Today I’d like to share my most mind-bendingly complicated position from the Far West Open last weekend. To set the stage, it’s round 5, I have a score of 2-2 and I’m playing a class-A player named Solomon Beilin. I’m going to show you three positions in a row, before my 38th, 39th, and 40th moves respectively. If you want to reconstruct the tournament conditions, set your chess clock to five minutes, because I had five minutes left to make these three moves. If you can get all three moves right in five minutes, congratulations! You are a super-GM, or else you were using a computer. I feel proud to have even gotten ONE of the moves right. It’s that difficult.
White has just played 38. Bb6-c7. What is the best move for Black, and what is your evaluation of the position?
I have just played 38. … Bf8 (using four of my five available minutes) and my opponent has played 39. e5. Resigning is not an option. What is Black’s best move and what is your evaluation?
And finally, our third position:
I played 39. … Bb7 and White played 40. Qxb7. What is Black’s best move, and what is your evaluation?
Are you ready with your answers? Here goes!
I. Black’s best move is 38. … Bg4, and he is winning! But you don’t get any credit for this unless you found the right follow-up. White is, of course, forced to play 39. Qg2. Of course I looked at this — it’s what I spent most of my four minutes on — but I just couldn’t see a followup for Black, and meanwhile because I have abandoned the queenside, the advance b6-b7-b8 is now a huge threat for White. This seemed too reckless, so I reluctantly played the defensive move 38. … Bf8.
After 38. … Bg4 39. Qg2 the solution is 39. … Ne2+!! An unbelievable move found by the computer. I never even considered this move, really. To be precise, I think my thoughts were “It would be nice if I could play 39. … Ne2+, but that just hangs a piece.”
What is the point, then? Let’s consider the possibilities, from easiest to hardest:
(A) 40. Kf1? White timidly declines the piece sac, and Black wins easily with 40. … Nxc1. Notice that Black has time to trade queens and still get his bishop back to c8 to defend the passed pawn.
(B) 40. Nxe2? Here again the win is not too hard to find, but there is a little subtlety on the first move: 40. … Qe3+! (You have to do this first. If you take on d3 first, White wins the g4 bishop.) 41. Kh2 (On either 41. Kf1?? or 41. Kh1?? Black wins the queen with a pin. On 41. Qf2 the queen is no longer attacking g4, and Black can take on d3.) 41. … Bf3 followed by 42. … Qxd3 with a big advantage to Black.
(C) The hard question is what Black does after 40. Bxe2! Qe3+ 41. Kh2. As in line (B), Black has to be careful — if he takes on c1 right away, then he loses his own bishop on g4. The correct move is 41. … Bxe2! 42. Nxe2 d3!! The two exclamation points are because this move is psychologically so hard to see, especially way back on move 38. As a friend of mine, Cliff, said when I showed him this position, “What makes it so hard is that you’re thinking of checkmating White’s king — you’re not thinking about queening the d-pawn.” In fact, the d-pawn is unstoppable here. But so is White’s b-pawn, so actually we should go a little bit farther in our analysis: 43. Ng3 d2 44. b6 d1Q 45. b7 Qg4! 46. b8Q+ Kh7 (diagram)
Position after 46. … Kh7 (analysis)
In this amazing four-queens positions, White has no good defense to the threat of 47. … Qh4+. Notice how secure Black’s king is, tucked in on h7, compared to his counterpart!
As Cliff said, “I HATE computers!” The thing that makes this combination so hard for a human to see is that it is so unclear where the combination is going. First you have to force yourself to look seriously at 38. … Bg4, even though it appears to be abandoning the queenside. Then you have to consider 39. … Ne2+, even though it seems to be losing a piece for no compensation. Finally you have to play the counterintuitive 41. … Bxe2, which seems to help White consolidate, but in fact opens the way for your passed pawn. Unreal.
Okay, now are you ready for the answer to problem 2?
II. This is the move I got right! In fact, 39. … Bb7!! is a miracle save, and Black should draw here. Of course, with just 50 seconds on my clock I couldn’t do any serious analysis; it just seemed as if this gave me a chance to make some mischief and reach the time control. If, after the time control, it turned out that White had an easy defense, then at least I wouldn’t have to suffer for too long.
But actually the piece sacrifice is quite sound. The strategic reasons, really, are the same reasons why the wild piece sacrifice in position I also worked. White’s knight on c1 is vulnerable; without that knight, the bishop on d3 is vulnerable; White’s king is exposed; White’s dark-squared bishop is unable to participate in the defense; and finally there is now an extra wrinkle that 39. … Bb7 lures White’s queen away from the kingside (at least temporarily).
If I couldn’t find the brilliant winning piece sacrifice, I’m glad that at least I found the brilliant drawing piece sacrifice! But the story is not over, because after 39. … Bb7 40. Qxb7 we get to position III.
III. Obviously the knight has to go somewhere, so that Black can threaten … Qe3+, but where? With virtually no time on my clock, I played 40. … Nf5? which should have lost. This is again a position where ordinary logical rules do not apply. I barely even considered 40. … Nh5! because it puts the knight on the rim and doesn’t make any obvious threats, aside from the pawn on f4. But the most important thing is that it keeps the knight on the board. Remember that a queen and knight are a very good attacking combination. If, after, 40. … Nh5, White plays an “ordinary” move such as 41. ed or 41. b6, then Black will draw effortlessly with 41. … Qe3+. If White plays the same way as in the game, moving his king to h2 after a check on e1, Black will have another check with … Qg3+. In the game (see move 44) I didn’t have this move, and therefore I should have lost.
So it is essential for White to prevent … Qe3+, and the only way to do it is 41. Qe4! This on the surface seems to be a strong move, centralizing the queen. Now Black has to find yet another “illogical” move: 41. … Qg4+! (and not 41. … Qg3+). The reason this move is better is that it keeps g3 free for knight checks! In fact, White’s king cannot go to f1. And on 42. Kf2, Black continues 42. … Qh4+! White’s attempt to run to the queenside is completely stymied, as the king cannot go to either f1 or e2 because of knight forks. So he has to come back to g2 or g1, and obviously White has made no progress. Finally, if 42. Qg2 Qd1+ 43. Kh2 then 43. … Nxf4! is the simplest draw, vacating h5 for the queen and incidentally creating some quite nasty threats. In fact, the computer actually shows a small =/+ advantage for Black, although I would be thrilled here to settle for a draw by repetition.
So what happened in the game? After the more “logical” and bad 40. … Nf5? White of course played 41. Bxf5 Qe3+ 42. Kg2 Qd2+ 43. Kg3 Qe1+ 44. Kh2! One more “illogical” thing about this position is that, even though Black is two pieces down, he would really prefer not to take either of White’s pieces that are en prise, because it costs him a tempo. But in this position I have to take back the piece, because I have no perpetual check. Notice that the story would be different if I still had my knight on the board, because I could play … Qg3+. As noted above, that is why 40. … Nh5 was better than 40. … Nf5.
So, reluctantly, I continued 44. … gf. And now we get to a position where, for a change, it’s White who has to find a good move.
White to play and win.
Now actually according to the computer there are a couple winning moves for White, but the one I think is the clearest is:
45. Qf3!! White gives back the second piece; he doesn’t need it. First notice that Black has no perpetuals: if 45. … Qd3+ 46. Kh3 or 45. … Qh4+ 46. Kg2 Black is stopped dead in his tracks. So Black has to take the knight: 45. … Qxc1. (Since when has it been so painful to take loose pieces?) And now after 46. b6! it is now White who has an unstoppable pawn, and White who gets a checkmate after both pawns promote!
Un-friggin-believable.
Fortunately, my opponent didn’t find this resource. I think that he had not yet entered the Twilight Zone mentality that this position required, where ordinary rules of chess are suspended, and one tempo is more valuable than a knight or a bishop. Instead he played the “normal” move 45. Qa8?, prehaps expecting me to defend my bishop or take his knight, but of course I did no such thing. I drew easily with 45. … Qf2+ 46. Kh3 Qe3+ 47. Kg2 Qd2+ 48. Kg3 Qe1+. We are back to the position we were on move 43, but the difference is that White no longer has his bishop and he therefore has no shelter from the checks. After 49. Kh3 Qe3+ we agreed to a draw. What an escape for me!
* * * * * * *
Okay, now let’s take a deep breath. There is one other thing I’d like to talk about in this game, and that has to do more with psychology, or the Anatomy of a Blunder. The fact is that I should have gotten an easy advantage earlier in the game, and should never have gotten into the desperate tactical scrum that we have just seen. After White’s 22nd move, 22. Nd2-b3, we arrived at this position:
Black to move.
I think that any player above 1200, and probably a lot of players below 1200, would have immediately found Black’s best move, which is 22. … R8a8. Black takes control of the a-file, and White cannot really contest it. It’s true that you have to watch out for 23. f5?!, but if you work out the tactics you’ll see that White just loses a pawn and does not really have compensation for it.
Instead I played the goofy 22. … Bg4?, after which Black’s advantage completely evaporates. White played 23. Qc2, of course, and after 23. … R8a8 24. Ra1 he was able to successfully challenge the a-file. Not only that, his queen was better placed and my bishop was worse placed than it was before.
What came over me? There were two interesting psychological factors at work here. First, I was trying to play rapidly in this phase of the game, so that I could avoid my usual time pressure. This experiment failed in two different ways. First, I played several sloppy moves (22. … Bg4? being the sloppiest); and second, I didn’t actually avoid time pressure! Usually I define “time pressure” as less than a minute per move. So in some sense I was successful; as we saw, we got down to a position where I had 5 minutes left for 3 moves, which is not time pressure according to my normal definition. But alas, it was time pressure for that particular position. To navigate moves 38, 39, and 40 correctly, I probably needed at least 30 minutes! (In fact, I think it’s debatable whether I would have ever found the winning combination on move 38, even with 30 minutes.) So this is an interesting lesson: time pressure is always relative to the position. An arbitrary definition like “less than one move per minute” may work most of the time, but it will not work always.
The second psychological factor behind the move 22. … Bg4 was a hallucination. After 23. Qc2 I intended to play 23. … Ra2??? The point was that in my mind, I had already played … R8a8! This is an extremely interesting type of blunder, and I think it’s a fairly common pattern. You analyze one variation (22. … R8a8), and then when you’re done you fail to mentally reset the pieces to their original positions! I mentally kept the rook on a8, and so I thought, “Oh, wow! 22. … Bg4 followed by 23. … Ra2 is a powerful idea!”
Fortunately, of course, I discovered my mistake before playing 23. … Ra2???, but still the damage was done — I no longer had any claim to an advantage.
{ 5 comments… read them below or add one }
Hey Dana,
I follow your blog dilligently and absolutely love training with your test positions and listening to your ideas. I was just wondering how do you come up with such good annotations to your own games? Do you use books, or engines, or just think about it a lot? Do you consult higher rateds on their thoughts? If I could annotate my own games as well as you could, I would be a master by now =)
Thanks, Michael! It’s nice to know that somebody appreciates the effort that goes into these posts.
There is no magic formula for good annotations, but I have some general ideas. First, I go over all my games “by hand” (i.e., without a computer). The importance of this is strongly emphasized by GM Jesse Kraai in his Chess Lectures, and I have been doing it since 1984. In this computer era there is a strong temptation to look for quick solutions by going over your game with the computer first. But you should resist this temptation. First, form you own opinion about all the critical moves and positions in the game. Then, if you want, check your findings with the computer. Here I disagree with Jesse — the computer is a useful “reality check” as long as you have done your own analysis first. Otherwise it’s just a crutch.
On occasion, for this blog, I have to violate my usual rule of analyzing by hand first, then by the computer. To put up narrative or analysis really quickly, I have to use the computer first. But I do this in the knowledge that it will not provide final answers, just a first-pass approximation to the truth.
So anyway, the first answer to your question is that I try very hard to construct my own truth or narrative about the game through uncompromising analysis. For me, a good analysis session is one that reveals at least one thing about the game that I didn’t understand when it was going on.
Second, there is a question of presentation. This part is no longer for my benefit, but for my readers’ benefit. I certainly don’t put all of my games up here. I try to think about whether somebody who doesn’t care about me would still find this game interesting or instructive, and if so, what is the interesting thing about it. So in this game I really focused on moves 38-40 and also the instructive mistake on move 22. That leaves about 90 percent of the game uncovered, but if I wrote a complete analysis of the game I think it would just dilute the message, in this case. (Of course there are other times when I think it’s really valuable to show the whole game.)
I do greatly value comments of stronger players (or even not-stronger players), but it is relatively infrequent that I can work them in. Sometimes, for example, I might do a post-mortem (e.g., in my game with Jesse Kraai) and I can bring some of those insights into the analysis.
Finally, I am a professional writer, and I would certainly like to think that some of my writing skills can be put to good use when I’m writing up my analysis of a chess game.
I hope this isn’t a ridiculously long answer to your question!
Good answer!
Ahhh I see now.
It’s just that sometimes I have to admit that some of the ideas that you find in your annotations are so shocking yet so brilliant that I have a hard time believing a mortal can even find such ideas, never mind having me find those brilliant improvements in my own games. Thanks for the tips.
Well, in this case most of the shocking ideas were found by the computer. I hope I made that sufficiently clear in my post.
When the computer finds moves that I didn’t think were good — or sometimes moves that seemed completely out of the realm of possibility — I do make some effort to understand why. Otherwise I’d go crazy! That’s why Cliff said he hates the computer!
I don’t think we can or should emulate how a computer thinks about chess, but the computer can broaden your mind and make you realize there are more possibilities out there than you realize. If you can see even 10 percent of those possibilities, then you will become a better player.