The Pursuit of Chess Truth, Part One
Why you should annotate your games, even though you may never reach the end.
I’ve played a fair amount of tournament chess in my life. Not as much as some of the true chess degenerates out there, but I’ve got around 1,100 scoresheets tucked away in a filing cabinet, covering tournaments from more than 30 years. Theoretically, each of these games was a learning experience, but there’s only so much you can learn from an unexamined game. A lot of the learning comes from digging around afterwards and trying to find the truth.
I’ve annotated my own games off and on since I started playing tournament chess seriously in the mid-1990s, but three substantial rounds of work stand out in my memory: the first in late 1996, when I picked eight of my most interesting games from that year to annotate, the second in 2002, and the third in early 2009. Each flurry of annotations came when I was treading water. In 1996 I was in high school, working hard to make expert but was finding it much harder to get through class A than class B. In 2002, as a college student, I was stuck in the high 2100s and wanted to get to master and stay there. By 2009 I had plateaued pretty significantly, stuck in the mid-2200s for four years; the reward for this work was finally breaking through and gaining the FM title in 2010.
There’s one other connection between these three periods of annotating games: although I worked on a bunch of my recent games each time, in both 2002 and 2009 I returned to the first eight I’d selected in 1996 and revised what I’d done before. This means that there are three sets of annotations for each of these games.
I originally conceived of a series of posts about each of the games in question, contrasting the quality of the annotations to get a snapshot of my chess understanding at each moment in time. The problem was that my old high school annotations from 1996, done without the aid of computers, held up pretty well under scrutiny in 2002 and 2009. There would usually be a big mistake or two somewhere, but I generally understood what was going on in each game. In the end, only one of the games was so difficult that my understanding of it changed dramatically each time I investigated it.
The game in question is Steven Gaffagan (1993)-Andy Lee (1876), Berkeley Chess Club, 5/24/96. Steven’s rating was shooting up from the 1900s to the 2100s at the time, and he had a dynamic style of play that often left me flummoxed. He was playing only king pawn openings, and I always played the French in those days, so we quickly rattled off the opening moves: 1 e4 e6 2 d4 d5 3 Nc3 Bb4 4 e5 c5 5 a3 Bxc3+ 6 bxc3 Ne7 7 h4 Bd7 8 h5 Qa5 9 Bd2 h6 10 Nf3 Nbc6:
We’ve reached a pretty normal Winawer French position via a move order transposition or two and I was feeling pretty comfortable until Steven played the surprising 11 dxc5!? This kind of move, voluntarily ruining the pawn structure for future piece activity, was the sort of thing I didn’t understand at the time and was incapable of predicting. I played a couple more normal looking moves: 11… 000 12 Rh4 Nf5?! and got hit with another shocker: 13 Rb4!
We’ve arrived at the first critical moment: I decided to retreat with 13… Qc7. Let’s compare the three sets of annotations:
1996: “The only possible move. Black gets squashed after 13… Qxc5? 14 Rb5 Qe7 15 Rab1 b6 16 g4! Nh4 17 Nxh4 Qxh4 18 Rxb6!! axb6 19 Be3! and white wins. Now black distracts white from attacking on the queenside by counterattacking in the center.”
2002: “Really the only move. 13… Qxc5? 14 Rb5 Qe7 15 Rab1 b6 16 g4 Nh4 17 Nxh4 Qxh4 leaves black in terrible trouble on the queenside, since 18 Be3 is strong, as is 18 Rxb6!? axb6 19 Be3! Also bad is 13… f6?! 14 Rab1 Nxb4 15 axb4 Qc7 16 Bf4 since black can do nothing to activate his extra exchange.”
2009: ! “Black must seek quick counterplay against the e-pawn, and in many variations it helps to get the queen off the a-file. It’s too late for 13… Qxc5?? 14 g4! Nfe7? 15 Rb5, trapping the queen, and 13… f6?! 14 g4 Nfe7 15 exf6 gxf6 16 Nd4 is also quite pleasant for white.”
There’s some consensus here about the need for 13… Qc7, but it’s interesting how my reasoning about the refutation of 13… Qxc5 changed over time. My youthful desire to sacrifice pieces comes through in the 1996 annotations; by 2002 18 Rxb6 has been downgraded from brilliant to interesting (spoiler alert: it doesn’t work, although it’s interesting to try to figure out why not). You can see the influence of working with a computer on the 2009 annotations, as it cuts through the speculative variations and simply picks up material.
Let’s go just a little bit further: white defended his e-pawn with 14 Bf4, and I had another decision to make:
Perhaps this is just an extension of the previous critical moment, but we’ll call it critical moment #2 to keep things organized. My move, 14… g5, was not well regarded by my future (and now past?) self.
1996: ?! “14… f6 looks interesting, and is probably better than what I played. It ties white’s king knight down to the e5 pawn, stopping it from hopping to d4. Plus, if white plays 15 exf6 e5 16 fxg7 Nxg7 17 Bh2, black has good chances to overwhelm white’s king in the center. The text gives white too much time, by taking en passant, to keep his attack going.”
2002: ? “This wastes too much time to be really effective. Black wants to strike back in the center and possibly take the exchange on b4 when the moment presents itself. Thus, it’s better to attack it right away with 14… f6!? 15 g4 (15 Bb5? Nxe5; 15 exf6 e5 16 fxg7 Nxg7 and black’s center is imposing) 15… fxe5 (15… Nfe7?? 16 exf6 wins a piece) 16 Nxe5 Nxe5 17 gxf5 (17 Qe2? Nh4 18 Bxe5 Qxe5! 19 Qxe5 Nf3+) 17… exf5 18 Be2, and although white has a pair of good bishops, his position in the center and on the e-file is very loose.”
2009: ?! “Black naturally wants to get moving on the kingside and the center, but this move gives white a valuable tempo with which to develop his attack. 14… Nxb4?! 15 axb4 Kb8 16 Qc1 a6 17 Qa3 is also very dangerous for black, so best appears to be a safety-first approach: 14… Na5! with the dual ideas of taking the annoying c-pawn and blocking the f1-a6 diagonal with a well-timed Nc4.”
I find the divergence in these annotations really interesting. 1996 and 2002 share a similar, logical idea: time is of the essence, so let’s attack the center before we get overrun on the queenside. By 2009 I’m rejecting this plan without even mentioning it, so there must be some kind of refutation (there is: 14… f6?! 15 Bh2! Nxe5 16 Nxe5 fxe5 17 Qe2, winning the pawn back on e5 and taking over the dark squares). But 2009’s proposed solution smacks of desperation. The idea of needing to act fast remains from the previous iterations—“Black naturally wants to get moving … gives white a valuable tempo”—but nothing about 14… Na5 is fast. Perhaps more to the point, on c6 that knight was going to be able to capture the Rb4 or a future Nd4, removing a key attacker. On a5 it will do nothing after 15 Rab1 Qxc5 16 Bd3 b6 17 Qe2.
What we’re seeing in each set of annotations is my denial that black stands significantly worse. A more honest appraisal recognizes that 14… g5 is a perfectly good move that works to open up the kingside and destabilize the center. It doesn’t solve all black’s problems because there is no move that can do so.
The game continued with a flurry of captures: 15 hxg6 e.p. fxg6 16 g4?! (all three sets of annotations make a compelling case for 16 Nd4!) 16… g5 17 gxf5 gxf4 18 fxe6 Bxe6 19 Nd4!
Here’s our third critical moment. I played 19… Nxb4, a move that I had up to this point avoided.
1996: “It might have been a little better to take on e5 first with the queen, but black might as well take the exchange. If 19… Qxe5+ 20 Kd2! Nxd4 21 cxd4 white is still on the attack.”
2002: ? “Black still cannot afford to open the a-file and fix up white’s ugly c-pawns. Better is 19… Qxe5+ 20 Kd2, but here it is hard to find good moves, since 20… Nxd4? 21 cxd4 is positionally terrible and 20… Nxb4? transposes to the game. However, black can play actively with 20… f3! 21 Qb1 (21 Nxc6 Qg5+ 22 Ke1 bxc6 and black is fine, since 23 Rab1 Qe5+ is a draw) 21… Na5 22 Rb5 (22 c6 b6 23 Rxb6 Nc4+ 24 Bxc4 dxc4 is good for black, who can now sac on d4) 22… Nc4+ 23 Kd1 (23 Bxc4? dxc4 24 Rxb7 Rxd4+! 25 cxd4 Qxd4+ 26 Kc1 Qf4+ 27 Kd1 Rd8+ and wins) 23… Rh7 and black is still very much in the game.”
2009: ? "Finally giving in and granting white his ideal attacking setup; now it appears that white is winning by force. After 19… Qxe5 20 Kd2 Bd7, black is still fighting (the main alternative, 20… f3 21 Nxc6 Qg5+ 22 Ke1 Rde8 23 Nxa7+ should be a hair-raising draw, but white can also play 21 Kc1 as in the game).”
No credit to 1996 me for giving in to the inevitability of … Nxb4. At some point in the six years that followed I learned some lessons about putting up resistance in an inferior position. (Lesson #1 is to fight as hard as you can to avoid giving your opponent exactly the kind of position they want!) 2002 and 2009 don’t agree about the best way to do this. 2002 gets into the weeds and of course makes some mistakes along the way: for example, 19… Qxe5 20 Kd2 f3 21 Qb1 actually loses to 21… Qg5+. None of this seemed worth the risk by 2009. There’s something to be said for playing moves that are simple and meet the needs of the position rather than trying to mix it up.
The next few moves should come as no surprise: 20 axb4 Qxe5+ 21 Kd2:
We’re at critical moment number four, which I’ll call “to f3 or not to f3.” I said yes and played 21… f3.
1996: ! “Winning a tempo by making white bring his king back to the first rank. White will then have to move his king up to b2 to get his major pieces over to the queenside.”
2002: “Trying to gain a tempo, but white can successfully ignore this attempt. 21… Kb8 22 Qb1! f3 23 Qa2 Qf4+ 24 Kd1 is also bad.”
2009: ?! “This seems to gain a tempo, but white is happy to take a few moves to bring his king to safety. The variation 21… Kb8 22 Qb1! f3 would have been far more challenging, but after 23 Qa2! Qg5+ 24 Kd1 a6 25 Bxa6 [see diagram below] 25… Qg1+ 26 Bf1! Qxf1+ 27 Kd2 Qxf2+ 28 Kc1 Qe1+ 29 Kb2 f2 30 Nb5 Kc8 31 c6! Qxa1+ 32 Qxa1 bxc6 33 Qa8+ Kd7 34 Qa7+ Ke8 35 Nc7+ white is winning.”
We can get a lot out of considering why the evaluation of 21… f3 moved from strong to dubious between 1996 and 2009. The first reason is that neither 1996 and 2002 saw this position as critical: note the lack of concrete variations, particularly from 2002, a time when I seem to have loved long, complicated lines. 1996 me is still focused on the white’s attack as something that can be solved through gaining a tempo; 2009 me is aware that white’s positional advantages aren’t going away and that the time white invests to bring his king to safety actually make his position quite a bit stronger.
There’s something else going on in the 2009 annotations. If you glance back at the 2009 notes to black’s 19th move, I make a pretty strong claim: “Now it appears that white is winning by force.” Having said that, I felt obligated to prove that it was true by providing the very kind of long variation that I had been trying to avoid. Here’s the position from my 2009 analysis after 25 Bxa6:
I remember getting the dreaded 0.00 from the computer here, and pushing it to find 25… Qg1+ 26. Bf1! Now the black queen is deflected from the g1-g5 perpetual check route and the white king can find safety on b2. There’s only one problem: 25… bxa6! 26 Qxa6 Qg1+ and the perpetual works just fine. It turns out that I bought into my own narrative, that 19… Nxb4 was losing, and let it cloud my assessment.
Neither side had much choice but to improve their king safety in the moves that followed. As I noted above, taking the time to do this substantially improves white’s position because he can find far greater safety for his king than black can: 22 Kc1 Kb8 23 Kb2:
We’ll stop here for now, at the fifth critical moment. This is probably the best position white’s had all game long. For just an exchange, his pieces are on much better squares, his king is safer, and he’s ready to dominate on the queenside. Part two will come out later this week with the second half of the game and more critical moments. In the meantime, feel free to suggest a defensive move or plan for black in the comments.
Andy,
I am glad to see this game that we played many years ago continues to vex you. In truth, I do not recall if you emerged victorious. I did not maintain the careful records that you have, nor do I remember the result of this game. I do faintly remember that the game was played at the Berkeley City Club, on Durant, if I am not mistaken.
A few years before this game, I played a bunch of these French Winawers with my friend, Jeff Ely, who was a graduate student at the time. The games were played on the north side of the Berkeley campus at a cafe on Hearst. Other times we would play at Brewed Awakenings on Euclid. Anyhow, the idea to play the line with h2-h4 probably followed studying the game between Kasparov-Anand (1992). I remember examining that game with Jeff, since he, like you, enjoyed playing the French Defense.
I am impressed that you keep returning to this game, seeking a sort of analytical truth. Many of the games that we played against each other in those days included some quite unusual and lively positions. When I was white, we contested the French, and when you were white, a King's Indian. For this game, you offer many concrete variations that help explain what is happening and what we may have been considering.
You mention the flurry of captures starting on move 15. I wonder what compelled white (me) to part with my dark-squared bishop. That piece is often an important trump for white in a Winawer. I cannot say whether there was a concrete reason, or that I just didn't want to retreat. White allows a passed pawn on the h-file, in a position in which white will not be able to exert any influence on that part of the board. I don't know if that was well-considered, reckless, or merely an accident.
The position after 23 Kb2 does look favorable for white--the dark squares a7 and d6 are vulnerable (from black's point of view). Of course, black would quickly trade a rook for a knight landing on d6 (all else equal). Each side's pieces are not currently engaged, but the rook on a1 already declares ominous intent.
White must press forward and attack on the queenside, as your variations suggest. Otherwise, black will win by slowly but surely advancing the h pawn down the board.
In order to defend with black, we must decide what is white's primary threat and make sure we meet it. I think black would be happy if white tried to capture the f3 pawn. Doing so would just half open another file for a black rook. Is white threatening 23 c6? No, black wouldn't blunder the queen by 23...bxc6 Nxc6 check. What about 23 Rxa7? Black must make certain that does not succeed.
Presuming Rxa7 does not succeed in this exact position, how does white improve? Likely, the white queen must enter the queenside, via b1 or d3. Putting the queen on d2 and threatening Re1 doesn't seem in the spirit of the position. Once the queen and rook line up on a7, black has a huge problem. This may be one of those positions where the only defense is counterattack! I was first drawn to pushing the h-pawn with the idea of laterally defending on the 6th rank. But that seems too slow since it takes three moves (h6-h5, Rh8-h6, and then moving the bishop from e6).
My next idea is to make use of the rook in a different way: 23...Rh8-g8. If 24 Qb1 Rg1 25 Qa2 a6 26 Bxa6 then black has 26...Rxa1, neutralizing the attack. But what if white plays 26 b5 instead? If I were defending with black, I would be concerned. If black is getting ripped apart on the queenside, the black king may be forced to flee! I am eager to see how our game unfolded.
My first idea for black would be to make a beeline for the Nd4 with one of the rooks, say Rdf8 Qb1 Rf4. I’m planning to take it no matter what you do, but after Qa2, I think black needs to play a6 first. I’m moderately optimistic that once black trades the N the white K will be sufficiently exposed that the position will be balanced.