Tuesday, December 25, 2007

How Far to Take Post-Mortem Analysis?

{I'm on the road for Christmas, and decided to post this with some minimal editing. I actually did end up taking some endgame resources with me--Karsten Muller's DVDs as well as Smyslov and Levenfish plus Chess Endgame Training by Bernd Rosen, which I may review in the near future.}

This post was partially prompted by BDK's question on how much time you should spend analyzing a game. The short answer, in my opinion, is: analyze it until you've learned about as much from it as you're going to. If it's "I blundered my queen in the opening and resigned", your work is limited: check what the best or book move would have been, understand the tactic or oversight that was responsible, resolve not to repeat it, and move on.

Then there's cases such as the following.

I'm going to show an extreme case of where analysis of a game may take you. This could have been a "throwaway" ICC blitz game, but there was a lot of information to mine from it. You get spared the scholarly dissection of the opening that I'd normally perform, because I uncharacteristically deviate from my repertoire on move 2. Instead I'll focus on how the game highlights weaknesses in my play, and suggests areas worthy of study.

Relentless Bastard-Grandpatzer, ICC 2007

1. e4 e5 2. f4 Nc6

This is quite uncharacteristic. My repertoire move is 2... Bc5. It turns out to be a happy experiment.

3. Nf3 d6 4. Bc4 Nf6 5. d3

I actually like the feel of this position. I intend to go back to playing 2...Bc5, but I feel that 2...Nc6 could be a good repertoire move. I imagine White players wouldn't be as prepared for it, although I have no idea what advice repertoire books for the King's Gambit give against 2...Nc6. This is an example of a place where you, a chess engine and a chess database could arrive at your own opening repertoire.


The logic being, "If he's willing to surrender the bishop pair, go for it." This is a common theme in these 1.e4 e5 openings. Apparently here ...Na5 is a novelty, suggesting it's not actually appropriate here. 5... Bg4 appears to be the main move. The pin strengthens e5, and it's a developing move. Challenging with h2-h3 would weaken White's kingside, yet I see Topalov has played it: Topalov,V-Morgenstern,H, Frankfurt 1997.

5... Be7 also appears to be common in this position, e.g. Larsen,B -Najdorf,M, Buenos Aires 1979.

6. Bb3 Nxb3 7.axb3 and here I ignore the threat to e5 by playing 7...Be7, which Fritz surprisingly agrees with. After 8. fxe5:

8... Ng4

Fritz prefers 8...dxe5. Both lead to positions where Black has the bishop pair, a two tempi lead in development (the customary rule of thumb is that three tempi is worth a pawn), and saddled White with darksquare weaknesses, whereas White has an extra pawn and a mobile pawn center.

Fritz surprisingly finds 8...dxe5 9. Nxe5 O-O to be about equal, suggesting a sound sacrifice.

9. exd6 Qxd6 and I don't think White had to be afraid of castling here, although the future development of the queenside pieces looks awkward. However, my opponent played 10. h3??

White already had serious darksquare weaknesses. Isn't this the sort of thing King's Gambit practitioners try to avoid? 10. Qg3+! 11. Kd2 Qxg2+ 12. Qe2 Qxh1 13. hxg4?! Bxg4 14. Nc3?

This is the first of many critical moments where Black, bit by bit, lets slip his overwhelming advantage by not finding the best move. I would be hard pressed to find a more extreme example of my thematic obtain-overwhelming-advantage-and-pee-it-all-away style of play Remember this position and compare Black's crushing advantage here to the game result.

14. Qxf3?!

The logic here was that, if I'm going to win the knight, I may as well offer to trade queens to simplify to a won endgame. 14... Bxf3! is actually clearly superior, e.g. 15. Qe3 Qg2+! 16. Ne2 (16. Ke1 Bh4+ 17. Qf2 Qxf2#) 16... Bg5 {and it's Good Night, Irene.

15. Qxf3 Bxf3 16. Nd5 Bg5+ (the bishop move looked good, but it seems that it would have been better to batten down the hatches and not allow ...Nxc7, e.g. 16...Bd8 or 16...Bd6) 17. Kc3

17...Bxc1 I felt I could let him have his knight fork, because it helps me simplify to a won endgame. 18. Nxc7+?! (18.
Rxc1) 18... Kd7 19. Nxa8

This is an instructive moment. My idea of allowing ...Nxc7 was correct, but it required a zweichenzug finesse to make it clearly good:

19...Bxb2+ ?

19... Rc8+ !? This or the immediate ...Be3 saves the bishop while the knight remains trapped. A typical "he takes, I take..." exchange series calculation oversight. Play might continue 20. Kb4 Rxc2.

20. Kxb2 Rxa8 Black is still a bishop and a pawn up 21. Rg1 Rg8

21... Bh5!? surprisingly works, because 22. Rxg7 Bg6 traps the rook.

22. c4

I seriously considered Fritz's 22...g5! in the game but I wanted to try and put some brakes on White's pawn advance first.

23. Ka3 Kc6 24. Ka4 b6 25. Rg3 Be2 26. d4 f6 27. Re3 Bh5 28. e5


Fritz-approved, but maybe too tricksy for Blitz. I didn't like the looks of 28... fxe5 29. dxe5, but it's really OK: e.g. 29... g5 30. e6 g4.

29. e6 Re7

Throughout this sequence I'm ignoring active play (pushing my kingside pawns) in favor of prophylactic play. This is a recurring theme in my games....not pushing pawns early enough. At this point I'm planning to play my Bishop to the a4-e8 diagonal and sac it if necessary for a passed pawn or two. Again, I'm in the mindset that I have the luxury of jettisoning material to make the endgame simpler.

30. d5+ Kc5 31. Rd3 Be8+ 32. Ka3 b5 33. cxb5

33...Kxb5 (Fritz found the far less emo 33... Kd6! The king blockades and Black will gain the b-pawn. The bishop sac is not essential) 34. d6 Rxe6 35. d7 Bxd7 {the planned sac of the bishop} 36. Rxd7 g5 37. Rxh7

This is the start of the rook endgame that I could spend a lot of time on. Black should be good here, but two features strike me as important: White's rook is behind the passers, while Black's defends from the side; Black's king is better poised to dash over to the kingside, but neither king wants to leave their queenside pawn and allow a passed a- or b- pawn.

37...f5 38. Rg7 g4

One factor I need to determine is whether the f- or the g-pawn leads. I know when the attacking king is there, it's typical to lead with the outside pawn and tuck the king in the short-side hole. For the defense, a typical fortress seems to be where the defending king is on f4 with a rook on the 4th rank. I have seen in endgame books that the case of a g- and an h-pawn is the most drawish, and there are fortresses like this to watch out for. I intend to bone up on this material.

39. Rg5 Rf6

Following the rule that rooks belong behind passed pawns. A cursory glance at these types of endgames indicates that there's a lot of exceptions, and in some cases having the attacking rook in front of the pawns can even work.

39... Re5 seems clearly best. The a-pawn will be protected, allowing the Black king to invade, and it also cuts the white king off from the kingside.

40. Kb2 Kb4 41.Kc2

We can see the problem emerging...White threatens to get his king in
front of the pawns. 41...Ka3 42. Kc3 a4?

I thought if I free my king (as well as his) from kingside defense that it made the win easier. It made the win harder or maybe even disappear. 42... Rc6+ (Fritz) appears to lead to the win.

43. bxa4 Kxa4 44. Kd4 Kb4 45. Ke5 Rf8 46. Kf4

Fritz is giving this a +2.38 still for Black, but this is looking like a draw {if you're not accustomed to the idisyncracies of computers, you need to realize that you can't rely on such an evaluation in an endgame. If the computer sees a material advantage, but can't see how to convert it via queening a pawn and/or mate, the line is still unclear.} Part of my homework for this game is studying these types of fortresses.

The game concluded: 46...Kc4 47. Rg7 Kd5 48. Re7 Kd6 49. Re3 Kd5 50. Re7 Rf6 51. Re1 Re6?! (or ?, depending on whether you thought Black still had winning chances) 52. Rd1+ Kc6 53. Kxf5 Re8 1/2-1/2

Compare this result to Black's position to that on move 14. Unfortunately, this game is all too representative of my style of play: obtain advantage, then fail to find the best moves and return material "to simplify to a won endgame".

Now, if you've read this far you may wonder why I said this was "extreme". Maybe an unusual amount of analysis for a blitz game, but not extreme.

What makes this extreme is that the analysis is not done, and there are several levels of "extreme" that I could justify passing through to get closer to the "truth" of this endgame.

I can't help but feel that a grandmaster would look at the position after White's 37th move and know exactly what's going on, and how each side must play. But the position is not one that you're likely to find described in an endgame book. The positions where the game has simplified down to two connected passers are more fundamental.

One less extreme option is to study these g+h-pawn endgames, see how the defender draws, and try to extrapolate to the case in my game.

A more extreme option is to open up Muller and Lamprecht, or Fine, or Dvoretsky, and work through the pertinent sections on R+2P vs. R endgames.

Another is to finally crack open Smyslov and Levenfish's "Rook Endings" and finally read the bastard cover to cover.

All of these options could be justified. Rook endgames are the most common, so even the last option mentioned isn't that extreme.

I think a characteristic of good players is that they don't put off learning something for a rainy day. They analyze their games and try to find the best moves for each side, and if something is unclear they try to figure it out. Here, I think that at least studying a few of the typical drawing fortresses is warranted, and I think I'll take this opportunity to brush up on my rook endgames in general.

I happen to have Muller's rook endgame DVD on order anyway, so I think I'm going to start with looking at the g+h-pawn endgames and then watch the DVD. But I think Smyslov and Levenfish is going to find its way into my luggage for my Christmas vacation.


Anonymous said...

Interesting analysis.

Doesn't 14... Bg5+ win the queen right away? For eg., 14...Bg5+ 15. Nxg5 Bxe2 16. Nxe2 h6 and the g5 knight falls.

After 41. Kc2, how about 41... Rc6+ 42. Kb2 Re6 43. Rxf5 Re2+ 44. Kc1 g3? The idea is to play g2, followed by Rf2, Rf1+, g1Q.

Grandpatzer said...

Thanks for your analysis, anonymous.

14...Bg5+ is also crushing, but it seems that ...Bxf3 spends less material to win the queen. In Fritz terms, we're talking something like +22 vs. +11 for black. There's crushing, and then there's squeezing-your-ass-down-to-a-neutron-star.

The actual game move is still +7-ish. What this game demonstrates is that making a series of good-but-inferior moves can make your game a lot harder to win, or even cost you the win.

Yes, my analysis after 41.Kc2 wasn't complete. Your line looks good. Fritz still liked 41...Ka3 fine, which partly explains why I didn't delve into your line, but I didn't work out proper play after my move either. In a nutshell, your check with Rc6 works in that line as well (e.g. 42.Kc3 Rc6+).

I think the theme here was that it was erroneous to think "If I eliminate the queenside pawns I eliminate White's counterplay". Because of the fortress, White doesn't need counterplay because his play (i.e. fortress) is good enough as far as I can tell. Thus the need to study these fortresses.

Anonymous said...

Talk is cheap, so I'll toss in my two cents:
Before move 22, I wouldn't say you had made any mistakes. Sure, you missed a couple of chances to gain even more crushing advantages than what you already had, but it's blitz - if you see something that leads to an obvious win it makes a lot more sense to just play that, rather than spend time on the clock trying to find something more precise (which may or may not be there). The same principle applies if you're in time trouble in OTB.
Unfortunately, your play after move 22 shows that the win wasn't obvious to you. At move 22, the proper thing would be to take a deep breath, relax, and ask: does White have any threats? You suggested pushing the queenside pawns, but how dangerous are they? White's only passer is the d-pawn but the black king stands in its way. It takes at least six moves for White to support it, e.g. d3-d4-d5, b3-b4-b5, and c4-c5-c6+. Even if you let White play all those moves for free, Black meets the check with ...bxc6 bxc6+ Kd6 and the white pawns are stopped (if White plays Rc1 and c7, just ...Rc8; trying to bring the king up takes even longer). And along the way I'm ignoring details like the e-pawn hanging to the bishop.
Simply put: White's pawns take way too long to get anywhere. Meanwhile it only takes Black's g-pawn four moves to get to g2, freezing White's rook. (If Rg3, ...h5-h4 breaks the blockade.) Then bring up the h-pawn and it's curtains.
Your problem wasn't failing 'to find the best moves and return material "to simplify to a won endgame"'; in fact, that's exactly what you did. The problem was that you haven't yet mastered the technique of winning won positions (and by "won" I don't mean some sort of theoretical win like a Lucena position, I mean endgames like these where you're up a piece for nothing). Try playing out the position after move 22 again as Black against Fritz, and repeat until you can beat Fritz with, let's say, a minute or two on your clock. And if you ever find yourself with another piece-up endgame in a blitz game that you fail to convert, try the same kind of exercise again. This kind of thing needs to be practised the same way you might practise tactics, for example: it's endgame technique at an elementary level.
By the way, these endings don't have anything to do with the "endgame knowledge" you're trying to acquire, and I don't think it would be a wise use of your time to make a serious study of R+2P vs. R. I think Dan Heisman once mentioned in a Novice Nook column how as a 2100-rated player, he once lost a game because he didn't know how to hold the Philidor draw with R vs. R+P. He didn't know because he had never NEEDED to getting to 2100. That's a little extreme, but he makes a good point: theoretical endgames aren't all that common in practice. First figure out how to win with an extra piece (which the books don't think is worth teaching you, so it's something you have to teach yourself), and then learn the more subtle aspects of the endgame later.
I'm a 2100 btw, but unlike Dan, I do know the Philidor setup.

Grandpatzer said...

Anonymous#2: thank you for your commentary. I have encountered in many masters' games instances where, after obtaining a positional advantage, the master takes time to prevent even the ghost of a chance of counterplay before going in for the kill. This was unnecessary in this instance, as you pointed out, but it explains why I tried to get all finesse-y instead of charging down the board. Your suggestion of playing such positions against Fritz is a good one.

As for studying the rook endgame: I've gotten sidetracked from chess a bit (including going through that Muller video, in part b/c I have to transfer files from my laptop to my desktop), so I'm not quite at that endgame yet. My thinking on these technical endgames is: if now's not the time to learn about them (after an actual occurrence in your own game), when is? One answer is, "not until you're x level". However, if you can actually master the endgame at your current skill level (or at least retain the main guiding principles) in minutes to hours of study, I say go for it. This goes back to my previous comments on Silman's endgame book, where he places some (to me) simple endgames and principles much later in the book than I'd expect. I also think that sometimes it's good to "punch over your weight" and push the boundaries of what you can do. If a particular technical endgame makes you throw up your hands in despair, move on. But if it's within your grasp, or tantalizingly close, give it a shot.

I think one trait of good players is that they don't leave inaccuracies encountered as "to-do" items, but instead "git 'er done!"

Anonymous said...

Preventing counterplay in technically winning positions is an important skill. However it's also important to be able to judge when it isn't necessary at all. In your game an experienced player would see quickly that the white pawns couldn't be dangerous. And if you're an avid reader of Silman, then you know you should be making use of favorable imbalances! So if "passed pawns must be pushed" then your correct plan was quite clear. Did you ask yourself "what wonderful thing does this do for my position?" when you pushed the queenside pawns? There's lots of thinking techniques one can use during a game; the trick is to use the right one at the right time.
Interesting question, when to learn endgames. Silman's an experienced coach, and if he suggests leaving some seemingly basic endgames until later, he's probably got a point. He's talking to the type of player who's trying to make maximum use of their limited studying time. Certainly if endgames fascinate you then by all means delve in; do what's fun for you. It's like studying openings; people study them often not because they don't realize it's a total waste of time, but because it's fun.