"Who's Winning?": Computers vs Humans

You'll see this kind of thing happening time and again. The neophyte reaches a position where he is (say) a full queen up, and asks (on the forums, if not elsewhere): "So how exactly do I go about winning this thing anyway?"

And so he consults the engine for an answer. And the gadget proceeds to spew out a lot of moves and numbers...but offers up precious little in the way of advice. In fact, it merely sidesteps those few continuations which might lead to the odd disaster and accepts everything else as "winning" (although perhaps "maintaining the win" would be more accurate).

So how is this poor (and hardly hypothetical) tyro to play for the win? To "aim at it," so to speak?

The Shortcomings of Silicon

First off, you have to realize what you can (and can't) get from an engine.

Generally speaking, the farther from 0 a position's evaluation is, the less reliable/meaningful comps tend to get. And after all, if you're a 3500 player who's 2 rooks up and you've (somehow) managed to drop one...well what the heck, you're still a rook up! No doubt the thing could win just as easily with a good deal less in surplus.

So was it a mistake to drop that rook? Or even an "inaccuracy"? Apparently not, if you're made of silicon. :)

In a similar vein, here's the sort of annotation you'll routinely see in engine analysis: "Mistake! This allows mate in 11. Better was QxN..."

Of course, the experienced player can often only manage a wry smile at that, since the suggested move usually just gives up the queen for nothing. In fact, to a master that might as well read: "Better was resigning."

Another bit of computer myopia involves putting up "the best defense"--or actually, failing to do so. To the comp, a wrecked position is a Confirmed Loss--and since one continuation is as good (or bad) as another, it doesn't much matter what the defender plays.

True for another engine, no doubt--but how completely untrue for human players!

The Voice of Experience

I remember once way back when (when I was playing my Fidelity microprocessor a game), I actually managed to win Q vs R+P...with 4 minutes left on my clock!. The device first jettisoned the pawn (apparently I was "inevitably" going to win it anyway). And a number of moves later, it did the same with its rook! :D

It seems that I had a forced win of the thing (in there somewhere). But instead of making it as tough on me as possible (as any sensible human opponent would've done), it just figured that "All roads lead to Rome" (or something like that) and proceeded to make it as easy on me as possible. :)

So don't get locked into thinking that the sign of Ultimate Mastery is announcing mates in 17 (or anything like that). Masters don't have to waste their time with such antics (and since most of the people here are playing blitz, there wouldn't be time for it anyway!). The real point is that humans don't have to calculate things anywhere near that finely to feel they've got a won game well in hand. Just let things go on normally and eventually you'll get to checkmate. :)

So if you see a forced mate in 3 with checks, and there's what seems to be a quiet, "problem-like" mate in 2 without checks...take the mate in 3. Because that problemesque mate in 2 might just have something wrong with it; and then--after your opponent has uncorked the Amazing Saving Resource that you somehow overlooked--you'll spend half the time trying to dig your way out of the mess you just made for yourself, and half the time kicking yourself for not taking the easy win when you had one.

And yes, this is (once more) definitely the Voice of Experience speaking! :D