I have a love-hate relationship with my computer. Analyzing with the computer is a great way to get discouraged and to make unrealistic assessments of positions. (You think that the position is good because the computer says so — but then it turns out that you have to make five “computer moves” in a row, or else your position sucks.)

Nevertheless, every now and then the computer will come up with something that completely blows your mind. Here is a position I was looking at with the computer yesterday. It could have arisen in a game I played six years ago against Jeff Mallett, if I (as White) had played correctly.

White to play.

This is what the position would have been after Black’s move 19 … Qc7. Obviously Black is in a lot of trouble due to his exposed king and White’s superior development, but how does White prove his advantage?

Here Rybka came up with a move that completely dumbfounded me:

20. Na5!!

Black to move.

My first thought was, “Wait, doesn’t this hang the knight in two different ways, and hang a pawn too?”

Well, one of the captures on a5 is obviously bad: If 20. … Nxa5? 21. Rd7+ wins the queen and even though the material is technically still equal, White must be much better.

But what is wrong with the other capture, 20. … Qxa5? If you think White’s first move is obvious, look again. The title of this blog post might serve as a hint.

Now, once you’ve solved that, you can go on to the next question. What if Black snaps off the pawn instead with 20. … Bxc5? This looks as if it could be okay for Black, as 21. Qxc5 is met by 21. … Qxa5. If you’re having trouble, think of Lasker-Thomas again. However, you’ll find it a little bit more challenging this time.

Answers are below. However, to keep you from looking at them, let me just chat a little bit about why I was looking at this game. Recently Brian Wall wrote in his discussion list about a wonderful win he had in the last round of the Colorado Closed championship, where he played a speculative rook sacrifice on e6 against Damian Nash’s Caro-Kann. His move 13. Rxe6! reminded me of this game I played against Jeff Mallett several years ago, which also featured 16. Rxe6+! in a Caro-Kann. It seems to be one of the occupational hazards for a Caro player that your opponents will sometimes smash you with sacrifices on e6.

For several reasons my game is not quite as good an example of this theme as Brian’s. First of all, I didn’t win — it ended up being a draw because of my frightfully bad endgame technique. And second, it turns out that 16. Rxe6+! in my case was “the wrong sacrifice.”

I had an opportunity to sacrifice the exchange, rather than a whole rook, on the previous move, and it turns out that the exchange sacrifice would have been better. It leads, with optimal play for both sides, to the position illustrated above.

This was a case where I let myself get a little bit too carried away by temptation… sacrificing a whole rook seemed ever so much sexier than sacrificing the exchange. But the irony is that the exchange sac could have led to the piece sac 20. Na5!! and then, possibly, to the queen sacrifices in the quiz! It would have been ten times more brilliant.

And so, with that lead-in, here are the answers to my two questions above.

(1) 20. … Qxa5 21. Qxe6+!! forces a mate in seven: 21. … Kxe6 22. Bb3+ Ke5 23. f4+ Ke4 24. Bd5+ Ke3 25. Rd3+ Kd2 26. Bf3+ Ke1 27. Rd1 mate. Isn’t that exquisite? Just like the famous Lasker-Thomas game, Black’s king is hounded all the way to White’s back rank.

(2) If 20. … Bxc5 21. Qxe6+!! still wins, but it’s a little bit more complicated. Black can decline with 21. … Kf8, but after 22. Qxf5+ White is obviously way on top. After 21. … Kxe6 the mate is trickier than in line (1) because White’s f-pawn is pinned. However, this is compensated by the fact that White still has his knight, which is also a very strong attacker. Here is the proof:

20. … Bxc5 21. Qxe6+!! Kxe6 22. Bb3+ Ke5 23. Nc4+! and:

(a) 23. … Ke4 24. Nd2+ Ke5 25. Nf3+ Ke4 26. Bd5 mate; or

(b) 23. … Ke6 24. Ne3+ (mission accomplished, the pin is now broken) Ke5 25. Rd5+ Ke6 26. Rd3+ (second mission accomplished, the knight is now defended) Ke5 27. f4+ Ke4 28. Bd5 mate.

I just love how all of White’s pieces work together. This last line looks like a composition, with the way that White uses the discovered checks (24. Ne3+ and 26. Rd3+) to position his pieces in just the right places.

And they said (30 years ago) that computers would never be able to play beautiful chess! Ironically, for Rybka working out these checkmates is a piece of cake. It takes about a millisecond to find 21. Qxe6!!