Although there are no cards in chess, and no “luck of the draw,” trump cards definitely do exist. A trump card in chess is a dangerous threat that you keep in reserve to play at the right moment — often in combination with another trump card.

Here’s a game I played against the computer yesterday that illustrates the principle beautifully. For readers of my earlier posts, A Fascinating Experiment and Matrix Chess… Reloaded, this was another “Matrix chess” game. It’s a blitz game (10 minutes for each side) with one wrinkle: at one time during the game I get to call a time out and analyze the position for as long as I want. The name Matrix chess refers to the “Matrix” movies, where the main character, Neo, has the ability to slow down time as much as he wants. Also, like Neo, I am fighting against a computer!

timeout 1Position after 23. … Rf5. White to move.

FEN: r1b4k/1p4pp/1bn1p3/3qPrN1/p2p4/P2Q1N2/1P3PBP/2RR2K1 w – – 0 24

Shredder was Black, and I had set its strength to 2206 for this game. I sacrificed a pawn in the opening (a Catalan) and, not satisfied with winning one pawn, the computer has moved its queen to d5 and rook to f5 in a misguided attempt to win a second pawn. However, both of these are risky positions: the queen can easily be hit by a discovered attack along the h1-a8 diagonal, and the rook is pinned because of the mate threat on h7. Those threats are my two trump cards. I had a “spidey sense” that I should be able to do something with them here, so I called my timeout.

Before I called the timeout I had two candidate moves: 24. Nxh7 and 24. Bh3, and I thought my choice would come down to one of them. The first move exploits the discovered check motif: if 24. … Kxh7 25. Ng5+ wins the queen. The second, of course, exploits the pinned rook on f5.

However, both of these moves fail on principle and they also fail on concrete analysis. The problem with 24. Nxh7 is that it’s really the queen that wants to go to h7. The knight on that square is out of play and in danger of being trapped. The problem with 24. Bh3 is that I really would rather keep the bishop on g2 where it can attack Black’s queen. In both of these moves I am making important concessions just for the sake of grabbing material. I’m giving away my trump cards. Also, note that in both of the above lines I am only playing one trump card at a time.

Also, both of those moves run into 24. … Nxe5. If 24. Nxh7? Nxe5 25. Nxe5 Qxe5 the knight on h7 is indeed trapped and White loses. And on 24. Bh3? Nxe5 25. Nxe5 Rxg5 tragically comes with check, so I don’t have time to fork the king and rook with Nf7+. Note that the move 24. Bh3? opened the king up to a check on the g-file.

But just when things were looking bleak for me, a third possibility entered my mind. Do you see White’s best move in the diagrammed position, which leads to a large advantage? Actually, you need to find two moves, this one and the next one, and neither of them was my top choice to begin with. That’s why it was so helpful to call a timeout right at this moment.

Hint: What other piece can attack the rook on f5? And what is preventing it from doing so?

The answer is that Nh4 would also attack the rook, but that fails because the knight on g5 hangs. But what if that knight weren’t there any more?

The solution: 24. Nf7+! As I tell my students, always look for checks and captures. Of course if 24. … Rxf7 then 25. Ng5! wins the queen, because of the mate threat on h7. This is what I mean by playing your two trump cards simultaneously. Individually they didn’t do much, but together they are crushing.

The only other possibility for Black is 24. … Kg8, and now White follows up with 25. Nh4! As I mentioned above, this too was not my first choice. The move I wanted to play was 25. Neg5?, which is nice and aggressive, but Black has the stunning answer 25. … Qb3! That move completely takes the air out of White’s attack. If White allows the queen trade, then the knights on g5 and f7 are just cannon fodder for Black’s rook. If White moves the queen, Black gets murderous counterplay with … d3 (or possibly … Qxb2) striking at the f2 square.

But after 25. Nh4! the queen trade defense just loses an exchange or more: 25. … Qb3 26. Qxb3 ab 27. Nxf5 etc. Other queen retreats are no better; for example, 25. … Qd7 26. Nxf5 ef and now 27. Qc4! rescues the f7 knight. If 27. … Qxf7 28. Bd5 snags the queen, and if 27. … Na5 the neat double check 28. Nh6+ brings down the house.

I thought that Black’s best defense to 25. Nh4! was 25. … Nxe5. Now Rybka likes 26. Qxf5 best for White, but I thought that a simpler approach was better: 26. Nxe5 Qxe5 27. Nxf5 Qxf5 (27. … ef?? 28. Re1!) 28. Be4. Here we have material equality, an exchange for two pawns, but I think White’s position is vastly preferable, as I control all the important lines, Black is not finished developing, and in particular my pressure on h7 and g7 gives me super chances for a successful kingside attack.

Evidently Shredder didn’t want any part of this, because it played something I hadn’t expected. After my first move 24. Nf7+!, it decided to give up its queen for two pieces. It played 24. … Rxf7 25. Ng5 Rf5 26. Bxd5 Rxg5+ 27. Bg2 Nxe5 28. Qe4 (diagram).

timeout 2Position after 28. Qe4. Black to move.

FEN: r1b4k/1p4pp/1b2p3/4n1r1/p2pQ3/P7/1P3PBP/2RR2K1 b – – 0 28

Well, superficially I can understand why Shredder chose to play this way. It has two pieces and two pawns for the queen, which is almost material equality. But you’ve got to look at the quality of the pieces. Black’s army is undeveloped and disorganized, while White’s rooks are perfectly posted and the centralized queen is a monster. I was nervous, of course, about the computer’s ability to come up with tactics, but after 28. … Ng6 29. h4 Rb5 30. h5! Ne7 31. h6! gh 32. Rxd4! the tactics were all in my favor.

timeout 3Position after 32. Rxd4! Black to move.

FEN: r1b4k/1p2n2p/1b2p2p/1r6/p2RQ3/P7/1P3PB1/2R3K1 b – – 0 32

No time-out required here! I was itching to get rid of the d4 pawn and the dark-squared bishop, Black’s best defenders. The proof of the pudding is that if 32. … Bxd4 33. Qxd4+ Kg8 34. Qd8+ Kf7 35. Rc7 the knight falls and the king soon after. Of course the computer saw this and played 32. … Nc6 instead, but now White is just dominating the board and we can omit the rest of the game. (I did go on  to win.)

Lessons:

(1) Don’t give up your attacking “trump cards” (here, the mate threat on h7 and the discovery threat on the main diagonal) too soon, just for the sake of winning material.

(2) In bridge you can’t play two trump cards at once, but in chess you often can! That’s often the best way to play them.