The diagram below shows where we left off in the previous post; I’d recommend going back and checking out Part One if you haven’t read it already. Part Two is going to slow down through the remaining nine moves, as the position becomes increasingly sharp and the evaluation swings wildly back and forth.
Black is facing the kind of unpleasant critical moment that you arrive at when almost all of your pieces are more poorly placed that their counterparts. I chose 23… Rhf8, a strange looking choice at first glance given the problems facing black. Here’s what the three sets of annotations say:
1996: !? “23… Bd7! was the strongest and repels any kind of attack white wants to start, like 24 b5!? Rc8 25 b6 a6! (not 25… axb6? 26 axb6 with threats of Ra8+) 26 Ra5 Qe7! White’s best try is 24 Bb5! Bxb5 25 Nxb5 d4! 26 Nxd4 Rhf8 27 c6 Rf4.”
2002: ! “Preparing to sacrifice the exchange back to get rid of the Nd4. 23… Bd7? defends well against 24 c6, but 24 Qb1! is a crusher.”
2009: “This is a nice idea, but black is helpless against the coming storm on the queenside. No better is 23… Bd7 24 Ra5! (avoiding 24 Qb1 Qc7! 25 Qa2 a5) with 25 Qa1 to follow, or 23… Rdf8 (an important nuance that will become clear in the next note) 24 c6! Bc8 25 Rxa7!”
The evaluation between 23… Rhf8 and 23… Bd7 illustrates one of the problems I often face as an annotator. Sometimes I really like a move because it’s more aggressive or more aesthetically pleasing than the alternative. Sometimes the move that feels good to me is objectively not very good. I like 23… Rhf8 a lot and I desperately want it to be good, but the truth of the matter is that it simply loses to 24 Qb1! (for more, see the 2009 annotations that follow the next diagram). 23… Bd7 feels weak to me, but it’s a decent move that forces white to solve more problems (although my 1996 analysis is not very good). There’s also a “wrong rook” issue, as it turns out that 23… Rdf8 is stronger than 23… Rhf8. My analysis of it in 2009 doesn’t do it justice: instead of 24 c6! Bc8? black should play 24… Rf4, the consistent and logical follow-up after playing a rook to the f-file. None of these options change the overall evaluation: black’s busted, but you can always make life a little more difficult for your opponent.
We’re staying here for the next diagram:
Critical moment number six, finally one for white, who has played very well to this point. He went with 24 Ra6, another move that wasn’t on my radar:
1996: !? “Of course, white can still play 24 c6!, but both he and black are proceeding with plans of their own, rather than the ‘best’ moves. One reason is that both sides are starting to run short of time, and carrying out a definite plan is a lot quicker than working out each move separately.”
2002: !" “24 Qb1? Rf4! 25 Qa2 a6 26 Bxa6 (26 Nc6+ bxc6 27 Qxa6 Rxb4+) 26… Rxd4! begins a very dangerous counterattack in the nick of time. The text presents a lot of new problems for black to defend against.”
2009: ?! “This is a pretty shot, but it walks right into the defensive resource set in action by black’s last move. There were a number of alternatives: 24 c6 is dangerous, but 24 Qb1! is the cleanest win: 24… Rf4 25 Qa2 a6 26 Qa5! (26 Bxa6? Rxd4! is the miracle black is hoping for) 26… Re8 27 Bxa6 Rxd4 and now, since the b-pawn is protected by the queen, 28 Bxb7!”
This set of annotations (and the one before, to black’s 23rd move) give us three competing narratives. The 1996 version is hesitant to criticize either player but suspects that there were better moves available. The 2002 narrative is that we should be impressed by the skill of both players through this difficult phase of the game. By 2009 that narrative has flipped: the moves played are relatively logical but don’t fully meet the needs of the position. The reason for this difference is my computer aided discovery of just one in-between move—26 Qa5!—that refutes black’s defensive idea.
Conveniently, after Ra6 black can still proceed as planned; there’s no need to move or defend the Be6, so 24… Rf4.
Here’s another critical moment for white, who played 25 Nxe6. The variations from each set of annotations are pretty much the same (I wouldn’t recommend playing through all three of them) but the evaluations are not.
1996: ? “A critical point in the game. White has many options, some of which are: A) 25 Rxe6? Rxd4! 26 cxd4 Qxe6 27 Qxf3 Qe1 28 Qd1 [sic] 28… Qxb4+ -+. B) 25 Qa1! Rxd4 26 cxd4 Qxd4+ 27 Kb3 Qxa1 28 Rxa1 =+. C) 25 Kb3 Bd7 26 Rd6 -+.”
2002: ? “The temptation of winning the bishop and attacking the Rd8 distracts white from the attack. Other tries: A) 25 Rxe6? Rxd4! 26 Qxd4 Qxe6 27 Qf4+ Ka8 28 Qxf3 should be winning for black; B) 25 Qa1 Rxd4 26 cxd4 Qxd4+ 27 Kb3 Qxa1 (27… Qd2? 28 Rxa7 d4+ 29 Bc4! Bxc4+ 30 Kxc4 Qxc2+ 31 Kb5 Qe2+ 32 Kb6 Qe6+ 33 Kb5 appears to be no more than a draw) 28 Rxa1 is another good ending for black; C) 25 Kb3! Bd7 26 Rxh6, and white is still doing well.”
2009: ? “One mistake follows another; now black is on the verge of victory. White’s chances are hardly improved by either 25 Rxe6? Rxd4! 26 Qxd4 Qxe6 or 25 Qa1?! Rxd4! 26 cxd4 Qxd4+ 27 Kb3 Qxa1 28 Rxa1, but 25 Kb3! (another strong king move!) 25… Bd7 26 Rxh6 keeps up the pressure.”
The verdict is clear: taking on e6 is a mistake. The only question is what white should do instead. Back in 1996 I thought 25 Qa1 was best, despite the fact that this leads by force to a position in which white is clearly groveling for a draw. By 2002 it was clear that 25 Kb3 actually brings white closer to material equality while maintaining many of the things that made his position good in the first place.
It didn’t take me long to give the advantage back. The next moves were forced: 25… Rxb4+ 26 Kc1.
It’s a fun position—all three rooks are hanging! It seemed obvious to me to play 26… Re4, saving one rook and threatening … Re1. Once again, the variations are virtually identical but the evaluations are not:
1996: ?? “Remarkably, even though three of white’s pieces are hanging, this loses by force. Best was 26… bxa6! 27 cxb4 (27 Nxd8?? Qf4+ 28 Qd2 Rb1+) 27… Qxe6. Also interesting is 26… Qxc3!? 27 Nxd8 Qb2+ 28 Kd2 Qd4+! 29 Bd3 (29 Ke1? Qc3+; 29 Kc1? Rb1+) 29… Qxf2+ 30 Kc3 (30 Kc1? Qe3+ 31 Qd2 Qg1+!) 30… Qxc5+ 31 Kd2 and then either 31… bxa6!? ∞ or 31… Qf2+ = .”
2002: ?! “There’s no need to complicate the position like this. 26… Qxc3? is the most tempting, but 27 Nxd8 Qb2+ 28 Kd2 Qd4+ 29 Bd3 (29 Ke1? Qc3+; 29 Kc1? Rb1+) 29… Qxf2+ 30 Kc3 (30 Kc1? Qe3+ 31 Qd2 Qg1+!) 30… Qxc5+ 31 Kd2 bxa6 32 Qxf3 is unclear and may be better for white. The text looks like it just threatens more of white’s pieces, but 26… bxa6! 27 cxb4 (27 Nxd8? Qf4+ 28 Qd2 Rb1+) 27… Qxe6 28 Qxf3 Qe1+ 29 Qd1 Qxd1+ 30 Kxd1 Kb7 is an easy win.”
2009: ? “It’s hard to imagine that this is bad: black saves one rook while threatening to win the white queen, but it simply does not hold up tactically. It was time for 26… bxa6! (26… Qxc3? requires still more effort: 27 Nxd8 Qb2+ 28 Kd2 Qd4+ 29 Bd3 Qxf2+ 30 Kc3 Qxc5+ 31 Kd2 bxa6 32 Qxf3) 27 cxb4 (27 Nxd8? Qf4+ 28 Qd2 Rb1+) 27… Qxe6 when the extra exchange and the passed h-pawn should decide.”
It’s clear that 26… bxa6 was the best move, but there’s little agreement about the correct evaluation of 26… Re4. 1996 thinks it’s the final critical moment; 2002 and 2009 disagree. We have to look at what happened next to figure out why. First of all, both 1996 and 2002 left white’s next move, 27 Qxf3, unchallenged, commenting only on Gaffagan’s draw offer and my rejection of it, but computer-aided 2009 had this to say:
2009: ? “Another mistake, this time accompanied by a draw offer. White’s best move, 27 c6! is extremely hard to spot, as it seemingly hangs everything. However, after the text I was convinced that I was winning and was also inspired to play against my opponent’s time trouble.”
I would add a little more to this variation: 27 c6! Rc8! 28 Nc5! (the only move) Re1! 29 Nd7+ Kc7 30 Nxe5 Rxd1+ 31 Kxd1 bxa6 =. I’ve never seen a tactical sequence quite like this one, and it’s a minor miracle that it ends in a balanced endgame. It’s hard to imagine either player navigating this path without error, but it does help clarify what’s at stake.
The game continued 27… Re8, a mistake that puts white back in the driver’s seat. 1996 goes mute here, but both 2002 and 2009 give the line 27… bxa6! 28 Nxd8 Qg5+ 29 Kb2 Qxd8 and wins. I’m no longer convinced that the win is so simple, but this variation does explain why 2002 and 2009 are less critical of 26… Re4.
1996 me made a pretty common mistake, assuming many of the moves played in the game were forced and that move 26 was black’s only chance to improve. 2002 me didn’t find 27 c6 and so slightly overrated black’s chances, whereas 2009’s assessment that the real mistake comes on move 27 was closest to the truth. After 28 Rd6 there’s only one more critical moment:
I played 28… Rxe6, a move that I reacted to in each set of annotations in pretty much the same way:
1996: ?! “It would have been harder for white to win, although the result is the same, if black takes the bishop instead of the knight by 28… Re1+ 29 Kb2 Rxf1 30 Qf7 Rc8 31 Rd7 Ka8 (31… Ra1!? looks close, but not quite enough) 32 Nc7+ (32 Rxb7?? Rb8 33 Nc7+ Qxc7 34 Qxc7 Rxb7+ 35 Qxb7+ Kxb7) 32… Kb8 33 Na6+! Ka8 34 c6!! with mate coming.”
2002: ?! “28… Re1+ 29 Kb2 Rxf1 give white more chances to go wrong, but 30 Qf7 Rc8 31 Rd7 Ka8 (31… Ra1!? 32 Rxb7+ Ka8 33 Kxa1 Qxc3+ 34 Kb1 Qe1+ 35 Kb2 Qe5+ 36 Ka2! avoids the perpetual) 32 Nc7+ (32 Rxb7?? Rb8 33 Nc7+ Qxc7! and black wins) 32… Kb8 33 Na6+! Ka8 34 c6! and black cannot avoid mate.”
2009: “No better is 28… Re1+ 29 Kb2 Rxf1 30 Qf7 Rc8 31 Rd7 Ka8 32 Rxb7 Rb8 33 c6, with mate on the horizon.”
I’m including this position not because there’s some miracle that changes the overall evaluation, but because my assumption that black should capture one minor piece or the other shows that I misunderstood what’s actually going on. Once we’ve established that there’s no safe way to win back material, we should be asking if there’s anything else that can be done to keep fighting. This question leads to either 28… a5 or 28… a6, giving the king some luft and forcing white to find 29 Bh3. White’s still much better, but any win is a long way off.
After … Rxe6 the final moves of the game were forced, and led to a lovely mating net: 29 Qf8+ Re8 30 Rd8+ Kc7 31 Rxe8 Qxe8 32 Qd6+ Kc8 33 Bh3+ Black Resigns. White’s bishop, left in reserve for so long, delivers the final blow:
If you’ve made it this far, I’d like to leave you with some reasons to annotate your own games:
Annotations force you out of the narrative that existed while you were playing and towards greater objectivity. With objectivity comes an increase in humility about how very difficult chess is; my annotations from 2009 reflect some of that humility by giving shorter and more concrete variations rather than the long variations of 2002 that often contained an early error.
Good annotations reveal hidden tactical and strategic ideas that expand your view of what is possible in chess. Even here, with a game I’d analyzed many times before, I became aware of the possibilities of 23… Rdf8 and 28… a5 for the first time.
When analyzing a game played long ago, annotations can help you think about how you understand chess better than you did in the past and what areas you still find difficult. In this game, I struggled to find many of white’s resources until I started using a computer for assistance.
Despite feeling scientific at times, good annotations typically enhance the artistry of a well-played game.
Thanks for reading and indulging my strange affection for this game that I lost almost 30 years ago. I’ll probably come back to some of the other games I annotated from 1996 at some point, as they shaped how I understood chess during the pivotal stage between class player and master.
Andy,
Thank you for sharing your process for revisiting the annotations to this game. Perhaps I have some old annotations in a computer file somewhere that might shed some light on what I was thinking during the game. As I mentioned before, I didn't recall how this game ended--my only memory was that we played the game at the Berkeley City Club and that 13 Rh4-b4 was a worthy idea.
Part One left off after Black's 23rd move. Remarkably, the only white pieces that have moved off their starting squares are the knight and king!? The rook, queen, and bishop haven't budged. White is also down an exchange. But when you consider the relative safety of the kings it is clear that black is in danger. If white can crack the shelter around black's king, the game will likely be decided in white's favor.
When you resumed the game and revealed 24 Ra6, I was stunned: why does white lead with a rook that already exerts pressure on the a-file; what about the white queen or the c pawn? I was also saddened that I didn't remember such a move. Still, I thought it was a terrible move. I couldn't understand why white would threaten a bishop that can easily move or be defended.
But then I realized: if the bishop is moved, white may capture black's h pawn. If given the chance, white can play Qd1-a1 and a7 is indefensible. Since 24...Rf4 indirectly defends the bishop, 25 Kb3 (which you point out) is a pretty way to continue the game. The king guards b4, fears not the bishop on e6, and steps off the a1-h8 diagonal, readying white to play Qa1.
After all these years, this game returns to haunt me, even though I escaped on that Friday night in 1996.
Really interesting game, wow.
After recent tournaments I've been going over my games and trying to identify the main reason(s) for the outcome, with the goal of identifying what I most need to focus on as I try to get better. But I've been using an engine from the start, rather than trying to figure things out on my own first. That's probably something I should consider...