Sunday, August 24, 2014

What is Advanced Chess ?

          Unless  you’ve been living in a mountain cave or on a remote island for the last twenty years, you already know advanced chess allows players to use chess engines to assist them in selecting moves in live games. Those who promote advanced chess apparently feel the need to justify it, even going so far as to claim advanced chess titles and ratings are comparable to titles and ratings awarded prior to 1980 (i.e., before computers). They shroud “centaur” chess in mystique, with solemn pronouncements such as those given below. Some of these assertions are true, others specious; all should be viewed in proper perspective.

            “Advanced chess is a new art form, creating the highest level

            of chess play possible.”

           In a recent article in CCLA’s Chess Correspondent magazine (Vol. 87, No. 1, Jan-Mar 2014 issue), IM Wolff Morrow discussed several high-level games and his approach to correspondence chess play. Analyzing his own game vs. David White, he properly credits the Kruger-Svacek game (ICCF WC31/ct06) for the “hot new refutation of 18.Bb3” in the Sicilian Sveshnikov line, giving Svacek’s 18… f4 an “!” with no explanation. Now 18… f4 is one of those “surprise” moves that doesn’t become top recommended move until many, many plies have been exhausted by an engine. For example, my older Stockfish 64-bit engine has reached 35 ply but still doesn’t rank 18… f4 as best move:

35 [+0.34]  18.... Be5 19.f4 Nxf4 20.g3 Bxc3 21.bxc3 Qb6+ 22.Kh1 e3 23.Qd4 Qxd4 24.cxd4 Ne2 25.Bc4 Nxd4 26.Bd3 Rfe8 27.Rae1 Rab8 28.Rf4 Nb5 29.Kg2 Kg7 30.Rxf5 h6 31.Rf3 Nc3.

Downloading and running the new Stockfish 5, I was “close but no cigar” – my system bogged down at 38 ply, still showing advantage white:

38 [+0.17]  18.... Qf6 19.f3 Qd4+ 20.Qxd4 Bxd4+ 21.Kh1 Bxc3 22.bxc3 a5 23.a4 Kg7
24.c4 Rae8 25.c5 dxc5 26.Bc4 e3 27.d6 f4 28.g3 Rd8 29.Rad1 Rfe8 30.Rfe1 Ne5 31.Be2.

These analyses took about four hours each on my x64, dual core machine with 3 GB ram, admittedly not a state-of-the-art computer. A colleague who ran the analysis on an i7 processor (4 cores) with 8 GB ram had better success, finding the elusive 18… f4 (and a tiny edge for black) with Stockfish 5 x64 at 39 ply:

39 [-0.10]  18...f4 19.Nxe4 f5 20.Nc3 f3 21.g3 f4 22.Kh1 Qd7 23.Rg1 fxg3 24.Rxg3 Be5 25.Qg5 Bxg3 26.Qxg3 Rf4 27.Re1 Rg4 28.Qxf3 Rf8 29.Qe3 Nf4 30.Ne4 Rg7 31.c3.

18 … f4 is an example of “bread crumbs”, clever use of “if” moves to lead an unwary opponent into disaster. In the old (pre-1980) days, postal players who sent losing “if” moves were roundly condemned. In advanced chess, this practice now means “guiding (the opponent’s) engine.” Once a player determines his opponent is only analyzing to a certain ply depth, he uses his machine’s superior capabilities to analyze deeper, looking for “surprise” moves beyond the horizon of his opponent’s system. When he finds such a move, he can offer his opponent “if” moves, ostensibly saving time on “obvious” replies. This technique is possible because the “if” move offered will be a top selection on the opponent’s inferior computer. Even if the player checks it out, his engine will conclude it’s the best move, so why not accept it? In Kruger-Svacek, after 16… 0-0, Black likely sent the sequence “if 17. Qd2 Ng6”.  White accepts this and plays his engine’s top move, 18. Bb3, expecting 18… Be5 with a slight edge for himself, as shown above at 35 ply. Instead, he received the surprise move 18… f4! Often the evaluations won’t change dramatically after such a move, but sooner or later the player’s engine will start reporting bad numbers and his game is inexplicably lost.

           ICCF players have been known to enter lengthy “if” move sequences into the server, hoping to lead an unwary opponent into one of those “surprise” (aka “trap”) positions. Small wonder players lobbied for an “if” move feature to be added to the ICCF server a few years ago. The uninitiated wondered why a game (whose moves were already transmitted at lightning speed) needed a “speed up” with “if” move capability. To borrow a phrase from the late, great Paul Harvey, now you know “the rest of the story.”

When reading about a move like 18… f4, wouldn’t it be nice to know how deeply the position was analyzed, which version of a chess engine was used and on what hardware it was run? No player will ever reveal such details! Advanced chess players are careful not to shoot themselves in the foot, tipping-off future opponents with such damaging disclosures.

Is there any sport, or art form for that matter, where the competitors’ equipment isn’t subject to rules and regulations, open to inspection by officials? Think “corked” baseball bats, chemicals on boxing gloves, fines for miniscule engine violations in NASCAR, illegal use of copyrighted photographs as the basis for paintings – the list goes on and on. Sports team owners are reined in by salary caps, preventing multi-millionaires from amassing super athletes into unbeatable teams. Competitors and fans alike expect a level playing field.

There’s no way to level the playing field in advanced chess as currently practiced, because everything is done in secret. It is debatable whether systematic if-move trickery constitutes a “new art form”. Not long ago, a bitter rival stated the world correspondence championship had been won by a Dutch millionaire with a “basement full of processors.” Advanced chess is all about ply depth, and that equates to being wealthy enough to afford some very expensive hardware. Advanced chess is a game for the affluent, not the masses.

            “It’s much more difficult to win an advanced chess game.”

            It must be, because over 90% of games in ICCF’s current world championship are draws. Historically, in traditional chess white wins about 40% of games, black wins 35% and draws are 25%. Percent of draws in this event has risen at an alarming rate since introduction of high-speed, multi-core processors and “deep” versions of chess engines a decade ago. For background and discussion of the fast-approaching draw death of advanced chess, see the blog posting:

and the statistical summary of all the ICCF world championships:

History indicates computers will get faster, software will improve and market pressure will drive prices lower.  More and more players will be able to upgrade their hardware and software. This “trickle-down” effect will spread the draw death to the rank and file. According to CCLA's Server TD, boredom is the number one reason advanced chess players quit. I wish boredom was the only problem!

           “There’s no way to detect when correspondence chess players

           are using a chess engine”

           Sadly, some engine addicts apparently believe this. Unable or unwilling to compete
in the arms race, they have turned to non-engine events where naïve opponents (and un-enforced rules) make for easy pickings. Some even have the chutzpah to annotate and publish their fraudulent wins on web pages and in print.

Per above, the statistical detection system developed by Dr. Kenneth Regan, University at Buffalo, has been accepted by FIDE to help police over-the-board events.

            Advanced chess relies on the engine’s calculation but human
            genius is required to guide the engine’s analysis.”

            It’s difficult to accept this notion when games from the ICCF world championships are analyzed and 100% of the moves match an engine’s top recommendations. For that matter, so do many games among untitled players. How can this be if players rely on their own “genius” to select moves? After a “book” opening, some of the moves played should differ from engines’ top recommendations, but they rarely do. Here’s a typical example:

[Event "26th CC World Ch Final"]

[Site "ICCF"]

[Date "2010.06.10"]

[Round "?"]

[White "Perevertkin, Vladimir Viktorovich"]

[Black "Jaulneau, Christophe"]

[Result "1-0"]

[WhiteElo "2614"]

[BlackElo "2551"]

[ECO "D43"]

1. d4 d5 2. c4 c6 3. Nf3 Nf6 4. Nc3 e6 5. Bg5 h6 6. Bh4 dxc4 7. e4 g5 8. Bg3 b5 9. Be2 Bb7 10. e5 Nd5 11. O-O Nd7 {end of book: 12. Nd2 Mamedyanov-Gelfand, Moscow 2008, 0-1 in 37 moves.} 12. Nxd5! cxd5 13. Ne1 Bg7 14. f4 O-O 15. fxg5 hxg5 16. Nf3 f5 17. h4 Qb6 18. Kh2 gxh4 19. Nxh4 Nb8 20. Bh5 Qd8 21. Qd2 Nc6 22. Ng6 Rf7 23. Bh4 Qa5 24. Qg5 Qc7 25. Rad1 a6 26. Nf4 Bc8 27. Qg6 Ra7 28. Bf6 Rxf6 29. exf6 Qd8 30. Rf3 Qxf6 31. Qe8+ Bf8 32. Qxc6 Bd7 33. Qb6 Ra8 34. Rg3+ Bg7 35. Qd6 Be8 36. Rf1 1-0

            Let me first congratulate the winner on 12.Nxd5!, very clearly a TN and likely to replace the two traditional “book” moves here, 12. Nd2 and 12. a4. But one must be able to run an engine to very great ply depths before it gives up on 12.Nd2 and makes 12.Nxd5 the top move. There’s nothing revolutionary about opening preparation helping to win chess games – chess masters have been unleashing opening surprises on their opponents for 500 years. What is revolutionary is the ability of chess software to find TN’s so deeply hidden that an unassisted human player could never discern them.

            After the first eleven “book” moves, Vladimir Perevertkin’s next 25 moves are all (100%) top engine moves at 30+ ply depth. Christophe Jaulneau’s next 23/24 (96%) moves are also top engine moves. The one move not even on the engine’s horizon is 25… a6.  Since Jaulneau dutifully followed his engine’s recommendations 96% of the time, are we to believe 25 … a6 is one of those “human genius” moves, an artful collaboration of man and machine? 

After this move, despite continuing to play his engine’s top selections, Black’s game goes steadily downhill and he resigns just ten moves later. The engine’s recommendation was 25… a5, (not …a6) and I submit Jaulneau’s move is not genius but a slip of the computer mouse when entering his move on the server. Once a player has sent a move, it cannot be retracted. The ICCF server gives players a last chance to avoid this, but one must notice an error to avoid clicking “send” the second (and final) time.

“When two opponents are both using engines, the better player
           will still win.”

            This would be theoretically true if both players were using identical hardware and software. If parity of equipment could be enforced, advanced chess would be restored to mano a mano competition. As discussed previously, this can’t occur in practice because players are left to their own devices (no pun intended.) There are no restrictions on, or verification of, players’ equipment. Some players extol the virtues of opening research, but their deep traps and surprise moves are only successful because of the disparity in equipment. As it stands now, when two opponents are both using engines, the player with the better computer system will win.

            “If you just play the engine’s top recommendation each move,
            you will lose every game.”

             The implication here is chess engines need guidance by “superior” human intellect. The statement is disingenuous, true only because of the disparity in equipment. At the highest levels of advanced chess, where there is equality of hardware and software, players do play the engine’s top selections nearly 100% of the time but they do not lose every game. In fact, they hardly ever lose, instead drawing over 90% of the time. For players below world championship level, following their engines’ output only gets them in trouble if opponents have better equipment. What is a top move for computer A running to 30 ply depth may well be an inferior or even losing move for computer B analyzing to 40 ply depth, or more. It doesn’t take very many faux top moves to drop a position’s evaluation below the critical range from which no engine can recover, no matter how many top moves its operator continues to play.

            The next time a break-through in computer technology occurs, more speed, more storage, a better chess engine and so forth, the ripple effect will be evident in advanced chess. One player will out-distance the pack and be hailed as the “next Bobby Fischer.” However, as more and more players acquire the new technology, results will level out again. Advanced chess is, and always will be, an “arms race”. As in the Cold War, eventually parity is reached and the “war” becomes a stalemate. Engines are not programmed to win, they’re programmed not to lose, and so the majority of games end in draws.
            Fifty years ago, chess players universally disliked Tigran Petrosian’s style (closed openings; paying more attention to his opponent’s possibilities than his own; keeping the draw in hand.)  Today’s proponents of advanced chess, many having grown up with pc’s and computer games, lack this historical perspective. They don’t realize they’re just “Petrosian clones on ‘speed’.” The coming draw death of advanced chess is an unintended consequence. What amazes me is advocates of advanced chess couldn’t see it coming.


No comments:

Post a Comment