The Computer Gets Ridonkulous

by admin on May 23, 2020

I’m glad that I got a couple of comments on my last post, “What’s the Best Move?”, saying that my readers found the positions just as hard as I did. It makes me feel better, because most of them were positions where I was playing against the computer and messed up. So now I feel as if I have company — a lot of my readers would have messed up, too!

Here’s a position that takes this theme to an absolutely berserk extreme. I’m White, and the computer has just played 42. … Bc5. What would you play here if you had White? By the way, the time control has passed (at move 40), so I have lots of time to decide.

Position after 42. … Bc5. White to move.

FEN: 1k1r4/2p2p2/1pP1nnp1/1Pb5/2Q4p/R5PP/5P1K/8 w – – 0 43

Okay, I’ll bet you noticed the same thing I did. Actually for the past several moves I’ve been trying to double on the a-file so that I can checkmate Black, and the computer has managed to prevent me from doing so. But with its last move, it seems to have walked into one of the oldest mate patterns in the book.

I looked and looked. Why can’t I just sacrifice on a8, bring my queen to the a-file and checkmate on b7? Would the computer really just walk into this combination? Its last move was really fishy, because Black moved his bishop right to the square that his knight needs (c5) to defend the mate on b7.

This last point raises a bit of an alarm. After 43. Ra8+ Kxa8 44. Qa4+ Kb8 45. Qa6 hg+, I have to avoid moving my king to a dark square, because that would allow the bishop to move away from c5 with a tempo. So 46. Kxg3? Bd6+ is off the table. Likewise, 46. fg? Rd2+ is off the table, because the rook move vacates d8 for the knight. However, after 46. Kg2! neither the rook nor the bishop can play a check, so I thought I was just winning. So that’s what I played.

Do you see what I missed?

First, Fritz played 46. … Nf4+! I mean, in one sense it’s obvious that it would play this move, because it’s the only way to delay the checkmate. I really thought this knight move was just a spite check. The thing is that Fritz can no longer move its knight to c5 or d8 to defend checkmate, and this means that I no longer have to worry about allowing bishop or rook checks. I can now move my king wherever I want, including taking the pawn on g3. Which is what I did: 47. Kxg3. (If you’re curious about the alternatives, after 47. Kf3 Black wins with 47. … Rd3+ 48. Kxf4 Bd6+ 49. Kg5 Ne4+ and the other knight comes to c5. And moving back to the first rank is not an option because 47. … Rd1 is checkmate.)

Of course Fritz played 47. … Rd3+, because again, what else is it going to do? I was still under the impression that it was simply playing spite checks. However, as I tried to figure out how I was going to escape the checks, I started to become a little bit worried. On the obvious move, 48. Kxf4, Black is able to defend in the same way as the parenthetical note above. Except it’s even worse this time, because I get checkmated! Here’s how: 48. Kxf4? Bd6+ 49. Kg5 Ne4+ 50. Kh6 (the other two possibilities also walk into mating nets) Bf8+ 51. Kh7 Nf6+ 52. Kh8 Rxh3 mate. Truly a picturesque finish! I like it so much that I’ll show the diagram here.

Position after 52. … Rxh3 mate. (Analysis.)

FEN: 1k3b1K/2p2p2/QpP2np1/1P6/8/7r/5P2/8 w – – 0 53

I can’t say that I saw this amazing variation, but even if it weren’t for the mate, Black could defend perfectly well with 50. … Nc5. So I had an uneasy feeling about taking the knight and decided I couldn’t do it. Besides, I had another option that I still totally believed was winning, which was 48. Kh2. Yes, it’s a bit scary to face those four rampaging Black pieces with my nearly naked king, but I still didn’t see how the computer was going to prevent checkmate on b7. Do you see what I missed?

Position after 48. Kh2. Black to play.

FEN: 1k6/2p2p2/QpP2np1/1Pb5/5n2/3r3P/5P1K/8 b – – 0 48

In some ways the solution is obvious. Black just has to keep playing checks until either he doesn’t have any more checks, or until he can move a knight to c5. For a computer this is all just a piece of cake. For a human, it’s completely bewildering.

So the computer plays 48. … Rxh3+ and I play the forced 49. Kg1. And now it has to keep playing checks, so of course it plays 49. … Bxf2+!

It would be nice if I could decline it with 50. Kf1, but Black would insist with 50. … Rh1+ 51. Kxf2 Ne4+, etc. Instead, I reluctantly accepted with 50. Kxf2 Ne4+. Now I have to be careful, because on 51. Kf1 or 51. Ke1, Rh1+ would be mate! So I have to play 51. Kg1. Black can now play … Nc5, with a won position, any time it wants, but the remaining moves are a clinic on how to win with a rook and two knights against a king. The game concluded 51. … Rg3+ 52. Kf1 Rf3+ 53. Ke1 Ng2+ 54. Ke2 Rf2+. This allows my king to break towards freedom with 55. Kd3, but it will cost my queen after 55. … Nc5+. So I retreated with 55. Kd1 Ne3+ 56. Ke1 Nc2+ 57. Kd1 Nc5 and … holy crap … my queen is trapped! The only safe square, a2, isn’t safe because of 58. Qa2 Ne3+. So White resigned.

What an amazing finish! Obviously, in retrospect, the “brilliant” 43. Ra8+ was in fact the losing move. Obviously it leads to White being checkmated or losing his queen after 16 moves. Obvious, that is, if you are a computer. In analyzing the position, Fritz never for a second doubts that Black has a winning advantage. Right there you see the enormous gulf between a human and a computer, because I thought that I was winning and Fritz was just playing spite checks.

What should White have done instead in the original position? According to Fritz, White should have just played the simple, quiet retreat 43. Ra2!! White’s idea is to play Qa4 followed by checkmate on a8. Black has only one way to stop this, 43. … Rd4. Now White plays 44. Qc3! Again, the threat is simple: White wants to play Qa1 followed by checkmate on a8. Again, Black has only one way to stop this: 44. … Bb4. Now, instead of 45. Qa1?, which Fritz says is only equal, White has to find 45. Qf3! This variation is starting to become reminiscent of the game, only with the roles reversed. Now it is White who is tormenting Black with annoying little threats, equivalent to “spite checks.” Again, Black seems to have a perfectly okay defense with 45. … Be7. And now comes the truly sick winning move, 46. Kg2!!

Position after 46. Kg2 (analysis). Black to move.

FEN: 1k6/2p1bp2/1pP1nnp1/1P6/3r3p/5QPP/R4PK1/8 b – – 0 46

The point of this move (which, by the way, is the only correct move according to the computer) is to take away spite checks. If Black tries to sit on the position with 46. … Rd5, White will play 47. Qb3 Rd4 48. Qb1 Bb4 49. Qa1 Ba5. This arrives at the same position we could have had four moves earlier, but now White’s king is on g2 and that makes all the difference. After 50. Re2! White threatens Rxe6 and there is no defense for Black. If White’s king were on h2, Black could play 50. … hg+ (“spite check”) 51. fg Ne8! 52. Rxe6? Rd2+ (“spite check”) and Black wins. With the king on g2, though, 50. … hg is simply answered by 51. Rxe6! Another interesting line from the above position is 46. … Nc5 47. Qc3 (same idea) Ra4 48. Rxa4 Nxa4. Black has almost thwarted White’s attack, but unfortunately his pieces are too scattered and disorganized, and he will lose one of them after 49. Qe5! Bd8 50. Qd4!

I don’t think that even one human player out of 1000 would have played this position correctly. A timid player might have played 43. Ra2 for the wrong reasons, because he couldn’t work out whether 43. Ra8+ led to mate. But the rest of the line requires not timidity but an incredible in-depth understanding of where the pieces want to be. The rest of us can only watch and marvel at the way a tiny, gentle gust of wind (46. Kg2!!) causes Black’s position to collapse.


  1. In theory, as Dan Heisman says, you should analyze a position until “quiescence.” That is, White should not really play the sacrifice 43. Ra8+ until he has worked out how to escape from Black’s checks. The trouble with Heisman’s advice here is that it takes so long — 10 moves or more, many of them not obvious — to see clearly that there is no way out.
  2. If you can’t work the position out to quiescence, you might just have to guess. In this position where Black has four attacking pieces, the odds really are in Black’s favor to either checkmate White or maneuver a knight to c5 somehow.
  3. The hardest thing to do when you are trying to checkmate your opponent is to play a quiet rearranging move, or prophylaxis. Your heart is pounding, your adrenaline is racing… but sometimes you’ve just got to channel Aron Nimzovich, slow down and play the simplest, calmest moves.
  4. As a general rule, it is better for you to be playing the “spite checks” and the “annoying little threats,” rather than your opponent. Compare the variations after 43. Ra8+, where White had to find some way to handle all of Black’s annoying little threats, to the variations after 43. Ra2, where it was Black who had to keep defending against White’s annoying little threats.
  5. Sometimes chess really is too hard for humans. When that happens, just accept it and go out for a walk.
Print Friendly


What’s the Best Move?

by admin on May 9, 2020

Old-timers like me can remember when every issue of Chess Life came with a Larry Evans column called “What’s the Best Move?” Evans would give you positions with three choices of possible moves. The idea, I think, came from a book of the same name that he wrote in 1973.

I’m actually not too fond of quizzes that do part of the thinking for you, by giving you three options. In a real game, part of the challenge is coming up with an appropriate list of candidate moves.

However, one good thing about Evans’ approach is that it teaches you not to make up your mind until you have looked at three possibilities. Many of my students only analyze one move. In fact, I have that problem too, when I’m playing speed chess or getting seriously low on time.

I also liked the fact that Evans called it “What’s the Best Move?” not “What’s the Winning Move?” In reality, there aren’t that many moments in our games when it’s our move and there is one and only one way to win. More frequently we just have to make the best choice out of several alternatives.

Anyway, I thought I would do a Larry Evans-style quiz today, except I’ll give you four options per move. The other interesting wrinkle is that every one of these games was played by me against the computer within the past week! In all positions it’s my turn to move. (Understandably, because I don’t care about how to improve the computer’s “thought” process — I care about how to improve mine.)

Position 1.

Position after 33. Re6. Black to move.

3r1k2/6pp/2p1Rp2/1pP5/3p3P/6P1/2P2P2/6K1 b – – 0 33

Your choices: (A) 33. … Re8; (B) 33. … d3; (C) 33. … b4; (D) 33. … Rc8.

Position 2.

Position after 36. Ke3. Black to move.

FEN: 8/2r2kpp/2pR1p2/ppP5/4P3/P3K3/6PP/8 b – – 0 36

Your choices: (A) 36. … Ke7; (B) 36. … Kg6; (C) 36. … b4; (D) 36. … a4.

Position 3.

Position after 25. … Bg5+. White to move.

FEN: 4r2r/ppp2Rpk/7p/6b1/8/2B5/PPP2PPP/2K1R3 w – – 0 26

Your choices: (A) 26. Kb1; (B) 26. Kd1; (C) 26. f4; (D) 26. Bd2.

Position 4.

Position after 33. Rf2. Black to move.

FEN: 5r2/1b4rk/1p2pq2/p2pN2p/4pP2/P1P3PP/1P3RQ1/4R1K1 b – – 0 33

Your choices: (A) 33. … e3; (B) 33. … Qe7; (C) 33. … d4; (D) 33. … Rfg8.

Position 5.

Position after 12. … a6. White to move.

FEN: r1bqk2r/1pp1ppbp/p4np1/1N1PN3/1nB5/8/PPP2PPP/R1BQR1K1 w kq – 0 13

Your choices: (A) 13. c3; (B) 13. Nc3; (C) 13. d6; (D) 13. Nxf7.


Position 1. Answer: (C).

This was a game I played this morning. I played (B) 33. … d3?, and the game rapidly headed toward a draw after 34. cd Rxd3 35. Kg2 (A little twist of the knife. White doesn’t even have to take on c6 yet because there’s nothing I can do to stop it.) 35. … Rc3 36. Rxc6 b4 37. Rb6.

Option (D), 33. … Rc8? is unforgivably passive. You should never passively defend in a rook and pawn endgame unless you absolutely have to. In this case, White wins a pawn anyway after 33. … Rc8? 34. Rd6.

Option (A), 33. … Re8 is not a bad idea, but White has good drawing chances. Ideally, White would like to trade his c-pawn for Black’s d-pawn and establish his rook behind the b-pawn. If he can do this and force Black’s rook to defend from the front of the b-pawn, then it will be a theoretical draw.

Variation (C) is best because it retains two different pawn-break options depending on what White does. If 33. … b4! 34. Rxc6? then I win by pushing the b-pawn b-pawn: 34. … b3! 35. cb d3 and White will have to sac his rook to stop the d-pawn. On the other hand, if 33. … b4! 34. Kf1 now I win by pushing the other pawn: 34. … d3! 35. cd b3. Now 36. Rxc6 Rb8! is too slow for White, so he will have to drop his rook back with 36. Re4 or 36. Re2. In either case White will force Black to blockade the b-pawn with his rook, and he will then stroll with his king to d5 and take Black’s c-pawn. Meanwhile Black can perhaps win the b-pawn, but then rooks get traded and White has an easy win in the K+P endgame.

Position 2. Answer: (A).

This position may look a little bit like Position 1, because it came out of the same opening (a Scotch Game where Fritz unaccountably likes to play a dubious pawn sacrifice). Anyway, this one should be pretty easy, but nevertheless I botched it in time pressure.

The question is: which way should the king go? The answer becomes obvious once you ask yourself, “What is my opponent’s plan?” Clearly White wants to bring his king to the queenside to stop Black’s soon-to-be-passed pawn on the b-file. But after 36. … Ke7, White can never set foot on the d-file, because 37. Kd4 (or d3 or d2) will always be met by … Rd7, trading rooks and easily winning the K+P endgame.

Black’s plan, therefore, is as follows: (a) Play … Ke7 to keep White’s king on the kingside; (b) push the queenside pawns and force White’s rook into a passive position in front of the b-pawn; (c) activate his king and rook and target the c-pawn, as in Position 1. As before, if Black can trade the b-pawn for the c-pawn and trade rooks in the process, it’s a win.

Instead, I played (B) 36. … Kg6?, moving my king in the wrong direction. The kingside is not where my advantage lies. I ended up drawing the game after 37. g3?! (Not really necessary, but it tempted me to go further in the wrong direction.) 37. … Kg5? 38. Kd4 Kg4 39. e5 fe+ 40. Kxe5 and I reluctantly, but correctly, concluded that I needed to go for a draw by repetition with 40. … Re7+ 41. Kd4 Rc7 42. Ke5 Re7+ etc. Move (C) is premature and lets White draw with 36. … b4 37. ab ab 38. Kd3. Move (D) makes no sense at all, as White is able to easily blockade White’s pawn majority.

Position 3. Answer: (B)

After two problems that I got wrong, turning wins into draws, here’s one that I got right. But I drew anyway!

Until this move, I just thought that I was completely winning. I have a rook on the seventh rank, with all sorts of tempting targets: the pawns on c7, b7, a7, and especially g7! Then the computer played 25. … Bg5+, a move that I at first thought was “just a spite check.” Then gradually it dawned on me that I had only one move to retain any advantage at all, and a couple of ways to lose!

First, the lemons. I put in (D) just to keep you honest. On the surface, 26. Bd2 looks plausible, but it’s actually a helpmate after 26. … Rxe1 mate! Brrrr! That’ll send cold shivers down your spine.

Also, (A) is horrible. After 26. Kb1 Rxe1+ 27. Bxe1 Rd8, it transpires that White has to give up the bishop to stop checkmate. Yikes! What’s going on here?

Option (C) is not quite as horrible, but it’s not good either. After 26. f4 (my original plan) 26. … Rxe1+ 27. Bxe1 Kg6! 28. Rxc7?? Bxf4+ wins the rook. So instead White has to play 28. Rd7, but after 28. … Bxf4+ my advantage is completely gone.

By process of elimination, the only move for White is (B). This was the last on my list of options originally. But after 26. Kd1, Black now actually has to face the problem of how to defend g7 and c7. He has only one way: 26. … Rxe1+ 27. Kxe1 Re8+ 28. Kf1 Re7 (all of which happened in the game). In this position White should still win. I have a solid extra pawn with no weaknesses, and my pieces are as active or more active than the computer’s. But the close call on move 26 affected me psychologically, I think. I never could quite regain my confidence, and I ended up making another mistake and drawing.

Position 4. Answer: (C)

I hope you were paying attention to my previous posts on pawn breaks! Black’s pieces are all loaded for bear — all except the bishop on b7, which desperately wants to join in. After 33. … d4! 34. cd e3! White has to give up the exchange: 35. Rf3 Bxf3 36. Nxf3. And now comes another terrific pawn break: 36. … h4!, prying open the g-file in spite of White’s attempts to keep it closed. If 37. Nxh4 Rxg3! the attack on the g-file is decisive. If 37. g4 Qxf4 and the pressure on the f-file is decisive. If 37. Ng5+ Rxg5! 38. fg Qf2+ wins. Wow, what great stuff!

If you chose (A), at least your heart is in the right place but you got the move order wrong. After 33. … e3 34. Rxe3 d4 35. Ref3 dc 36. bc Qe7 37. Qf1 Bxf3 38. Rxf3 Black has a better endgame with an exchange for a pawn, but there is still some work to be done. Clearly this is not quite as awesome as line (C).

Variations (B) and (D) are wrong-headed though not actually losing. Both of these moves shift pieces that were already well-placed to new positions that are not clearly better. It’s much better to bring a new piece into the attack (the bishop on b7) than to monkey around with your already well-placed pieces. And both moves give White time to prevent the pawn breaks that put his position in such peril.

So, of course, I played… variation (D). Proving that I had not learned my own lesson about pawn breaks. After a few more inaccurate moves I had to tender my resignation.

Position 5. Answer: (B), (C), or (D)!

I wanted to end with an easy one — the only one of these five games that I actually won! Both (C) and (D) are absolutely crushing, and I’m glad to say that I picked (D). After 13. Nxf7! Kxf7 14. d6+ e6 15. Nxc7 Rb8 16. Nxe6 Bxe6 17. Bxe6+ Kf8 I already have three pawns for a piece and Fritz’s king is in grave peril. The game finished 18. c3 Nc6 19. d7 Qe7 20. Bf4 Qd8 21. Bd6+ Ne7 22. Bxb8 Nxd7 23. Qxd7 and Fritz threw in the towel.

Answer (B) gets maybe half credit. It’s a move that fails to take advantage of the (rare) opportunity to blow the computer off the board with tactics. However, it turns out to be a “sneaky good” move. After 13. Nc3 White is threatening simply to win Black’s trapped knight with 14. a3. Thus, Black does not have time to castle, which he desperately needs to do. Instead he has to play something like 13. … c6 to rescue the knight, and after 14. d6! e6 15. Bg5 White still has tons of pressure.

If you chose (A), you were focused on the wrong part of the board. The king should be the main target. After 13. c3? ab 14. Bxb5+ c6 15. dc bc 16. Nxc6 Nxc6 17. Bxc6+ Bd7 18. Bxa8 Qxa8 we get a position I would certainly not want to play as White, where White has R+2P versus two pieces, but Black’s extra pieces look as if they will have Benko Gambit-style pressure against White’s denuded queenside.

Hope you enjoyed these positions!

Print Friendly


3 Variations on a Theme

April 27, 2020

Here’s a challenge of a kind I haven’t seen before. I’m going to show you three very similar positions. In the first, it’s Black to play and win. In the second, Black has to play exactly the same first two moves, and you have to show that the game is a draw. In the third, […]

Print Friendly
Read the full article →

Pawn Breaks

April 24, 2020

During the Pro Chess League season this year (which still has not ended, by the way — the finals have been postponed until September) Robert Hess made a comment that jolted me awake. “When I’m evaluating a position,” he said, “The first thing I look at is who has the pawn breaks?” What?! The first […]

Print Friendly
Read the full article →

Long Think, Wrong Think

April 19, 2020

There’s a saying that I’m really starting to like: “Long think, wrong think.” In game after game, both against the computer and in tournaments, it seems as if my weakest moves are the ones I take the longest on. A long think is usually a sign that I’m trying to talk myself out of the […]

Print Friendly
Read the full article →

Learning and Not Learning… Humans and Computers

April 12, 2020

Coronavirus check-in! It’s Easter Sunday of 2020, and here in the U.S. the coronavirus epidemic seems to be near its peak. That’s both good and bad. It’s good because “peak” means that the number of new cases and new deaths should start going down every day. In New York and California, the number of intensive […]

Print Friendly
Read the full article →

Wild Fritz Endgames, No. 1 (R vs. 3P)

April 6, 2020

With nothing much going on the world of chess, I’ve been playing against my computer a lot. It’s a bad habit, really. But one somewhat unexpected benefit is that I’ve gotten to play a few interesting endgames recently. When I play against Fritz, I generally set it on “Rated Game,” because that setting allows you […]

Print Friendly
Read the full article →

Reality Intervenes (Candidates Postponed)

March 26, 2020

This morning FIDE President Arkady Dvorkovich announced that he was suspending the 2020 Candidates Tournament, which had just reached the halfway mark. As most or all of my readers probably know already, this was the tournament to select a challenger to Magnus Carlsen for the world title. It’s been going on for the last two […]

Print Friendly
Read the full article →

Halfway! (Candidates, Round 7)

March 25, 2020

The only significant sports event in the world got a little bit more interesting today! In the seventh round of the 2020 FIDE Candidates tournament, the two leaders faced off against each other, and chess fans were rewarded with an exciting and decisive game. Maxime Vachier-Lagrave defeated Ian Nepomniachtchi to catch up with the latter […]

Print Friendly
Read the full article →

It’s All About Nepo Now (Candidates, Round 6)

March 23, 2020

It was a good news, bad news kind of day for Ian Nepomniachtchi and for chess fans everywhere. On one hand, he won his second straight game and put more distance between himself and his pursuers. He beat Ding Liren in a very solid Ruy Lopez where Ding’s position just spiraled downhill. This gives Nepo […]

Print Friendly
Read the full article →