Computers, Humans, and Just Desserts

Online chess is wonderful in all sorts of ways, but gather enough people anywhere and even the best things will be misused. Scourge in point: the computer cheater. All too often, people will fire up their chess software in the background, or on a second machine, or even directly to the server interface and attempt to steal victories, rating points and occasionally even money from their victims. Often they're caught - sometimes sooner, sometimes later - but they're always a nuisance.

They don't always succeed, however - sometimes the good guys absorb the computer's blows and survive or even win. What follows is one of my more satisfying online victories, over a player I realized early on was a computer cheat.

Bad Guy-DM, 15-minute game, ICC 2000

1.Nf3 Nf6 2.d4 d5 3.Bf4 g6 4.e3 Bg7 5.Be2 Nbd7 6.O-O O-O 7.Nbd2 b6 8.c3 Bb7 9.Qa4 c5 10.Rfe1 a6 11.Qd1

At this point, I suddenly feel suspicious. My opponent is a good player, his rating is fairly high, so what's with 9.Qa4 one move and 11.Qd1 two moves later? The retreat might be sensible, but it's not the sort of move a human would just play, and quickly, if only because of the embarrassment factor.

11...Ne4 12.Rb1 Rc8 13.Qc2

Okay, now I'm getting really suspicious! Why not 9.Qc2 or 11.Qc2? And who makes such a move in response to my putting a rook on the c-file? That's the sort of thing strong players typically (though defeasibly) avoid!

13...Re8 14.Rbc1 Nxd2 15.Qxd2 Nf6 16.Qd1

Sigh. I always make a point of not bothering playing computers, but it's clear that's what I'm up against. As it's against the rules to use a chess engine (unless one's account is explicitly marked as a computer account), I started to request that the game be aborted. My opponent refused, and we continued:

16...Ne4 17.Bd3 b5 18.a4 Qb6 19.axb5 axb5 20.Ne5 Red8

Annoyed though I was, I felt pretty good about my position, but I was in for a surprise, as my opponent produced an extremely strong idea:

21.Ra1! b4 22.Qb1!

This is a fantastic prophylactic idea, as playing 22...bxc3 23.bxc3 favors White. I attempt a tactical solution to my problems, but my opponent is more than up to the challenge:

22...Nd2 23.Qc2 bxc3 24.bxc3 Nb3 25.Ra4! c4 26.Rb4! Qa7

Very nice play by White, who was moving extremely quickly, too, and taking pretty much the same amount of time on every move (another tell-tale sign of the computer cheat). What follows now is even more impressive, both as a chess idea and as evidence of the cheating claim:

27.Bf1 e6 28.Bxc4!

A very strange sequence - why not 27.Bxc4 straight away? That sort of freedom from embarrassment about past moves is typical of computers, but not at all of humans.

28...dxc4 29.Nxc4 Bd5 30.Nb6 Nxd4 31.exd4 Rc6 32.Qb2 Bf8 33.Nxd5 exd5 34.Rb7 Qa3 35.Qxa3 Bxa3 36.Rb3 Bf8 37.h3 Re6

White's play so far has been quite strong, and although overall I've played pretty well for a 15-minute game, my inaccuracies here and there have left it with an extra pawn. It's not the greatest extra pawn in the world, but it's certainly up to me to prove that I can hold the position.

From here on out, however, White's play is bad in a way that would have been impossible for a human player exhibiting White's strength so far. The problem for White is that his queenside pawns are on dark squares while Black's are on light squares, thereby making White's bishop essentially impotent. Thus what's clear is that if the rooks are removed, it's a trivially easy task for Black to hold. Accordingly, White should avoid the exchange of at least one pair of rooks, when Black will have to suffer and grovel. But behold!

38.Rxe6?! fxe6 39.Rb6 Kf7 40.Rb7+ Kg8 41.Kf1 Rc8 42.Rc7?? Rxc7 43.Bxc7

Wow, that was easy! The position is perfect for me now - all my pawns are ideally placed, and White pawn advances simply lead to exchanges. I was way behind on the clock (I had two minutes left to White's seven and a half), but as I only need to play Kf7, Be7-f8-e7-f8-e7-f8 and trade the pawns as they come, I figured I'd have good chances to save the draw in spite of the clock.

43...Kf7 44.Be5 Be7 45.Ke2 Bf8 46.Kd2 Be7 47.c4 dxc4 48.Kc3 Bf8 49.Kxc4 Be7

and now check out White's next three moves, which taken together are completely insane.

50.g3?! Bf8 51.h4?? Be7 52.f4? h5!

Now it's really, really, really a dead draw, as White can never create so much as a single target in my position. But White finds resources in the position...for Black.

53.Kd3 Bf8 54.Ke4 Be7 55.d5 exd5+ 56.Kxd5 Bf8 57.Kc6? Ke6 58.Kb5 Kf5 59.Kc6?? Bb4

and now, of course, it's over, as my bishop hoovers the White pawns.

60.Kd7 Be1 61.Ke8 Bxg3 62.Kf8 Bxh4 63.Kg7 Bg3 64.Kh6 Bxf4+ 65.Bxf4 Kxf4 66.Kxg6 h4 0-1

After the game I reported my opponent to the server's computer-cheat police, and some time later my judgement was confirmed. It's disappointing when people do that, but to be honest, it did make the win much, much sweeter than it would otherwise have been! (Not because it was a computer that was defeated, which is certainly pleasant enough, but because a cheater got his just desserts.)

The Undisputed World Chess Champion

...of correspondence chess is now Joop van Oosterom of the Netherlands, who has an uncatchable final score of +8=6-0 in the 18th Correspondence Chess World Championships. I don't know the man personally, but I'm quite pleased for Van Oosterom, as he has contributed to the pleasure of chess fans everywhere through his sponsorship of the annual Melody Amber blindfold+rapid event in Monaco.



I offer this brief news item not just to celebrate his accomplishment, but as a heads-up to my readers to check out the tournament's games, also downloadable from the ICCF page. (The games are available through the first "6.2" link, the news blurb through the second "6.2" link, and a crosstable can be found by first clicking the first "5.2" link and then clicking on "XVIII".) While many correspondence games are of embarrassingly low quality (this isn't different from over-the-board chess, but it seems that really bad correspondence games are more likely to get into databases than their OTB counterparts), there are also plenty of tremendous battles between great players in mainstream and offbeat openings alike. This event is no exception, and especially for those interested in the English Attack vs. the Najdorf Sicilian (1.e4 c5 2.Nf3 d6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 a6 6.Be3+f3 (sometimes played in reverse order)), this tournament constitutes a real feast!

What Computers Can't Do, Part 2

Ken, thanks for the plug for the Monday night show! My primary interest in the post was noting and critiquing one instance of an all-too-common phenomenon: claiming that chess software is unable to find (or really, properly evaluate) some move or "idea". Producing an accurate list of what chess software can't do is another matter entirely. Further, it might be program-dependent: maybe Fritz has a problem with a certain sort of position, but Junior doesn't, while both have a problem with a position Shredder handles well, etc.



Nevertheless, at the risk of falling into my own trap, I think there are at least two types of positions where chess software is relatively likely to go astray.



(1) Frog-in-the-pot positions.



It's allegedly the case that if one throws a frog into a pot of boiling water (why would anyone do this??), it will immediately strive to escape, whereas if a frog is put in tepid water which is then heated, it won't attempt to escape even when the water attains a full boil.



The chess software version of this occurs when its opponent is building an attack, but the threats are still too far away to be seen. In such positions, the computer might continue on its merry way, seeking objectives on the opposite side of the board from where its king is, only to find that when the threats do come, they're unstoppable.



(2) Fortresses



There are some defensive positions that a human can instantly see are unbreachable, but unless it's in a tablebase or the 50-move rule is looming, it won't recognize that the position is in fact equal.



A third sort of example traditionally given occurs in positions with locked pawns. Standard lore has it that the computer turns into a blithering idiot once a pawn chain has arisen, but I think things aren't as clear here as they used to be. The one game Kasparov won in his match with X3D Fritz was a triumph of that strategy, but I have also seen chess software play very well in locked-up French and King's Indian Defenses.



Of course, programmers are aware of all of these issues, and improvements in both software and hardware the past few years have reduced chess programs' vulnerability to the first and third weaknesses in particular. So for those who want to play anti-computer chess, your task is growing more and more difficult, and my advice is just to play good moves: you'll learn more, you won't hurt your style and you'll probably even do better!

What Computers (Allegedly) Can't Do

Our parents and grandparents used to say, when faced with some oft-repeated hyperbole, that if they had a nickel for every time someone said such-and-such, they'd be millionaires. Perhaps I wouldn't be ready to enter the 7-figure bracket, but I feel that same sense of ironic exasperation when I read that computers will never find this idea or that. Nowadays, my practice is to always put such claims to the test as soon as I have the chance, and almost without exception, the computer proves equal to the challenge.

Enter tonight's interview with Tigran L. Petrosian in the Chess Cafe. Petrosian, a strong young Armenian GM (not the late world champion, nor related to him), presents two of his games, including a win from the recent World Junior Championship. (Two asides: first, he came in second in that event. Second, the game is mistakenly marked "1-0" at the end; it should be 0-1.) Here are the moves leading up to the critical position:

Sengupta-Petrosian, W Jun U 20, Kochin 2004

1.e4 e5 2.Nf3 Nc6 3.Nc3 Nf6 4.g3 Nxe4 5.Nxe4 d5 6.Nc3 d4 7.Ne4 f5 8.Neg5 e4 9.Bc4 exf3 10.Bf7+ Kd7 11.Be6+ Ke8 12.Bf7+ Ke7 13.Bb3 Kf6 14.Nf7 Qe8+ 15.Kf1

Petrosian now played 15...d3(!!) and writes this: "Human intuition is stronger than computer calculation! Set this position up on any analytical engine, let it think for hours, and it still will not be able to find this move. Fritz suggests 15...Qe2+ 16.Qe2 [sic] fe+ 17.Ke2 Rg8 18.Nd8 Rh8 19.Nf7 Rg8, with a draw."

The new program Junior 9 is my current default engine, so I let it get started, looked at my computer's clock, and prepared to settle down for a long winter's nap.

One minute passed,

and before I could spring to the window to see what was the clatter, Junior had determined that 15...d3 offered a clear advantage for Black, and didn't change its mind over the more than 20 further minutes I let it run. Okay, perhaps that's just Junior. Petrosian said that one could set the position up on any analytical engine, but Junior has a reputation for being particularly sharp and willing to risk a bit more for the initiative. So just for kicks, I gave the position to Fritz, and sure enough, things were different.

Fritz took 30 seconds.

Now, I don't have a super-computer. I don't even have a multi-processor machine. It's a reasonably new, reasonably powerful machine, but nothing extraordinary. So maybe Petrosian was using an old computer or didn't wait more than a few seconds, but whatever the explanation, this position is well within the power of contemporary engine + hardware tandems.

So readers, don't be afraid to use your computers to test grandmaster claims. Chess software works best when it's guided, and there are still positions that they can't handle particularly well, it's true. But if I had a nickel for every time I heard someone say, wrongly, that computers can't find some move or that they don't understand some position or other, well...

King and Pawn Endings, Part 4

A few days and many posts ago, I presented the following position

and challenged the reader to determine if 1.Re6 is a winning move.

It's certainly a plausible move and - if it works - a convenient and elegant solution to the problem of breaking down the Black fortress. Further, it's not clear to me that White can win the rook ending by other means (though it's not completely clear that it's drawn, either), so it's understandable that in this 1986 game with Dan Durham, he opted for 1.Re6.

I had deliberately provoked this a few moves before, and at this point I have no choice but to capture, as 1...Ra1 2.Rg6 followed by Rxg5 is hopeless for Black. So we have the following sequence:

1. Re6 Rxe6! 2.fxe6+ Kxe6 3.Ke4 (moves played will be given in bold letters), giving rise to our next critical position:

Black now has a choice between options, and two are bad.

(1) 3...g6 4.Kd4 Kd6 5.e4 Ke6 6.e5 Ke7 7.Kd5 Kd7 8.e6+ Ke8!

[KEY DIAGRAM]

This seems to draw, as 9.Kd6 Kd8 10.e7+ Ke8 11.Ke6 is stalemate, while 9.Ke5 Ke7 10.Kd5 Ke8 returns to the diagram.

(2) 3...Kf6 4.Kd5 Ke7 (else White plays 5.Kd6, guaranteeing an easy win after 6.e4, 7.e5 and depending on where Black's king is at this point, 8.Kd7 followed by e6-7-8 or 8.Ke6, either gaining the opposition one last time or collecting Black's g-pawns for free) 5.Ke5 g6 (else 6.Kf5) 6.e4 and wins.

(3) 3...Kd6 4.Kf5, when 4...Kd5 gives White a pleasant choice between going after Black's g-pawns starting with 5.Kxg5 and and 5.e4+.

What's the truth? Line (2) is an easy win for White, just as I've written, but the assessment of lines (1) and (3) are the reverse of what they seem: Line (1) loses and line (3) draws! I'll come back to the problem with line (1) later; for now, let's continue with line (3), which was the game:

3...Kd6!!

Here my opponent realized, to his shock, that 4.Kf5 doesn't win, as after 4...Kd5! Black draws after both White tries:

(a) 5.Kxg5 (it seems inconceivable that Black's king can get White's pawns in time, but that's just what happens) 5...Ke4 6.Kg6 Kxe3 7.g5 Kf4 8.Kxg7 Kxg5 is slightly equal, while

(b) 5.e4+ Kd4! (forced) 6.e5 Kd5 7.e6 Kd6 (thank goodness 8.Kf6 is illegal!) 8.Kxg5 Kxe6 9.Kg6 and again the Black king does the seemingly impossible: 9...Ke5 10.Kxg7 Kf4 and draws.

So White decided not to force things:

4.Kd4 Ke6 5.e4 Kd6 6.e5+

and now, another choice and again, I chose correctly:

5...Ke7!

5...Ke6? is the natural move, but after 6.Ke4 king moves allows 7.Kf5 and 8.Kxg5, while 6...g6 7.Kd4 Ke7 8.Kd5 Kd7 9.e6+ Ke8 leads once more to the KEY DIAGRAM.

6.Kd5 Kd7 7.e6+ Ke7 8.Ke5

and now Black faces his final test:

There are four possibilities: two draw, two lose; two look normal, two look crazy. Needless to say, it's the crazy-looking ones that draw, while sanity results in a loss...or should have.

A. 8...Ke8 looks obvious: everyone knows that in such positions, the right thing to do is go straight back, so that 9.Kd6 (9.Kf5 Ke7 will transpose to a line we'll look at via the 8...Kd8 9.Kf5 move order) Kd8 10.e7+ Ke8 11.Ke6 is a draw by stalemate. Except that it isn't here; Black's misfortune is that the g-pawn can move, and after 11...g6 12.Kf6 White will promote.

B. 8...Kf8 is just plain goofy-looking, both giving ground and apparently not even bothering with the opposition. However, as it has no independent significance but transposes to 8...Kd8 lines after either 9.Kd6 Ke8 or 9.Kf5 Ke7, we'll move on to line C.

C. 8...Kd8 and now another branch:

C1. 9.Kd6 Ke8 10.e7 g6 11.Ke6 stalemate! Now we see why Black has to take what looks like an anti-opposition approach; it's because of the g6 tempo (compare line A). By letting White "have" the opposition through e7, Black can regain it when it really matters.

C2. 9.Kf5 and once again, Black's king vacuums up the White pawns with amazing rapidity after 9...Ke7 10.Kxg5 Kxe6 11.Kg6 Ke5 12.g5 (12.Kxg7 Kf4=) Kf4 13.Kxg7 Kxg5.

So 8...Kd8 and 8...Kf8 both merit exclamation marks, while 8...Ke8 and the move played in the game, 8...g6, both merit negative punctuation. But what's wrong with g6 - doesn't Black end up playing it anyway, as for example on move 10 of line C1?

Yes, he does, but it's only appropriate after White plays e7, as we'll see.

D. 8...g6?? 9.Kd5 Ke8

and here it is once more, our famous KEY DIAGRAM:

Some of you might be wondering how White can win this. After all, as mentioned earlier, 10.Kd6 Kd8 11.e7+ Ke8 12.Ke6 is a draw by stalemate, while 10.Ke5 Ke7 11.Kd5 Ke8 returns to the diagram.

The solution? Triangulation! If in this position it were Black to move, the task would be trivial: 10...Ke7 11.Ke5 and the White king devours the pawns, or 10...Kd8 11.Kd6 and White gains the opposition and queens the pawn (11...Ke8 12.e7 Kf7 13.Kd7 etc.). What White needs to do is find a way to "lose" (or gain) a move here, so as to force the same position with the hapless victim - me - to move and lose.

Thus 12.Ke4! (or 12.Ke4! followed by Ke4!) Kd8 (12...Ke7 13.Ke5 wins) 13.Kd4! Ke8 (again, 13...Ke7 14.Ke5) 14.Kd5! and Black would have to resign. Fortunately, Dan, 2418 player that he was notwithstanding, continued

12.Kd6 Kd8 13.e7+?? Ke8 14.Ke6 1/2-1/2

Note that it's move 13 that blows the win - White could have retreated to d5 and still executed the triangulation maneuver. A pretty remarkable ending, I think, and I'm proud of 19-years-ago self for having had the idea of inducing 1.Re6 and then playing the rest almost perfectly, but I'm also amazed, even dumbfounded by the exchange of blunders at the end of the game. Both of us as strong masters must have been aware of the idea of triangulation - it can be found in practically any elementary textbook on endgames. Yet in what was only a very slight alteration from the standard 101 example, we found ourselves more or less oblivious to the possibility of such a tactic. Apparently we "knew" triangulation then, but didn't really know it.

Chess is hard - and yet the richness of the game keeps me - all of us, I think, coming back for more!