Event horizon

Lately my chess calculation breaks down in the range of six to eight half-moves. Either I overlook important branches at that point, or I simply cut off the line. Case in point: last fall in a game against Avraam Pisminneyy, I calculated eight forcing half-moves and then rejected a line, when in fact mate was forced in two more half-moves. My “event horizon” manifested itself again tonight, when I made a couple of crucial errors in calculation right in that same range.

Slater (2035) – Shmelov (2391)

1.e4 e6 2.d4 d5 3.Nc3 Nf6 4.Bg5 Be7 5.e5 Nfd7 6.Bxe7 Qxe7 7.f4 a6 8.Nf3 c5 9.dxc5 Nc6 10.Qd2 Qxc5 11.0-0-0 b5 12.Ne2 b4 13.Ned4 a5 14.f5 Nxd4 15.Nxd4 Qe7

Here’s the first calculation error. I rejected 16.Nc6 because of the strange line 16…Qh4 17.g3 Qh6 18.Qxh6 gxf6 and now he’ll kick my knight and win my e5 pawn. Overlooked: The incredibly simple 19.Re1. After the game Denys said black’s prospects are dim in that position. (He thought 16…Qh4 was ridiculous and was planning 16…Qc5 with a probable repetition because he didn’t like 17.Nd4 Qb6 18.Qg5.)

Second error: After 16.fxe6 fxe6 17.Nf5 exf5 18.Qxd5 Rb8

shmelov21.jpgNow 19.Bb5 is better but I played 19.e6. Great fun is 19.e6 Nc5 20.Bb5+! Rxb5 21.Qc6+ (splat!) and I believed black also to be busted after 19.e6 Nf6 20.Bb5+ Kf8 21.Qd8+ Ne8 22.Bxe8 Qxe8 23.Qd6+ forking the rook on b8. What I missed was instead 22…Qxd8 23.Rxd8 Ke7 and the best white can do is 24.Rxc8 Rxc8 25.Bf7, which we agreed after the game is probably lost for White.

So after 19.e6 Nf6 20.Bb5+ Kf8, I discovered that defensive resource and varied with 21.Qxf5, just trying to further denude his king, but he calmly untangled and retained his winning material advantage.

So here’s the point. It’s easy and formulaic to say “I need to get better tactically,” but more helpful to be more specific. In my tag post I mentioned Kotov’s calculation exercise. The analysis tree method is particularly helpful for measuring and stretching your event horizon. Knowing the average depth (i.e. how many moves into the calculations) at which I’m making tactical errors or cutting off my analysis, I can do the exercise with a very specific task: “Must calculate all branches out to ten half-moves”. If through repetition I manage to start hitting that mark with reasonable breadth and accuracy, I can push on to twelve half-moves. Try it.


12 thoughts on “Event horizon

  1. The psionic flow was definitely happening last night at the club. The first 11 moves… identical to my game ( as you commented).

    I typically play 12.Bd3 first before to get it out before retreatign the knight. A typical continuation : 12 Bd3 b4 13 Ne2 a5 14 Kb1 etc… then I focus on trying to exchange queens.

    Last night I took a risk and played the immediate 12 Nd4 to throw Mark off his memory game. It worked. ;)


  2. Hey George – see, this is part of my problem. I never, ever, ever ever EVER try to exchange queens. Regardless of positional considerations. So I can’t play your line :)

  3. Ah but in this line of the french… it is critical… Exchange …purge…clean out the queen side… especially that pesky Queen.

    Its a winable endagame as long as you have the stronger light squared bishop, more active rooks to the queen side and the initiative. Black’s light squared Bishop remains imprisoned in the tomb of the advcancd center pawns, the King Rook is still back in the corner ( unless they make the mistake of castling king side) Utilize teh pawn majority on the queen side to your advantage. Keep the center locked and exhcnage knights along with the queen. Your pawns are on dark squares and Black’s are on light squares…. I like it.

    But look at me… I’m talking “as if” I know soemthing. ;)

  4. Yes but you play the Caro Kann. You LIKE positional chess.

    But now that you have mastered the 7 Circles and conquered de la Maza’s terrain and memorized all of CT-ART, I believe that you should start using your hard-won tactical acumen for good and not for evil, i.e. start wheeling out some crazy openings and throwing down some monster sac attacks.


  5. Derek,

    I’m wondering if what you call an “event-horizon problem” is really a pattern recognition problem. It seems like you got to a ‘plateau’ in your calculations, looked around at the scene, and decided there wasn’t anything worth pursuing. Now, I would argue that most GMs only look ahead 8-10 ply, at most, anyway. But they have better pattern recognition than you or I have. These patterns are either tactical or positional in nature. The position you gave up on after 19.Re1 seems, to me, an instance of a family of positions you might consider doing more work on. Ask yourself why you gave up on that position. Instead of training to look deeper into variations, take this position before 19.Re1, break it down into its tactical and positional motifs, and then study these weaknesses of your ‘personal evaluation algorithm.’


  6. Howard – Couple of thoughts.

    – Pattern recognition… this may be a matter of semantics. If you were to look at the Pisminneyy position I mentioned, you could argue that was a pattern failure. More simply a failure of common sense; any idiot (except me apparently) would look at the analysis position and say “His king’s in the middle of a crowded board! Play it!” But in this Shmelov game, I wouldn’t call Re1 a pattern issue. If you showed me the position on the board I’d pick out Re1 in one second, as would most anybody over 1400. So I’d still characterize this as a problem with lacking a clear vision of that position in my analysis, which is something you can clear up via practice. I think your method of examining positions and families to identify misunderstandings is a great method, but not the problem in this particular case.

    – GMs, 10 ply at most? In nontactical positions that’s probably the case but in sharp positions, I think it’s categorically false. My experience in postmortems with people 2350+ is that they have seen really long lines, and lots of them. GMs may skip some of that because of pattern recognition but the analysis they publish does quite frequently say “At this point I anticipated XYZ” and the variation is many moves long.

  7. I see your point. What’s still interesting is how calculation and evaluation intersect at higher ply. For instance, if you take a computer and limit it to two ply, then its strength is purely on evaluation. If you concoct a machine that calculates 100-ply, with zero evaluation (the epitome being a Deep-Blue-like computer), then you get, perhaps, something stronger than the first case; however, as Rybka proves, evaluation is very important when your calculating abilitiees are limited.

    Perhaps some GMs calculate past 10 ply, but probably only for lines where their evaluation proves promising. Certainly they don’t do this for every variation. That’s impossible except for the very fastest computers.

    If you’re saying that you need to extend your event horizon for promising lines, I can see where this would help, for example with 19.Re1 in your game.


  8. Seeing some of those long non-forcing variations GM’s calculate boggles my mind. But it may be more that I don’t understand that some of the moves are practically forced.

    Then again, in Andy Soltis’ The Inner Game of Chess, he refers to a quote by Botvinnik after he lost a game. It was something to the effect of “It shows that I still need to work on perfecting two move calculations.” I suppose even for GM’s, many positions can’t be calculated very deeply, and Derek’s favorite “+.001 chess” comes more into play.

  9. No question that learning this balance between calculation and evaluation is both difficult and critical. It’s also a bit idiosyncratic, i.e. how two GMs handle the same position may be quite different. When Tal found a tactical position or a sacrifice where he couldn’t calculate the outcome, he’d play it! Because he figured if he couldn’t find all the answers, the other guy couldn’t either & Tal was confident in his ability to outplay his opponent.

    Once in college my friend Glenn Etter walked by my board, looked at the murky Benoni position for about one second, and went back to his game. Afterwards he told me he didn’t bother to analyze any lines because he said, “I saw that it was a mess and you always win messes.” (Ah, the good old days!)

Comments are closed.