysis window, and 4. the game being played by you while viewing.
There also may be the game as seen by a live commentator. I was following on Chessdom, where Canadian IM Aman Hambleton supplied comments on Carlsen-Radjabov in excellent prose that explained strategy and variations at perfect picth for the audience. I myself had time only to throw in one comment in the separate chat window, before leaving to attend an early afternoon concert. But this is enough to illustrate the differences among all four different games.
In all four games, Carlsen played the early Rossolimo Sicilian trade 1.e4 c5 2. Nf3 Nc6 3. Bb5 e6 4. Bxc6 used by Anand in last year's match against Boris Gelfand, whose complexity gives much latitude for outplaying the opponent in a fashion Carlsen has trademarked. Radjabov allowed 7. e5 and 8. exd6 splitting Black's pawns, but seems to have done his homework, as he mobilized his center and pointed his pieces Kingside in a forthright manner. Then came a phase that struck me in "my" game as nervous: 12..Be6, 13...Qe7 (awkwardly defending c5 under fire from the Rook on e1), 14...Rad8, 15...Bc8 (going back), 16...f4 (again forced to defend c5 and weakening), and 17...Kh8. I thought Carlsen was all set up to win his game, but Radjabov's shuffle made it possible for his game to begin with 18...Bg4!
Carlsen admitted in the press conference that he underestimated the frank strategy 19...Bxf3! 20. Qxf3 Nh4 21. Qe4 f3 22. g3 Ng2!, partly under the hallucination that 23. Rd1 Qe6 24. Kh1 Qh3 25. Rg1 threatened 26. Rxg2 protected by the Queen on e4. This makes me wonder what game he was playing---but I myself have "seen" brilliant combos ending with Ng6+ intending to meet ...Kg8 by Bh7+. The Houdini analysis window was less nervous---it was showing evaluations consistently near 0 (even game) right up until suddenly agreeging with Hambleton and liking Black after 21...f3 was played. Meanwhile, however, I was fixated on the problem of Black's blocked and lifeless Bishop on d6, even making a crack comment that it was just a big pawn.
How does all this come together for the game? After Carlsen's decision to jettison the Exchange and bring the other Knight back for defense with 23.Nc3, Radjabov concentrated on a strategy to further the Kingside pressure and open up that side with 23...Qe6(?!) 24.Re3 Nxe3 25.fxe3 f2+(?). He was quite set in the conference that he needed to do this to avoid a blockade with Nd1-f2 and felt he should win---underestimating Carlsen's later retreat 29.Qh1! which kept everything covered. But I (and Houdini) think the best chance was missed at move 23: 23...Nxe1 24.Qxe1 Bc7! to follow with 25...Ba5 and trade the awful Bishop off! Then I think White will just be flattened by the pressure and pawn on f3.
Neither player noted this, and the only place I've found it mentioned so far is in comments at http://www.chessgames.com/perl/chessgame?gid=1713445&kpage=20, player DcGentle. After 29.Qh1 I think White is holding pretty comfortably, but perhaps others will find tries for Black.
[UPDATE: Ken has added remarks on the other games as well, and the replayable games now include comments to all four games.]
Levon Aronian's game was another case of differing perceptions. I thought Black was positionally awful around Move 19, but Black does have a Kingside activity plan, and Aronian as White felt obliged to blow the center open first with a pawn sac. I don't think it gave chances as strong as some other annotators seem to believe, however. Gelfand's game with Kramnik featured a double mistake at Move 18-19, and I haven't yet looked for the explanation in what resumed being a fairly chanceless draw. Ivanchuk-Svidler may be important for Scotch Opening theory but stayed dynamically balanced.
Back to DM. Here are the standings after round 7 - the end of the first of two cycles:
1-2. Carlsen, Aronian 5
3-4. Kramnik, Svidler 3.5
5-6. Grischuk, Radjabov 3
7-8. Ivanchuk, Gelfand 2.5
Next up, the round 8 pairings:
- Carlsen - Aronian (the battle of the leaders)
- Radjabov - Gelfand
- Grischuk - Ivanchuk
- Kramnik - Svidler
About a week and a half ago, I mentioned on these cyberpages the remarkable play of one Borislav Ivanov in the Zadar Open. Despite a rating of 2227, Ivanov's managed a spectacular tournament performance rating (TPR) of 2697, and this occasioned a partial search of his person and his pen (both were negative) and a long list of accusers on the web ready to proclaim his guilt.
My preference was to wait for the work of a sane, qualified researcher before drawing any potentially libelous conclusions - the work of Ken Regan, to be specific. He has weighed in, and has offered some thoughtful remarks (and invited more from others) on the more general topic of cheating in chess. (Be sure to check out his letter to the Association of Chess Professionals (ACP) Board for more on the matter, and - for those with the relevant technical background - the appendix attached thereto.)