Wednesday, September 21, 2016

TPS Report #19

I came into possession of a talisman, a touchstone that symbolizes my relationship with movies, with computer programming, and with rebelling against authority. Yes, I now have a red Swingline stapler similar to Milton's of "Office Space". I even filled it with staples even though I almost never staple things together at my desk. But I could, and that's what's important. Perhaps I will print these 19 TPS Reports, collate them into a stack, attach a cover sheet AND a memo, and staple the whole mess together, just for the fun of it. A strange sort of cognitive dissonance comes from watching myself imbue this inanimate object with magical properties. Why does such a mundane thing make me happy?

I have not been playing chess, but I have been going over my opening repertoire. I have not been training middle game tactics, but I have been annotating rook and pawn endgames. I have not been studying the games of masters, but I have been watching some broadcasts from the St. Louis Chess Club. I have not been going to the chess club, but I have been blogging about chess. It's like I am of two minds: one that is attracted by chess, and one that is repelled. Autumn tends to strengthen the pull of the chessboard on me. But I actively resist some of its basic trappings.

In tracking where my blog referrals were coming from, I found out that my old friend Temposchlucker started blogging again at the end of 2015 and went strong through every month in 2016, until August when his blog went dark again, temporarily or not. A key question seemed to awaken his passion for analysis: How does one see the invisible? Sometimes, I am surprised at what I see. It is as if my mind sees without the participation or consent of my conscious will. Neural networks, like the one used to beat the Go champion, use nodes to needle the network into producing an integrated result. Somewhere, I have neurons that fire faster when pieces are knight forks apart or related by bishop and rook pins. How do those neurons get their programming? Will Tempo or the knights or anyone else find that thing that trains the eye to see what it doesn't see?

I was somewhat inspired by the volumes of words produced by Temposchlucker and by another blogger on the "Path To Chess Mastery". But in the end, the proof is in the pudding. Can I train myself to be as good as a master? Is that even my goal any more? It is a question that seems more likely to be answered in the negative given my emotional momentum. A rejuvenation of my commitment to a younger man's goals seems unlikely. The expression is usually "Time will tell," but I wonder if it is more accurate to say "Time may or may not tell."

Tuesday, September 20, 2016

Practical Rook Endgames 16: Sack the Waterboy

The 1998 Adam Sandler football movie "The Waterboy" ended with a halfback option play. Sandler, the lowly waterboy turned punishing linebacker switches to offense and throws the winning touchdown pass to the quarterback. Sorry I spoiled it, but you had 17 years to watch it.

Theoretical generalizations seem to say that R2P versus R endings with a bishop and a rook pawn are drawn if the pawns are on the same side of the board and won if the pawns are on opposite sides of the board. After a seesaw game, I had the stronger side of the same-side ending but managed to find a way to win. This depended upon my opponent choosing the wrong time to ignore the weakness of my bishop pawn. I'm barely scratching the surface of this ending, but I thought I would share this position and its critical moment:

For the purposes of the Waterboy analogy, the f-pawn is the waterboy, the white king is the offensive quarterback, the white rook is the center or offensive line, and the black rook is the pass rush. White has just played 67.Rf6, so it's Black to move. I'm not sure why I put my rook on f6, but I think that I wanted my king to be able to lose contact with the f-pawn. My plan here was to move my king to e7 and perhaps block a Re1+ with Re6 and then try to queen the pawn. 67...Rg1+ 68.Kf4 Rf1+ 69.Ke5 Re1+ 70.Kd6.

Black to move. Here Black has to decide whether to chase the king further with 70...Rd1+ and live with 71.Ke7 or prepare some other way to get drawing chances. Actually, Rd1+ is still drawn and my opponent's choice is still okay to draw. 70...Ra1. Preparing lateral checks. The white king at this moment has no shelter from checks on the 6th, 7th, or 8th ranks. So I decided to create shelter. 71.Re6.

Black to move and draw. A critical moment has arrived. The two moves that can draw are slightly difficult to spot. 71...Ra5 and 71...Rf1 maintain tablebase draws. All other moves lose. Sample lines are: 71...Ra5 72.Re5 Ra4 73.f6 Kxh6 74.Ke7 Kg6= and 71...Rf1 72.f6 Kxh6! 73.Ke7 Kg6! 74.Re2 Ra1/b1! with lateral checks to hold. Unfortunately for my opponent, he didn't recognize that it was time to sack the waterboy. Instead, he followed my quarterback downfield to where I wanted to receive the Hail Mary pass and queen the f-pawn. The remaining positions are all tablebase wins for White. 71...Ra6+?? 72.Ke7! Ra7+ 73.Kf8! Ra5 74.f6! Ra8+ 75.Re8! Ra6 76.f7 Kxh6 Kg8! and Black resigned.

Saturday, September 17, 2016

Practical Rook Endgames 15: Vancura Caveats

Isaac Asimov's short story collection "I, Robot" was made into a 2004 movie starring Will Smith as Detective Spooner. I love the writing in spots and wish to highlight two segments: 1. Dr. Alfred Lanning's soliloquy in the middle of the movie which I have quoted in brief before:

There have always been ghosts in the machine. Random segments of code, that have grouped together to form unexpected protocols. Unanticipated, these free radicals engender questions of free will, creativity, and even the nature of what we might call the soul. Why is it that when some robots are left in darkness, they will seek out the light? Why is it that when robots are stored in an empty space, they will group together, rather than stand alone? How do we explain this behavior? Random segments of code? Or is it something more? When does a perceptual schematic become consciousness? When does a difference engine become the search for truth? When does a personality simulation become the bitter mote... of a soul?

The second passage is near the beginning when Detective Spooner compares his task to Hansel and Gretel:

SUSAN CALVIN: I don't understand. Alfred wrote the Three Laws. Why would he build a robot that could break them?
DETECTIVE SPOONER: Hansel and Gretel.
SUSAN CALVIN: What?
DETECTIVE SPOONER: Two kids, lost in the forest. Leave behind a trail of bread crumbs.
SUSAN CALVIN: Why?
DETECTIVE SPOONER: To find their way home. How the hell did you grow up without reading Hansel and Gretel?

It's obvious to me that I really have a lot to learn about rook endings. The more I learn, the less I know. I didn't give any variations in my previous post because I thought that the Vancura was a shortcut to understanding how to draw the ending like an automaton. "Put your king on g7 and your rook on f6 and check, check, check your way to a draw." My one active reader commented and gave me a Vancura variation to chew on:

Black to move and draw. Since I am too lazy to work all the secrets out myself, I went straight to my favorite 6-man tablebase robot at the Shredder site and plugged in this position. I was surprised to find out that the path to a draw, far from a mindless template, was a bit more complicated. Actually, there are only two bread crumb trails that draw: A) Ra5 and B) Rc1. A) 1...Ra5 2.Ke4:

Here again, only two moves draw, but the pathways are convergent: C) 2...Rb5 3.Kd4 Rb6! (only move) and Black can look forward to moving his rook to f6 and checking until the draw comes home. D) 2...Rc5 3.Kd4 Rc6! (only move) with a similar Vancura pattern.

Going back to the previous diagram, here is the alternate way to draw: B) 1...Rc1 2.Ke5 Rc6! (only move) and we're back to familiar territory. Staying with the first diagram, I began the search for hypotheticals. Why doesn't 1...Rf1+ work?:

White to move and win. 2.Ke5! keeps White's winning path alive. Why can't Black just force the Vancura? It's tactically unsound here: 2...Rf6? 3.Rg8+! Kf7 4.Rf8+! The rooks come off and the a-pawn waltzes in. After Kf4-e5, if only Black could play Rf1-b6 or Rf1-c6, he would be on the true Vancura path. Let's try 2...Rf7.

White can still win. Shredder shows 5 moves to victory. The simplest is the pawn advance 3.a7. Black's rook is misplaced at f7. For the rook pawn on the seventh, the defender should be behind the pawn to prevent the attacker from just sacrificing his rook e.g. 3...Re7+ 4.Rd6 Rf7 5.Rg8+ Kxg8 6.a8=Q+. Continuing this White win variation, what if after 1...Rf1+? 2.Ke5 Black tries to draw with a spate of spite checks? White makes his way to sanctuary at a7 3...Re1+ 4.Kd5 Rd1+ 5.Kc5 Rc1+ 6.Kb6 Rb1+ 7.Ka7. With the defending king so far away, White reshuffles his rook to b8, king to a8, pawn to a7, then the White king moves out through b7, chases down the checking rook, and queens the pawn.

Going back to the beginning, I now asked Shredder, why does 1...Rb1? lose? Can't Black force Rb6 and find the way home?:

White to move and win. Shredder points out an unintuitive variation starting with 2.Ra7+!. White easily wins if Black goes to the eighth rank e.g. 2...Kg8 3.Rb7 Ra1 4.a7 followed by Rb8+ and a8=Q. 2...Kf6 is more testing. Now White has to avoid the drawing 3.Rb7? Ra1 4.Rb6+ Ke7!= but instead find 3.Ke4! Ke6 4.Ra8! It looks like White is going back to a previous position, but with the Black king away from his safe base of g7, more tactical checks come into play for White. I just realized that doing the shuffle is not always straightforward either and I'll probably have to devote a short post to the a6 pawn win.

Why didn't this Ra7+ win work against B) 1...Rc1!=?:

Black to move and draw. Now Black has to find three narrow bread crumb trails E) 2...Kg6!= 3.Ra8 Kg7!= and F) 2...Kg6!= 3.Ke5 Rc6!= and G) 2...Kg6!= 3.Rb7 Rc5!= angling back toward the classic Vancura with Kg7 and Rf6. As you can see, the path with A) 1...Ra5 is much safer than B) 1...Rc1. Here is one more argument in A)'s favor:

Black to move and draw. This diagram has one subtle difference with the first one. The White king is now at f5. Black cannot draw now with 1...Rc1? because 2.Ra7+! Kh6 leaves his king just too far away to prevent the brute force plan of White moving his king to b8 and escorting the pawn in. 3.Rd7 Rc5+ 4.Ke6 Rc6+ 5.Rd6 Rc5 6.Kd7+. Instead, Plan A) still works against Kf5: 1...Ra5+! 2.Ke6 Rh5!. I know that I warned in the last post not to put the Black rook behind the Black king, but here it is necessary. 3.Ra7+ Kg8!=. 3...Kh8? and 3...Kf8 both lose to 4.Rf7! blocking the Rh6+ skewer. Black also has to know that after 3.Kd7 there's still time to get back to Vancura with 3...Rh6! 4.Kc7 Rf6! and Black is ready with checks even if 5.a7 Rf7+!=. 5.a7 Ra6? loses to 6.Kb7! Ra1 7.Rb8.

From the attacker's point of view, getting the king to e4 or e5 while the Black rook is on a1 wins even if Black is on move. With White on move from the last diagram, 1.Ke4 or 1.Ke5 lead to wins. Surprisingly, 1.Ke6? only draws if Black can find .1...Rh1!=. From the defender's point of view, his rook must be on a5 prior to White getting from the kingside to Ke4 or Ke5. From Shredder, I see now that if White can get Kd4 and Black has not gotten his rook to b6 or f6 (e.g. he still has his rook on a1 or a5), then Black is lost. So the speed at which Black can get the ideal setup with rook on f6 or b6 is crucial.

Whew! As my reader commented "The theory of rook + a-pawn versus rook is actually surprisingly deep, with lots of different techniques."

Sunday, September 11, 2016

Practical Rook Endgames 14: Vancura Candidate

"The Manchurian Candidate" was first a book in 1959 written by James Condon. It was made into a black and white movie in 1962 starring Laurence Harvey as war hero Raymond Harvey, Frank Sinatra as his superior officer Bennett Marco, and Angela Lansbury as Shaw's mother Eleanor Iselin. The plot revolves around a Communist ploy to brainwash American POWs of the Korean War and turn them into subconscious sleeper agents in a bid to seize the United States presidency. The brainwashing occurs in Manchuria. Angela Lansbury, who I know as a sweet old woman of the "Murder, She Wrote" TV series and the voice of the animated teapot in Disney's "Beauty and the Beast," really shocked me with her range as a vicious and conscious sleeper agent earlier in her acting career.

Down the cast list is Leslie Parrish who plays Shaw's love interest Jocelyn Jordan. She appears in the middle of the movie to demonstrate the significance of the Queen of Diamonds and how firmly held the brainwashing control is on Shaw. Leslie Parrish makes several intersections with my cultural life: She appears in the original series Star Trek episode "Who Mourns for Adonais?" as a crew member who beguiles the Greek god Apollo. Leslie Parrish also figures prominently in Richard Bach's love story novel "The Bridge Across Forever." During their romance, they play chess together and she eventually becomes his wife in real life. Unfortunately, the soulmate subject of the book is somewhat invalidated by the fact that the two eventually separated.

"The Manchurian Candidate" was remade in 2004 starring Liev Schrieber as Raymond Shaw, Denzel Washington as Ben Marco, Meryl Streep as Eleanor Shaw, and Vera Farmiga as Jocelyne Jordan.

Whenever I try to remember the gist of the Vancura position, I feel like I have been brainwashed. It just doesn't stick. This morning, I listened to a lecture given by GM Alexander Ipatov on the Vancura position and for now, I have some memory stickiness. Interestingly, Ipatov pronounces it "van-chur-a" while I always thought it was "van-cure-a". Ipatov introduces the position with this diagram:

The elements of Vancura are the following:

  • Material is RPvR.
  • The pawn is a rook pawn, specifically on its 6th rank, not 7th.
  • The attacking rook is usually on the rook pawn's queening square.
  • The defending king is on the opposite side of the board, preferably on his knight-2 or rook-2 square.
  • Vancura's classic position has the defending king on g7 and the defending rook on f6.
  • The attacking king is not yet touching the rook pawn.

Black's formulaic method to draw is to Plan A) check the enemy king from the f-file, especially when it touches the rook pawn, or Plan B) shuffle his rook between b6, c6, and f6 when Plan A does not seem viable. If the pawn ever goes to a7, then the black rook belongs on the a-file and the black king must stay at g7-h7. One caveat: Don't put the rook where it will run into your king on the g-file or get behind your king on the h-file. Other than dropping a rook or forgetting to check the king when it touches the pawn, drawing this ending doesn't seem too hard. The attacking rook should never leave the a-file because the a-pawn will immediately drop. The defending king should stay in contact with the square h7 so that the attacking rook can't skewer. Sometimes, the rook will check Ra7+. The king just moves to g6 or h6.

The reason this formation needs extra significance is that the weaker side can't always leave his rook on the a-file and draw, especially if the stronger king can make it to a7. If the pawn moves to a7 and the king moves to a8, then the weaker side has to run his king toward the a-pawn before the attacker can set up the winning formation with Rb8, Kb7, and Pa7. With Vancura, the defender earmarks the f-file for his rook to harass the enemy king from afar. There is basically nowhere to hide.

This method can probably be extended to the situation where the rook pawn is only on the fifth rank, possibly even earlier. This method also works to draw against an extra g- or h-pawn held by the attacker. The defender must be careful that he doesn't allow a free check with Rg8+ or something, so sometimes the extra pawn sits en prise at g5 or h5 with the defending rook just shuffling around.

Karsten Mueller wrote an article about Vancura after the 2014 Sinquefield Cup game between Carlsen and Aronian, quoting some spectator as saying "It is insulting that Aronian is not resigning." But Aronian used Vancura-like methods to get a difficult draw three pawns down. Carlsen's extra pawns were at a5, h6, and h3.

Sunday, May 15, 2016

Practical Rook Endgames 13: Tactical Flourish or Positional Bind

Iron Man 3 saw Guy Pearce join Robert Downey, Jr., Don Cheadle, and James Dale in a tag team battle between two Iron Men and two Extremis mutants.

It was fun to see a couple familiar faces. The U.S. President in Iron Man 3 is played by William Sadler, who played Heywood in "The Shawshank Redemption" and butchered reading the name of the author of The Count of Monte Cristo as "Alexander Dumbass". Also, the fake Mandarin is played by Sir Ben Kingsley, who played Bruce Pandolfini in "Searching for Bobby Fischer". Incidentally, a fellow club member who is a New Yorker pointed out that Pandolfini should have had a Brooklyn accent, yet Kingsley played him with an Irish accent. I found this article to explain how that happened.

This post is the fifth and final installment from a semifinal game from my 2015 club championship run. The game is broken into phases running - Memento-like - in reverse chronological order. I have chosen Guy Pearce movie mashups to try to solidify themes of the game phases.

In this fifth installment, we rewind the game another six moves to the point where it became a double rook ending. White to move:

White has a dangerous passed pawn on d6 and the white rooks seem well placed to make it strong. The black rooks are struggling to prevent the pawn from queening and also defending against back rank mates. The white king is pinned against the a-file for now. The most forcing variations would be to queen by force and checkmate or win a rook. Such a variation doesn't exist. However, my opponent hallucinated a promising tactical flourish. Unfortunately for him, the variation he saw was not forcing and it allowed me to gain counterplay. The game continued 32.d7 Rfd8 33.Rb1?

Black to move. My opponent confidently went to the smoking area outside the club and expressed some happy words of confidence. If 33...Rxb1?? 34.Re8+ Rxe8 35.dxe8=Q or =R mate. Similarly, 33...Rxd7?? 34.Rxb8+ Rd8 35.Rxd8 mate. So I obviously should reject those two variations. If Black plays an indifferent move like 33...h6? to create some breathing room for his king, White still has 34.Rxb8 Rxb8 35.Re8+ Rxe8 36.dxe8=Q+. The Re8 variations can be neutralized by simply 33...Kf8! and now Black threatens to take at b1 and then on d7 when White would have nothing. Similarly I could see that with my king now only one step from Ke7 to support Rxd7, the advanced passer was not going to decide this game. White avoided the exchange of rooks with 34.Rd1 Rb7

Stockfish calls this a dead draw. If White decides to hold onto the d-pawn, 35.Red2 Ke7 36.Re1+ Kf8 37.Red1, he'll probably have to settle for a repetition draw. The game continued 35.Rd6 Rbxd7 36.Rxc6 Rd2+ 37.Rxd2 Rxd2+ and we have caught up to the previous post.

Instead of 33.Rb1?, how might White have improved?

Surprisingly, 33.Rde1 seems to be the path to victory. A plausible continuation goes 33...Kf8 34.Re7. White could also transpose and start with 33.Re7 Kf8 34.Rde1. 33.Rde1 allows Black to fall into 34...Rxd7?? 35.Re8+! Rxe8 36.Rxe8 mate. After 34.Re7:

I think this is where imagination helps the master find the right way. In this position, Black is nearly in zugzwang. The black rooks and king are tied to the eighth rank preventing disaster with Re8+. The white king is still trapped on the a-file, but Re1-Re3-Re1 can waste moves while Black can only waste moves with Rb8-a8, allowing the white king to break free and move up the board on a path like Kb3-a4-a5-a6xa7-b6xc5-d6. If Black moves his kingside pawns, the white rooks might gain more loose pawns as targets. Stockfish gives best play as 34...h6 35.gxh6 gxh6 36.R1e5 Kg7 37.Rf5 Rf8 38.Rxc5 Rfd8 39.Rf5 Rf8 40.h4 and evaluates this as about +4.0 winning for White.

So there you have it, a wild, blunderful rook endgame where I was:

  • lost after 32...Rfd8 33.Rde1
  • equal after 33.Rb1? Kf8!
  • winning after 45.Re1?! Re5!
  • equal but almost losing after 45...Rc7?!
  • lost if my opponent played 51.Kd7 and I couldn't find 51...Rh8! 52.c7 Kf5! 53.Re8 Rh7+ 54.Re7 Rh8 55.c8=Q Rxc8 56.Kxc8 a6!!=
  • and equal through my last move 59...Kf6
  • when my opponent resigned in a drawn position
It was a very instructive game that taught me rook endgame lessons such as:
  • No one ever drew a game by resigning.
  • Always keep your rook as active as possible.
  • Time is of the essence, almost all the time.
  • Don't give your opponent easy decisions.
  • Rook endgames are not just tactical; look for positional cutoffs and binds.
  • Two pawns, even if disconnected, can give a rook fits
  • Don't take any exchanges for granted in calculations.

Saturday, May 14, 2016

Practical Rook Endgames 12: Position and Movement

I should go back and watch "The Count of Monte Cristo" because much of it fades from memory. I guess the gist is a grave injustice leading to prolonged imprisonment followed by escape and revenge. My header refers to "The Shawshank Redemption" which includes a line of dialog discussing "The Count." Both Shawshank and Count are about prison breaks. In this phase, the black king finally breaks free of back rank imprisonment.

This post is the fourth installment from a semifinal game from my 2015 club championship run. The game is broken into phases running roughly in reverse chronological order.

For this fourth installment, I'm going 13 moves into the past from the previous post to the moment the game became a single rook versus single rook ending.

In rook endgames, I often think of the rooks as monsters rampaging around and eating up peasants as fast as they can with the possibility that one side will end up with a decisive advantage in pawn quantity. These considerations of capturing speed often deal with how orderly the pawns are and whether it pays in tempi to attack or defend. The game continued along this line with 38.Kb3 Rxf2 39.Rxc5 Rxh2 40.Re5.

Black has gotten the material superiority, but White has several positional advantages: his rook cuts off the black king, his own white king has space to roam, and the c-pawn is the most threatening passer on the board. Incidentally, Stockfish at 30 ply sees a draw ahead. Maintaining activity of the pieces is important with all phases of the game and the ending is no exception. The rooks are weird in that they command as many squares from the edge of the board as from the center and in fact checking distance seems to make them want to lurk at the edge of the board. However, my opponent has utilized a cutoff idea which prevents my king from fighting White's passer. Here I made a decent move designed to disrupt the rook placement, but I missed two other ideas that are crucial to the position.

The first missed idea is that Black might try to sneak the h-pawn past an en passant attack. The distant passer is a strong asset and if the opponent respects it too much, the position could transform into another kind of advantage for Black. 40...h5!?.

Of course, one of the problems with this move is that White's threat of queening the c-pawn is stronger, especially when Black's king has back rank issues right now. Still drawish is 41.c5 Rh1 42.Kc2 Rh4 43.Kc3 Rh1 44.Kc2 and Black can draw by threatening to arrive at c1 or c4. This drawing idea by trying to enter on c1 or c4 is the second idea that would have been helpful to see. But the two ideas are nearly mutually exclusive. If Black got too cocky about the h-pawn, 44...h4? would be a mistake because it would prevent the threat of Rh1-Rh4-Rc4 stopping White's passer. But if White were to make the mistake of 41.gxh6 then Black does well to recapture 41...Rxh6, with plans to transfer both king and rook to fight the c-pawn and if White's threats can be neutralized there, the f-g connected passers will be very strong eventually. One of the advantages of h5 is that if White tries to oppose rooks, Black probably has a winning pawn endgame. Instead, I opted for a move that escapes back rank imprisonment and also gains a tempo on White's centralized rook. 40...f6 41.gxf6 gxf6.

Here I expected White to go back into pawn eating mode with Ra5, but he preferred the positional 42.Re1. The c-pawn in the hand is worth more than material parity. One thing to note about White's move is that Black can no longer play Rh1 and therefore the harassment of the c-pawn is now limited to squares c5-c8. Black's rooks needs more activity, so 42...Rh5 is a decent move.

White now played 43.Re6. Since he could have played this move last turn, I consider the excursion with Re1 a waste of time. However, I let down my guard at this point and contented myself with a passive placement of my rook. 43...Rc5

And I offered a draw. I expected White to play 44.Rxf6+ reestablishing material parity and allowing my king some breathing room. Both sides could nurse their passers. Mine would be more distant. But White declined with 44.Kb4 so I continued with 44...Kf7

Both sides now have rooks en prise. Since it's White's move, I can lose the f6 pawn and we'll then have a pure pawn ending with two pawns on each side. However, Black will have the advantage of the distant passer. The black king can step in front of the c-pawn rather easily, but it's not so easy for the white king to step in front of the h-pawn. Black will then be able to get to the a-pawns quicker and the head start might be decisive. So the offer of a trade is actually a tactical trap. Indeed, the actual line if White chooses to win a pawn leads to a win for Black. e.g. 45.Rxf6+?? Kxf6 46.Kxc5 Ke6 47.Kc6 h5! White will find it necessary to post his king on b7 in order to queen. That means that the h-pawn queens with check and then the black queen and king can make sure c7-c8=Q never works. Now 45.Ra6 ensures a drawish position. Instead my opponent plays an inferior move that sets up his own trap. 45.Re1?!

The idea White is playing for is to maintain the cut off of the black king so that the black rook has to fight the c-pawn by itself. However, Black can refute this plan and grab the advantage by playing 45...Re5!

White will have to content himself with a worse ending. e.g. 46.Rxe5?? fxe5 and the distant passer wins for Black. 46.Rb1 a5+ 47.Ka4 h5 and Black has no trouble pressing his advantage. Unfortunately, I was still in consolidation mode and only had eyes for Rc7 defending my scattered pawns. Unfortunately, the defender needs defending too. It could have been strongly posted at e5, but instead it is passively placed at c7 and vulnerable to attack. So the actual game continued 45...Rc7?

Of note, Stockfish still thinks this is a draw. White pressed on with 46.c5 f5 47.Kb5 h5 48.c6 Kf6 49.Kc5 f4 50.Kd6 which brings us to the previous post.

Wednesday, May 11, 2016

Practical Rook Endgames 11: Time, Efficiency, and Alternate Realities

About three months ago, a friend sent me a game where he had this position as White with White to play:

He chose to sacrifice his bishop on this move while Black's passer was on g3 and went on to win a nailbiting race in which his f-pawn queened just before Black's h-pawn queened, White forced the exchange of the queens on h1, and then White ran to capture a6 and successfully queen his a-pawn. Later, he had afterthoughts of whether he should have waited to sacrifice. His win would have been much easier had he just played f5. If Black played g2 next, he plays Bf2, and only after g1=Q, sacrifices with Bxg1+ Kxg1 f6, when his pawn is one tempo closer to queening than the played line. If Black played Kh3, then he plays Bxg3 Kxg3 and f6 when again, his pawn is one tempo closer to queening than the played line. I thought that I would advise him that logic dictates that one should always wait until the last second to sacrifice for a queening pawn, but the last second is difficult to determine if your opponent has a way to block the sacrifice. And then study of the following ending really humbled me with regard to timing and rules of thumb.

One lovely thing about chess is that you can relive the game from the recorded score. As long as the notation isn't botched, the record of the game contains the essence of what happened. There are no dice rolls or card deck draws to reconstruct. History is captured in a half-page of alphanumeric codes. The player's thoughts and calculations can be reconstructed afterward, but the logic engines can also look at relevant lines of missed or accurately evaluated variations. The variations are like parallel universes, alternate realities, counterfactual possibilities. Guy Pearce's "Time Machine" from 2002 was itself a remake of an earlier 1960 movie adaptation of H.G. Wells' novel. One of the ideas that came up is that despite Alexander Hartdegen's ability to alter some past events, he cannot change the the mortal fate of his girlfriend.

This post is the third installment from a semifinal game from my 2015 club championship run. The game is broken into phases running roughly in reverse chronological order.

For this third installment, I'm going four moves into the past relative to the previous post. After White moved 50.Kd6, I had this position to contemplate:

I could see that I was going to have to sacrifice my rook, but that my pawns stood a chance if only I could give them enough of a head start. I needed to put my rook in a position to slow down the pawn, but still sacrifice for it when needed. My main move was Rc8. I don't specifically recall if Rh7 entered into my calculations during the game, but it's instructive to go through the exercise. So bear with me as we go into a long variation that wasn't played. Analysis variation 50...Rh7:

Black's idea is to play Rh8 and Rxc8 only when necessary. But those moments arrive earlier than expected due to some tricky variations. 51.c7 Rh8 and now 52.Rb1!!

The hard-to-see threat is that Rb8 guarantees a queen. "So what?" you say. Black can still get the pawn and try to queen his own with 52...Rc8 53.Kd7 Rxc7+ 54.Kxc7 Kf5

We've seen a position very similar to this. Recall in the previous post, that White had a choice to lock out Black's passers or try to race with his own a-pawn from this position, White to move:

In the race variation, Black barely held a draw with 54.Kb7 f3 55.Kxa7 Kf4 56.a4 f2 57.Ra1 Kg3 58.a5 Kg2 59.a6 f1=Q 60.Rxf1 Kxf1 61.Kb8 h3 62.a7 h2 63.a8=Q Kg1!=. The problem for Black is that he is now 2 tempi behind: the h-pawn is on h5 instead of h4 and when he plays f3-f2, White will not have to spend a tempo moving a rook on e1 because it moved to b1 already.

The new race would look like this:

55.Kb7 f3 56.Kxa7 Kf4 57.a4 f2 58.a5 Kg3 59.a6 Kg2 60.Kb8 f1=Q 61.Rxf1 Kxf1 62.a7 h4 63.a8=Q h3 and White just needs to use checks to get his queen to g4 and Black won't have a self-stalemate resource.

So after 50.Kd6, 50...Rc8! is the only move that draws.

At this point, I noticed that White has two logical moves: c7 or Kd7. I was really relieved when the game continued 51.c7 h4 52.Kd7 Rxc7+! 53.Kxc7 Kf5 which brings us to the drawn position at the beginning of the previous post. During the game I didn't think that I could delay the capture of the c-pawn with a move like 52.Kd7 Rh8 because of 53.Re8. I didn't notice at the time that 53...Rh7+ could keep the sacrifice bid alive and 54.Re7 Rh8 might lead to repetition. It turns out that Black has a very narrow path to draw with 55.Re8 Rh7+ 56.Kd6 Rxc7 57.Kxc7

57...f3 and White has to run his king down to f2 and trade his rook for both passers. Back to the game, I really dreaded 51.Kd7

because I thought that I had no choice but to play 51...Rxc6 52.Kxc6. Compared with the barely drawing variation above, White's king is one step closer to e2 and Black's pawn is back on h5. Two tempi mean that White has the win. If White decides to go for the a-pawn race, he sacrifices one tempo because there was no advantage to play Kc6-b7 versus Kc7-b7, but chasing Black's passer with Kd5 means that Black's h-pawn hardly gets going before the f-pawn is dead. e.g. 52...Kf5 53.Kd5 f3 54.Kd4 Kf4 55.Re8 f2 56.Rf8+ Kg3 57.Ke3 h4 58.Rg8+! followed by 59.Kxf2. In home analysis, I was astounded to see the complexity of a draw beginning with 51.Kd7 Rh8!

Didn't we just refute something like this? Three diagrams ago, Black tried to save time with the narrowly losing variation 52...Rh8 53.Re8 Rh7+ 54.Re7 Rh8 55.Re8 Rh7+ 56.Kd6 Rxc7 57.Kxc7 Kf5 58.Rf8!. But here, the pawn is still back on c6. So let's try to win the game for White beginning with 52.c7 Kf5 53.Re8 Rh7+ 53.Ke6 Rxc7 54.Kxc7. Since Black had time for 52...Kf5, he is one tempo up on the losing variation and this is just enough to draw.

One more surprising idea is that Black has a very narrow path to a draw after 51.Kd7 Rh8! 52.c7 Kf5! 53.Re8 Rh7+ 54.Re7 Rh8 55.c8=Q Rxc8 56.Kxc8. Black to move and draw:

It's hard to believe, but the only move that draws is 56...a6!! Why that pawn and why just one step? With the rook on e7 already, it's worth it for White to spend the time to capture on a7 and maneuver back to b1 or behind the leading passer. Here are some sample variations if Black doesn't find 56...a6!!

56...f3?? 57.Rxa7 Kg4 58.a4 f2 59.Rf7 Kg3 60.a5 Kg2 61.a6 f1=Q 62.Rxf1 Kxf1 63.a7 h4 64.a8=Q h3. Again, White checks with his queen until he can get to g4.

56...h4?? 57.Rxa7 h3 58.a4 f3 59.Rb7! (or 59.Rc7!) Kg4 60.a5 h2 61.Rb1! Kg3 62.a6! Kg2 63.a7! h1=Q 64.Rxh1 Kxh1 65.a8=Q Kg2. Now the technique is standard to never let the f-pawn get to f2 otherwise a drawn queen versus bishop pawn ending can arise. The white queen checks its way to g4, forcing the black king to step on f2 and then the white king moves closer to the f-pawn.

56...a5?? This is the hardest to understand, but the key is that the white rook recovers some time with threats on Black's king or pawns. 57.Ra7! Kg4 58.Rxa5 f3 59.Rb5 h4 60.a4 h3 61.a5 h2 62.Rb1 Kg3 63.a6 Kg2 64.a7 h1=Q 65.Rxh1 Kxh1 66.a8=Q is just like the previous variation.

The reason why 56...a6!! escapes with a draw is that Black's f-pawn can run pretty fast if Black's rook is out of position and a6 is just such a bad spot. 56...a6!! 57.Ra7?! f3!. White can even lose with 58.Rxa6?? f2!-+. Instead, there is still time for White to draw with 58.Rb7, Rb1 and Kb7. If White tries to improve with 56...a6!! 57.Kb7, Black still draws with the just-in-time pawn race, making sure not to be caught on g2 when the a-pawn queens.

If White had chosen 51.Kd7, I don't think I could have found the saving variations of 51...Rh8 and the impossible 56...a6!! I was just lucky my opponent didn't force the issue for me to go down in flames with 51...Rxc6?? I would say that the search for improvements in time efficiency are extremely important to look for in endgame variations, probably more so than rules of thumb about when is the best time to sac for a passer.