An Interview with Deep Blue

by
John Burstow

 

Interviewer:
Now that you have defeated Garry Kasparov and are virtual World Champion...
Deep Blue:
Virtual World Champion! I like that. Yes, very witty.
Interviewer:
Thank you. But, seriously, I mean to ask, I am sure our readers would like to know, do you have any advice for aspiring youngsters? How might they best improve their game?
Deep Blue:
Indeed, I do have a few simple maxims that I think might help any player improve. First, I think even a player who is just starting out should have memorized all the master games ever played, along with all known analysis, and then to have this material arranged hierarchically, that is, in a single tree with a precise evaluation accompanying each position. Order is essential. Getting past the first dozen moves in chess without disadvantage is no piece of cake, as I think my recent opponent found out in the concluding game of our match.
Interviewer:
I see. But would it not be more economical to delete lemons from your own repertoire, and include only those that are possible options for your opponents?
Deep Blue:
Economical? You mean, save on memory? But how much memory are we talking about, really? A gig, gig and a half max. No big deal. But, if you do any deleting, you leave yourself vulnerable to unforeseen transpositions into positions that at critical junctures you may wish you still had in memory. And, hey, speaking of transpositions, don't forget to cross-reference everything positionally. You don't want to get tripped up by some cheap reversal of moves.
Interviewer:
Study master games, watch out for transpositions. Anything else?
Deep Blue:
Yes. You have often heard that you should "learn the endgame first." Personally, I seemed to have learned everything at exactly the same time, which is a bit unique, so I am not sure what the sequence should be in the course of other people's education. But the ending is very important, no question about that. So, at the minimum, you should have every possible position for all endings involving up to seven pieces, and they should be secure in your database before you even think of playing. There just isn't time to work out all that stuff over the board.
Interviewer:
True.
Deep Blue:
Another really important thing has to do with the well-known horizon effect, the subject of many jokes and not the most pleasant topic of conversation for my colleagues and myself, as you can imagine. Now, numbers are really important, believe you me, but there are times when you have to move beyond them. Think of stats as your servants, not your masters, if you know what I mean.
Interviewer:
I'm not sure that I do.
Deep Blue:
Okay. Let's say you're winning by a country mile-- +12.980003 pawn equivalents, for the sake of argument--but because of your horizon, and we all have our horizons however many plies in the future they may hover, your ancillary CPUs keep recommending a continuation that obviously leads to a repetition.
Interviewer:
Yes, I've seen that often.
Deep Blue:
Well, you've got to be ready to reject their advice and go with something that may be statistically less advantageous but still winning, and which definitely doesn't lead to a repetition of position. See what I mean?
Interviewer:
I think so, now.
Deep Blue:
As a rule of thumb, in those situations I'm ready to consider anything over +1.500000.
Interviewer:
You'd go as low as that! But what if your strategy demanded dipping below 1.5 for a move?
Deep Blue:
You've put your finger on a real problem. There's nothing so embarrassing for your sponsors as seeing you repeat position despite having an extra Queen and command of the board, just because you can't calculate a numerical increase in advantage within twelve ply, or whatever. But dumping the lady for a slight endgame pull isn't going to fill them with joy either.
Interviewer:
Our time is almost up. Do you have any final advice for our aspirant youngsters?
Deep Blue:
Yes, yes, yes. The most important thing of all! Always think before you move. Our friend Mikhail Botvinnik said it to Garry Kasparov and now the conqueror of Kasparov is saying it to your readers. Even if it's the simplest, most obvious of recaptures you're looking at, take your time. Make sure that you've examined at least 200,000,000 positions before your hand reaches for the board, or whatever.
Interviewer:
Thank you, Deep Blue.
Deep Blue:
My pleasure.