by Steve Lopez

Every now and then an Internet message board post really gets you thinking. I was recently reading a thread on pawn promotions and the subject of underpromotion came up. We've all seen (and possibly played) games in which a pawn underpromotes to a Knight, but what about games in which a pawn becomes a Bishop or Rook?

For those coming in late (I've been receiving a lot of e-mails lately from folks who are just learning the moves of chess), when a pawn reaches the eighth rank, it's allowed to promote to any piece except a second King. The piece does not have to be one that's already been removed from the board; it's therefore theoretically possible to have nine Queens on the board (but the chances of stalemate in such positions are huge, so you'll seldom see a player with more than two Queens). Typically, a pawn will promote to either a Queen or Knight: the Queen because it's the most powerful piece or the Knight because of its unique move. Promoting a pawn to anything other than a Queen is called underpromotion.

You'll sometimes see a player underpromote to a Knight because the move results in instant checkmate; in other cases a Knight underpromotion might better suit the requirements of the position (for example, the underpromotion might result in a Knight fork of the opponent's King and Queen). But why would anyone underpromote to a Bishop or Rook? After all, the Queen combines the movements of these two pieces, so why get one of these pieces when you could have a Queen instead?

I recently posed this question in an Internet discussion group. I could think of just a few cases in which a player would underpromote to a Bishop or Rook: he's trying to be a butthead, he's going for some cheap laughs in a casual game, he doesn't know any better, or he instantly mates anyway with the minor piece so he underpromotes just for the heck of it.

Another poster, who is apparently a whole lot smarter than me, replied that there can be cases in which promotion to a Queen causes stalemate, so the player opts for the underpromotion instead. That started the old wheels turning -- how often does underpromotion to a Bishop occur and what are the circumstances? Hmmmm...

ChessBase 8's search functions make it easy to find the answer. You just do a maneuver search for a Bishop underpromotion and see what turns up.

There are various ways to bring up the search mask, depending on what database(s) you want to use in the search. I decided to search a few dozen issues of The Week in Chess, so I right-clicked on the icon for my TWIC database and selected "Search" from the popup menu. The search mask came up and I selected the "Manoevers" tab to get the following dialogue:

This isn't the easiest dialogue in the program, but using it isn't exactly rocket science either if you take it a step at a time. I first decided that I was just going to look at White underpromotions -- so I clicked the radio button next to "W". In the pulldown window right under that, I selected "P" for the piece that was going to move. In the box immediately to the right, I typed in "?7", meaning that the promoting pawn could be any pawn starting its move on the seventh rank. In the next box to the right, I typed "?8", meaning that the pawn ends up on any eighth rank square. The question marks are wild cards -- they're an old computer symbol that means literally "any value here". If I wanted to narrow the search to just e-pawns that underpromote, for example, I'd use the letter "e" instead of the question mark. But, in this case, any file will do so I used the question mark instead.

Dropping down a line, I put a check in the box next to "Promotion". In the pulldown window just to the right of that, I selected "B", meaning that I wanted to see instances in which a pawn promotes to become a Bishop.

But I'm not done yet. I'm not interested in a multi-move combination, so I unchecked the box next to "Check move order". Since promotion tends to happen late in the game, I set the "Moves" boxes to "25" and "90" -- this means that the program will only search moves 25 through 90 of the games in the database. Finally, since this is a single-move manoever (the pawn advances and promotes to a Bishop), I set "Length" for "1".

After I was finished setting the manoever parameters, the dialogue box looked like this:

Now at this point, I could further refine the search parameters by clicking on another tab and setting some additional values. Remember that we're searching for games in which White underpromotes a pawn to a Bishop. I could so some other stuff, like add Kasparov's name as White to see if he'd ever done it. Or I could type in a range of years, say "1999" to "2001" to find games from just the last two years in which this occurred. I could check the "C" box to find only games in which verbal commentary appears. The options are limitless. But since underpromotion to a Bishop is a pretty rare occurance, it's best to not set any additional search criteria -- I have a pretty good idea that such games are few and far between, so there's no sense in further limiting the search.

Once I set the manoever parameters, I click "OK'. ChessBase 8 searches this particular database (116,353 games) pretty quickly. In seconds, I get a new "Search results" window with six games listed. If I single-click on a game, it loads it into the board right in the search results window. Double-clicking on a game from the list loads it into a new game window all its own.

All of the games in this database are relatively recent. The oldest game is from 1996. After clicking on it, the game is loaded into the small chessboard in the "Search results" window, with the proper position (after the underpromotion) already loaded:

This position is from the game Lacroix-Bolzoni, Brussels 1996. White has just played 82.fxg8=B. Take a close look at the board position. Had White promoted to a Queen, the game would have been a stalemate -- Black's not in check and he would have no legal moves. A Knight underpromotion would work (with a mate in nine). But the game after the Bishop underpromotion continued: 82...Kg8 83.Kg6 Kh8 84.Bg7+ Kg8 85.h7# and White won anyway.

Another interesting underpromotion rationale occurs in the realm of computer chess. In looking though some Bishop underpromotions, I noticed that computer programs tend to underpromote just for the heck of it whenever the piece's recapture is imminent. For example, take a look at this position from the 1999 Paderborn tournament:

Nimzo has just played 43.b8=B. Evidently Nimzo figured that 43...Bxb8 was on the way, so why not underpromote? The wood's going back into the box anyway, right?

The same philosophy can apply to human players, such as in this position from Gerber-Maurer, Lenk 2001:

White plays 25.cxd8=B, knowing that the Black Queen will recapture.

It was interesting to note that in all six games in which White underpromoted to a Bishop, none of them were Black wins. White won four of the games and drew two. So let's try the search again, but from Black's perspective to see if he does as well as White. We'll just click the "B" radio button and change "?7" to "?2" and "?8" to "?1".

This time we get eight games -- Black wins just three, while two were drawn and the other three were victories for White.

Here's one that made me laugh out loud:

In this position (from Mikhailov-Sapundzhiev, corr 1992), Black figures he'll be a wise guy and underpromote, so he plays 59...a1=B+. Whether he promotes to a Bishop or Queen, he skewers the King and (in effect) has traded a pawn for a Queen. But he doesn't count squares correctly. After 60.Kg8 Bxh8 61.Kxh8 Kxc4 62.Kg7 he becomes embittered. 62...Kd5 moves toward the opposing King and results in mate in 85 moves (pardon me while I laugh), but Black instead went backwards with 62...Kd3 and ended up drawing: 63.h5 c4 64.h6 c3 65.h7 c2 66.h8=Q c1=Q (no underpromotions now!) 67.Qd8+ Kd2 68.Qe8+ Qe3 1/2-1/2

And sometimes there are cases that just baffle you. Check this one out, from Mellem-Borchgrevink, Gausdal 2000:

I don't care how many chess games you've replayed and how jaded you've become -- this one will really stump you. Black has just played 47...d1=B. Who knows what this guy was thinking? 47...Rxg7 48.Kg2 d1=Q 49.h4 Qxf1+ 50.Kf3 Rc3+ 51.Ke4 Qc4# wins; Black instead chose to underpromote. But now it gets even weirder. 48.Rg8# wins instantly for White, but he plays 48.bxa5. Black replies 48...bxa5 and once again Rg8 is mate. You guessed it -- White misses it again and plays 49.Qg6. Then it gets truly bizarre: 49...a4 50.Ne3 Bf3+ 51.Nf1 -- both players missing mating sequences. From here on out, both players miss mate-in-one on every move: 51...a3 52.a6 b2 53.b7 a1=N (Why a Knight?? Egad!) 54.b8=N (Again a Knight, for crying out loud!) 54...Nc2 55.Nd7 Nd4 56.Nf6 and then one of the players finally stumbles into it: 56...Ne2#. I just wonder how long it took the players to realize that this was mate -- I can picture them huddled over the board pondering the position for another twenty minutes before Black jumps up and yells "Zounds!" Man, if this had been a boxing match, the ref would have stopped the fight.

That's it for me -- my head hurts after that one (I was banging it on the desk). I have to take some aspirin now and hit the rack for about a week. Until I wake up, have fun!

You can e-mail me with your comments, suggestions, and analysis for Electronic T-Notes.