Now how would I know what Keres exactly got for letting Botvinnik win? Obviously I have no idea of knowing that, therefore I was joking.

I decided to take another look at this endgame after getting Keres’ book < World Chess Championship 1948>. The book was originally published in Estonia in 1949 under the title <Maailmameistri Turniir Haag-Moskva 1948>. It was republished, translated into English but with no other changes, as far as I can tell, by Verendel Publishing in 2016. It’s a nice book, but it’s a shame that the publisher didn’t make use of computer analysis or commission an introduction discussing the controversies around this tournament.

I used Shredder 13 running on my MacBook Air. There are much better engines and analysts on this site, so I’d be happy to be corrected.

As you can see from the kibitzing, this game, and the R+P endgame in particular, has long been a centerpiece in the argument that Keres deliberately threw games to Botvinnik in this event. Of course Keres writing in 1949 was hardly going to admit to losing on purpose, so his notes can’t settle the controversy. But they shed light on some of the things we argued about, including the level of play (from both sides!), and what he saw (and didn’t see) during the adjournment, and whether White’s position really was hopeless heading into the endgame. Keres’ notes take up 20 pages of the book, so I just post excerpts.

The game was played in Round 15, the last round of the third cycle. (Each contestant played four games per cycle — five rounds, one of which was a bye.) After beating Euwe and Smyslov earlier in the cycle, Botvinnik had suffered his first loss of the tournament to Reshevsky in Round 14. Keres, meanwhile, having scored only 50% in the first two cycles, had notched wins over Euwe and Reshevsky and a draw with Smyslov in the third cycle. As a result, going into this game, Botvinnik was in first with 8/11 points and Keres was second with 6.5. Keres describes this game as a “last attempt” to catch the leader.

I am going to skip for now his notes to the opening and middlegame. Keres’ notes are in brackets and my (and Shredder’s) comments are in plain text.

This was the adjourned position, with Keres sealing 42.Rd1-d2:

click for larger view

As I noted then, Kmoch wrote that White’s position was hopeless, and Keres agrees. <In his home analysis White convinced himself that Black has a logical strategic plan of campaign against which White has no satisfactory defense. White assessed his prospects in this endgame to be only of a practical nature, if Black perhaps did not carry out his plan accurately enough.>

42….g7-g5?

<It is very rare for Botvinnik not to play the strongest first move after home analysis. But here Black commits a bad tactical mistake by opening up the position of his king too early. In so doing, he seriously jeopardizes his win by offering White various tactical chances. Probably Botvinnik did not find a clear winning continuation for Black in his home analysis, and resumed the game without any clear thought-out plan. This circumstance should have been fatal for him.

Among other things, in his home analysis White could not see how to defend himself in an attack beginning with 42…f4!. After this Black threatens in several lines to play …f4-f3 in combination with an assault against the h-pawn, or also the maneuver …Kh7-g6-h5 followed by …g7-g5 and a subsequent …Kh4 followed by an advance of the g-pawn. The presence of the pawns on the a-file ties White’s rooks so tightly to the defence of the a-pawn that they are unable to adequately ward of Black’s attack on the kingside. It is of course almost impossible to demonstrate an accurate winning variation analytically, but by executing the attack in accordance with the general lines given above, Black could in my opinion win the game without any great problems.

If there is a win after 42….f5-f4, Shredder and I are having a very hard time finding it. Initially I let Shredder run overnight, and it wound up with an evaluation of -0.75 or so and a position that looked nothing like a win. Playing through the lines, a complicated story emerges.

Shredder responds to 42…..f4 with 43.R8d5.

If rooks are exchanged immediately, a draw seems pretty clear. The engine continues 43….Rxd5 44.Rxd5 Re6 45.Ra5 g5 46.Kg1 Kg6 47.g3! fxg3 48.Kg2 followed by Kxg3 — this is quite similar to some positions that could have occurred later on, which Keres (as well as Levenfish and Smyslov in their book on rook endgames) assess as drawn.

If Black keeps both rooks on with 43….Re4, then 44.Rc5 and if Black follows up with 44….g5, White can strike with 45.h4! If now 45….g4, then 46.Rd7+ Kg8 (46….Kg6?? 47.h5#) 47.Rc8+ Rf8 48.Rcc7 forces an immediate draw. If 45….gxh4 46.Kh3 f3 47.Rc7+ Kg6 48.Rf2 Ra4 49.Rxf3 Rxf3+ 50.gxf3 Rxa2 51.Kxh4 looks like a fairly straightforward draw.

But if 43….Re4 44.Rc5 Rfe6 45.Ra5, it gets murky. One plausible continuation is 45….R4e5 46.Ra4 Rf6 47.h4 g5 48.hxg5 hxg5 and now White can go after the a-pawn with 49.Rd7+ Kg5 50.Ra7 Kh5 51.R4xa6 Rff5

click for larger view

How can that be anything but a draw? Black is able to make some progress, though, after 52.Ra3 Rd5 53.Rh3+ Kg4 54.Raa3 Rd2. Black’s threat is to eventually double rooks on the second rank. At least in Shredder’s lines, if White doesn’t give the a-pawn back, he eventually winds up in a losing zugswang. But if he does, eventually a pair of rooks gets exchanged and we wind up in an endgame roughly like this:

click for larger view

Lamprecht and Mueller have several examples of 2:1 rook endings with no passed pawns, and I came away thinking that they were harder to draw than I had realized. But I’m sure Keres could hold this.

To sum up, I’m not convinced the adjourned position was really lost for White. But I’m not confident it was drawn, either.

43.g2-g4!

<Naturally White immediately utilizes the chance to exchange off one pair of pawns, which considerably helps his position. It goes without saying that Black cannot play 43….f4? due to 44.R2d7+, and thus he has to accept the exchange of his f-pawn.>

43….Rf6-f7

<After 43….Kg6 White does not need to transpose to the variation given in the previous note 44.R2d7 Re2+, as instead he has the stronger continuation 44.Rg8+ Kf7 45.Ra8 at his disposal, with various threats. The exchange 43….fxg4 44.hxg4 deprives the black king of the important penetration square h5, and is favorable only for White. Hence the only defence, besides the text move, that came into consideration for Black was 44….Re7. But White would answer this by 44.R8d7 Rxd7 45.Rxd7+ Kg6 46.Kg3 f4+ 47.Kf3 and obtain a position where no forced win can be seen. The same situation however also arises after the text move, so it is very difficult to decide which line offers Black better practical winning chances.>

click for larger view

44.Rd8-d7 Kh7-g7 45.g4xf5

<This exchange is sooner or later forced, since Black threatened in several lines to play the very strong ….f5-f4. >

45….Re5xf5

click for larger view

46.a2-a3?

<White had come to resume an adjourned position that promised him only small chances of survival. However, suddenly he had achieved excellent drawing chances, and spent too much time on the first moves and once again was in time-trouble. The text-move naturally did not even merit consideration, and should immediately have placed White in a clearly lost position.

But before going into the details, one more general note: the question of whether this endgame is a win or a draw will be decided by which player is first to take control of the squares a3-a5 with his rook. By knowing this idea, the following part of the game will be much easier to understand.

As Black threatens to immediately play 46….Ra5, White’s reply 46.Rxf7+ is almost forced. If now 46….Rxf7, then 47.Kg3 and we have achieved the same position as occurs a few moves later in the game, only with the irrelevant difference in pawn position between a2 and a3. Hence 46….Kxf7 is stronger, followed by 47.Rd6! Rf6 48.Rd7+!.

If now 48….Kg6, then 49.Kg3 Rc6 50.Rd3! followed by Ra3; if however 48….Ke6, then 49.Rb7 Kd5 50.Kg3 and again White has excellent defensive resources at his disposal. Finally 48….Ke8 49.Rh7 also does not offer Black genuine winning chances.

From this we can draw the conclusion that with 46.Rxf7+! White could have forced a position that occurred a couple of moves later in the game.>

The critical question here is whether Black can win after Keres’ 48….Ke6 49.Rb7 Kd5 50.Kg3 — none of the other lines promise him anything. Here is the position:

click for larger view

After 50….Kc5 Black has a simple plan: win the White a-pawn and queen his own. I (and more importantly Shredder) don’t see a defense, e.g. 51.h4 gxh4+ 52.Kxh4 Rf2 53.a3 (more intuitive moves also seem to lose) Rf3 54.Rc7+ Kb5 55.Rb7+ Ka4 56.Rb6 a5 57.Kg4 Rxa3 58.Rxh6 Rc3 is a tablebase win; provided Black plays sensibly in these lines the White king seems to wind up too far away to save the game. If 51.Rb3 Rf4 (a key resource) 52.Ra3 Kb5 53.Rb3+ Rb4 54.Re3 a5 55.Re6 Rh4 56.Re3 a4 57.Re5+ Kb4 58.Re3 Rc4 59.a3+ Kc5 60.Rd3 Rc2 61.Kg4 Kc4 62.Re3 Rc3 . If White shuttles his rook between b7 and b8, Black pushes his a-pawn to a4 and then plays …Rd6-d2, winning. 51.Re7 a5 52.Re3 Rf4 is similar to what we’ve already seen: the black king and a-pawn advance, and at the proper moment the black rook either wins white’s pawn or shields the king’s path to it. The rook always has h4 available to defend his own h-pawn.

So unless I am missing something, the position at the diagram is a Black win.

46…..Rf5-f2+?

<Here 46….Ra5 would have won without any great difficulty. For example: 47.Rxf7+ (47.R7d3 Rf4 followed by …Kg7-g6-h5 would significantly change matters.) 47….Kxf7 48.Rd3 Kg6 followed by …Kh5 and now the maneuver …a6-a5-a4 followed by a rook transfer to b3 decides the game no matter whether White exchanges a couple of pawns on the kingside with h3-h4 or not. A similar position also occurs later in the game.>

47.Kh2-g3?

<Here 47.Rxf2 Rxd7 48.Rc2 Rd6 49.Rc4! would lead to a drawn position, as later occurred in the game. If however Black, instead of 48….Rd6, plays 48….a5, then 49.Rc4 Rd2+ 50.Kg3 Rd3+ 51.Kg4 Rxa3 52.Rc7+ Kf8 is no longer sufficient for a draw, because now Black can get his rook to h4 with check and later defend the a-pawn by playing …a5-a4.

After 48….a5 White could secure the draw with tactical tools by playing 49.h4!, since 49….Rd4 (or 49….Rd3) 50.hxg5 hxg5 51.Rc5 would attack two pawns at the same time, and after 49…Kg6 50.hxg5 hxg5 51.Rc5 Black must place his rook in a passive position to hold the a-pawn. Also insufficient for a win is 49….g4 50.Rc5, for example: 50….Rd3 51.Rxa5 Rh3+ 52.Kg2 and White has ensured a draw.>

48….Rf2xd2?

I<t is curious to observe how many mistakes are made in this rather simple endgame during the span of two to three moves; and this in a position that had been analyzed at home!

Here Black naturally had to proceed with 47….Rf3+ 48.Kg4 Rf4+! (After 48….Rxa3 49.Rxf7+ Kxf7 50.Rd7+ Ke6 51.Rh7 Ra4+ 52.Kg3 White achieves a drawing position, which could also have occurred in the game.) 49.Kg3 Ra4! and wins, as been shown in the note to Black’s 46th move.

Now a drawn position occurs again.>

In this case the double rook variation after 49….Ra4 is worse for White than the comparable variation Keres gave at Black’s 46th, so it does appear that White can’t hold it.

48.Rd7xd2 Rf7-c7

<The attempt 48….Rf4 49.Rd7+! Kg6 50.Rd6+ Rf6 51.Rd5, followed by Ra5, would also hold the draw. For example: 51….Rc6 52.Ra5 Kh5 53.Kg2 and there is evidently no way for Black to strengthen his position further. The text-move also fails to offer substantially better prospects.>

click for larger view

49.Rd2-d4!

<Despite the time-trouble, White comes up with an interesting drawing continuation, which however does not seem to be the only one. 49.a4 Rc6 50.Rd5 Kg6 51.a5 also looks possible although this line is not so clear, as Black can still make some winning attempts with a king sally to the queenside. The text-move is simpler.>

Keres appears to be right that 49.a4 also holds the draw, which should make it a little less surprising when he plays it a move later.

49….Rc7-c6

<After lengthly thought, Black decides to reject the pawn sacrifice offered to him, since the endgame arising after 49….Rc3+ 50.Kg4 Rxa3 is drawish.

White however does not play 51.h4?, as advocated by Boleslavsky, due to 51….gxh4 52.Kxh4 Rf3! followed by ….Rf6, and 52.Rd7+ Kf6 53.Kxh4 (53.Rd6+ Ke5 54.Rxh6 h3) 53….Re3! 54.Kh5 Re7 55.Rd2 Kg7 would not work, as in both cases Black has a winning endgame.

But the draw could have been obtained after 49….Rc3+ 50.Kg4 Rxa3 with the continuation 51.Rd7+! Kf8 52.Rh7 Ra4+ 53.Kg3, and if now 53….Rh4, then 54.Ra7. In any event, Black loses back a pawn, and White would not have any difficulties holding the draw.>

50.a3-a4?

<Even though this move still does not throw away the draw, it makes White’s defence considerably more difficult. The natural move 50.Ra4!>

click for larger view

<would have secured a simple draw. After this there was no need for White to hurry with playing h3-h4, for example 50….Kg6 51.Ra5 Rc3+ 52.Kg2 Rc6 53.Kg3 and Black cannot make further progress, or also 50….Kf6 51.h4 Kg6 52.hxg5 hxg5 53.Kg4, or finally 50….Kf6 51.h4 gxh4+ 52.Kxh4 Ke6 53.Kg4 and White draws easily.>

Note that after 50….Kg6 51.Ra5 Kh5 White can simply play 52.Kg2, because if 52….Kh4 53.Ra4+ and the king has to go back again.

<Boleslavsky’s recommendation to play 50.h4 is, on the other hand, of much more dubious value, since Black would play 50….Rc3+ 51.Kg4 gxh4 52.Kxh4 Rxa3 and retain practical winning chances due to White’s king being cut off too far away from the back rank.>

Indeed the position after 52….Rxa3 is a tablebase win for Black. But if White remembers to throw in 52.Rd7+, then 52….Kf8 53.Kxh4 Rxa3 is a draw!

It <is> amazing that Keres missed 50.Ra4. But it becomes less amazing when we remember that Kmoch and Boleslavsky missed it too, and less amazing still given that both Keres and Botvinnik had repeatedly missed the idea of getting a rook to the a-file in front of their own pawn earlier in the ending. Also, as Keres demonstrates, 50.a4 really isn’t that bad.

50….Kg7-g6 51.h3-h4

<It was also possible to proceed along the lines of the note to White’s 49th move by playing 51.Rd5 Kh5 52.a5 Rc3+ 53.Kg2, and Black cannot make further progress as he is forced to play the retreat 53….Rc6 to thwart the threat of 54.Rd6. But after 54.Kg3 Kg6 55.Kg4 Kf6 56.Kh5 White would have built such a strong position that Black scarcely has any winning chances.

The text move is also still possible, but it complicates White’s defence even more by clearing the route for Black’s king to penetrate through h5 and h4.>

If Keres had found 51.Rd5 and 52.a5, his 50th move might have gone unnoticed. But now White’s position gets dicey. If I was in charge of punctuating this game, I might take the question mark away from White’s 50th move and put it on his 51st. (Maybe I’d give 50.a4 a “?!”.)

51….Kg6-h5 52.h4xg5 h6xg5

click for larger view

53.Rd4-d3?

<The mistake that finally ruins White’s position. 53.Rd5! would still have secured a draw, for example: 53….Rc3+ 54.Kg2 Kh4 55.Rd6 a5 56.Rd5 g4 57.Rxa5 Rc2+ 58.Kf1! Kg3 59.Ra8 and White draws. But White was unable to accurately calculate this possibility in time-trouble, and thus decided to choose the passive defence played in the game. However, in rook endgames such a tactic almost always leads to defeat.>

Levenfish and Smyslov continued Keres’ line in their book on rook endings: 59….Rc1+ 60.Ke2 Kh2 61.a5 g3 62.Rh8+ Kg1 63.a6 Ra1 65.Rh6 g2 66.Rg6=. <euripides>, kibitzing back in 2005, pointed out that if Black tries to win from here using the Lucena method, he fails short by a single tempo. Keres’ explanation for not playing 53.Rd5 — he couldn’t calculate the lines in time pressure — certainly seems plausible.

53….Rc6-c4 54.Rd3-a3

<White’s rook now stands so unfavourably that is is unable to organize efficient counterplay against Black’s pawns. Black wins the endgame easily by bringing his king over to the queenside. Of course, White could have put up more stubborn resistance at some points, but his position is lost in any event.>

I omit most of Keres' remaining notes. The game was adjourned for the second time at move 63; as he points out Botvinnik didn’t commit to the decisive transfer of his king to the queenside until he had the chance to analyze at home. At move 64, Keres notes out that 64….Rxa4? would have allowed 65.Rb5+ Kc4.Rxg5 Ra1 67.Kh2! with a draw (the tablebases confirm that 67.Kh3 or 67.Rf5 also draw, but everything else loses). Anyway, that resource may explain the odd-looking 67.Kg3-h2 in the game. Of course it made no difference what he played at that point.

As you can guess, I believe Keres’ errors at move 50 and 53 are understandable in context. At move 50, the problem was several reasonable alternatives (including the one he chose, which was quite good enough as we’ve seen). At move 53, the problem was making a committal move in time pressure. In all the words that have been spilled on this game, his 51st move hasn’t gotten a lot of attention, but (as Keres pointed out) it greatly complicated White’s defense. The drawing lines at move 50 and 51 (whether White plays 50.Ra4 or not) are fairly simple and easy to understand; the drawing variation at move 53 is long and difficult.

Looking back earlier, I still don’t know if the adjourned position was won for Black. It looks like Keres’ preferred continuation at move (46.Rxf7+) loses, but I’m not sure the game continuation (46.a3) does. It appears that both sides erred at move 47.

I’ll give Keres the final word on the endgame: <It is just a pity that both sides made so many mistakes in such an interesting rook endgame that the lion’s share of its details were lost on the spectators.>

I may post Keres’ notes for the opening and middle game at some point. Like the endgame, the opening and middlegame were tense, hard-fought, and with more than their share of mistakes.

< But White was unable to accurately calculate this possibility in time-trouble, and thus decided to choose the passive defence played in the game. However, in rook endgames such a tactic almost always leads to defeat.>

Basically, it says he knew deactivating the rook was lost, but he played it anyway because he could not see 53 Rd5 far enough.

I was curious though about Keres note about 46...Ra5 being easily won, so I left SF on all night.

<Here 46….Ra5 would have won without any great difficulty. For example: 47.Rxf7+ (47.R7d3 Rf4 followed by …Kg7-g6-h5 would significantly change matters.) 47….Kxf7 48.Rd3 Kg6 followed by …Kh5 and now the maneuver …a6-a5-a4 followed by a rook transfer to b3 decides the game no matter whether White exchanges a couple of pawns on the kingside with h3-h4 or not. A similar position also occurs later in the game.>

Following <keypusher> 46...Ra5 47 R7d3 Ra4 (SF sees this as significantly better than 47...Rf4) 48 Rb2 a5 49 Rb8 Rf6 50 Rd5 Rxa3 51 Ra8 a4 52 Rd4 Rff3 53 Rdxa4 Rxh3+

click for larger view

11.08/41

The engine's decision to abandon the a pawn with 50 Rd5 looks rather weird, so I ran it again, and noticed that 50 Rd5 which was about a tenth choice gradually moved up by attrition as the sturdier looking ones like 50 Rbb3 pushed past -3.5

Part I

OK, here are Keres’ remaining comments, in brackets, with comments from me and Shredder in plain text. I should say I didn’t run Shredder for as long or try to figure out the “truth” to the same extent in the opening and middlegame as I did in the ending. Keres described the overall course of the game as follows:

<The key game in the last round of the third cycle was Keres-Botvinnik. White made a last attempt to catch up with the leader of the tournament. A French Opening gave rise to a tense middlegame where White maintained a slight initiative, and finally forced a position with a strong knight against Black’s inefficient bishop. But when trying to make use of his advantage, White continued weakly and enabled Black to open the position; later White also lost a pawn. The following time-trouble was rich in mutual inaccuracies until the game finally reached an approximately equal endgame. But a few moves before the time control, White made another mistake and thereafter had to adjourn in a rook endgame a pawn down, with poor chances to save the game.

However, when resuming the game, Botvinnik did not choose the strongest continuation, and White was once again able to equalize. But the second batch of time-trouble finally ruined White’s position completely. Black activated his rook and forced White’s counterpart to take up a passive defence, and Black steadily won the endgame after it had been adjourned one more time.>

That is a fair summary, except I would say Shredder evaluates the opening less favorably for White than Keres does. As Keres points out, Botvinnik missed a clear win at move 28; the world could have been spared a very durable endgame controversy if he hadn’t.

1.d2-d4 e7-e6

<Botvinnik almost always chooses this modest move against 1.d4, thus he keeps the possibility open to transpose into either the French or Dutch Defences. As Botvinnik plays both these openings, this reply suits his playing style well.>

2.e2-e4 d7-d5 3.Nb1-d2

<With this method of play, White usually gets a lasting initiative without any greater risk. Therefore I abstained from the sharper 3.Nc3, which I first employed in the final round.>

3….c7-c5 4.e4xd5 e6xd5 5.Ng1-f3

<In the game Euwe-Botvinnik from the second cycle, where some explanations have been given about the moves made so far, White played 5.Bb5+ Nc6 6.Qe2+ Qe7 but did not achieve any significant advantage after the exchange of queens. The text-move is certainly more interesting, and gives rise to a more tense position where immediate simplification is not possible.>

5….a7-a6

<A novelty, which however seems to be more dubious for Black than the normal moves 5….Nf6 or 5….Nc6. It is true that Black in some lines prevents the troublesome check Bb5+, but in so doing he loses a valuable tempo and enables White to gain a clear edge in development.>

According to the database, Botvinnik’s novelty became fairly popular, though it never supplanted …Nc6 or …Nf6. Practitioners include Korchnoi, Andersson, and Dreev.

6.d4xc5?

<With this exchange White gives back the tempo Black lost on his previous move, and is not getting any significant advantage besides isolating the d-pawn. Hence normal development of the pieces by 6.Be2 Nc6 7.0-0 Nf6 8.Re1 looks stronger, when Black has to overcome much greater difficulties than in the game.>

According to the database, 6.dxc5 and 6.Be2 are about equally popular responses to Black’s fifth move, but 6.Be2 scores a lot better.

6….Bf8xc5 7.Nd2-b3 Bc5-a7 8.Bc1-g5

<Again weak, since after Black’s natural reply, White is forced into a time-consuming knight manoeuvre that deprives him of any prospects of gaining a genuine advantage.

If White wanted to employ the system of development utilized in the game, then it would have been better to immediately play 8.Be2 Nf6 9.Nfd4 followed by Be3, also keeping the possibility open of playing 0-0-0.>

8….Ng8-f6 9.Nf3-d4!

<9…..Bxf2+ followed by …Ne4+ was threatened, and after the natural defence 9.Bd3 the pin after 9….0-0 10.0-0 Bg4 would be very troublesome for White. With the text move White places his knight on a strong central post, and poses Black relatively big problems.>

9….0-0 10.Bf1-e2 Qd8-d6

<Black naturally wants to free himself from the troublesome pin as fast as possible, and in addition he carries this out with gain of tempo due to the threat of …Ne4. But owing to his 9th move, White has the good retreat square e3 in reserve for his bishop.>

11.0-0 Nf6-e4 12.Bg5-e3 Nb8-c6

<With this move Black is not threatening so much to play 13….Bb8, which White could answer quite well with g2-g3, but rather 13….Ne5, centralizing both knights and threatening to make an eventual sally to c4. White’s following exchange is thus more or less forced, but nevertheless it secures him an excellent game.>

click for larger view

13.Nd4xc6! Ba7xe3!

<With this small tactic Black saves himself from things getting even worse, since after 13….bxc6 14.Bxa7 Rxa7 15.c4 White would stand somewhat better. Now White gets a pawn weakness on e3 in his camp, which compensates Black for the weak central squares on d4 and c5.>

Continuing Keres’ line, Shredder sees the position as quite equal after 15….Re7 16.cxd5 cxd5 17.Bf3 Rfe8. I’m curious whether other engines would agree, or if Shredder tends to evaluate the isolated d-pawn favorably.

14.f2xe3

<After 14.Qxd5 the continuation 14….Nxf2! 15.Qxd6 Ne4+ 16.Kh1 Nxd6 17.Ne7+, followed by Nxc8 and Bd3, gives rise to an approximately equal position. The text-move, on the other hand, retains the tension in the position and still offers White a slight initiative.>

14….b7xc6 15.Be2-d3?

<White has prospects of getting an attack on the kingside, but the bishop is unfavorably placed on d3, and enables Black to carry out the …c6-c5 advance with gain of tempo due to the double threat. In addition, Black’s knight is much more unpleasantly placed on f6 for White than it is on e4.

Instead 15.c4 was much stronger, with the main threat being to play 16.Qd4 followed by c4-c5. This would have secured White a long-term initiative, whereas now after the text-move Black equalizes without any greater effort.

Another good continuation for White was 15.Rf1-f4.>

Again, Shredder thinks it is even after 15.c4. A sample line is 15….Rb8 16.Bf3 dxc4 17.Bxe4 Qxd1 18.Rfxd1 cxb3 19.Bxc6 bxa2 20.Rxa2 Rb6. The engine thinks White is slightly worse after 15.Rf4 Re8 16.c4 Nf6 17.Qd4 Qe7, bearing down on the e-pawn.

15….Ne4-f6

<Naturally not 15….f5? 16.Bxe4 fxe4 17.Rxf8+ followed by Qd4 with a clear edge for White. But with the text-move Black threatens to play 15….c5 and also in a number of lines to play ….Ng4, creating some difficulties for White.>

16.Qd1-e1

<According to Boleslavsky, this move is a mistake that once again puts Black in difficulties. But as a matter of fact, the lines that occur in the game are so complicated that it is very difficult to say whether Black’s attack is sufficient to balance the position or not. In any case, a very lively middlegame now takes place where both sides operate with a number of hidden threats and manoeuvres.

Instead of the text move, Boleslavsky advocates 16….Re8, and believes it would yield Black an edge. However, this opinion is far too optimistic, since 16….Re8 would also give rise to complicated variations that are very difficult to finally assess: White can, for instance, play 17.Qh4, and if now 17….Rxe3, then 18.Rxf6! Qxf6 19.Qxh7+ Kf8 20.Qh8+ Ke7 21.Rf1 and White has a dangerous attack as compensation for the sacrificed exchange; if Black however plays 17….h6, then there follows 18.Qf4 whereupon White at least is not standing worse.>

Here is the position at the end of Keres’ sacrificial line:

click for larger view

Shredder cold-bloodedly assesses White’s attack as insufficient, e.g. 21….Qg5 and White plays the plausible 22.Nd4, then 22….Bh3! threatening mate and the queen. After 23.Nf5+ (23.Nxc6+ Kd6 loses instantly) Kd7 24.Qxg7 Qxg7 25.Nxg7 Be6 the attack is broken. But it’s hard for a human to play that way.

In the quieter line, Shredder thinks White is a little worse after 17….h6 18.Qf4 Qe7 19.Rae1 c5 20.c4 Be6.

<The most dangerous reply after 17.Qh4 thus seems to be 17….c5, which White answers most easily by 18.c4 in order to after 18….Rxe3 again proceed along the lines give above by playing 19.Rxf6!.

But if Black would immediately play 16….c5, then besides 17.c4 there could also follow 17.Qh4 h6 (17….Re8 leads to the variation given above) 18.Nxc5 Qxc5 19.Rxf6 gxf6 20.Qxh6, and White’s continuing attack assures him of at least a draw.>

17.Qe1-h4 f7-f5

<After 17….h6 White would not play 18.e4, as suggested by Boleslavsky, because after 18….d4 Black would have an excellent game due to his control of the e5-square, but instead 18.Rf4 followed by 18….g5? 19.Rxg4 or 18….Nxe3 19.Qg3 g5 20.Rxf7.

With the text move Black in some lines threatens the unpleasant ….Rf8-f6-h6.>

After 17….h6 18.Rf4 Shredder suggests 18….c5!, which seems to maintain a slight advantage for Black. After 19.Rxg4?! Bxg4 20.Qxg4 c4 21.Bxc4 dxc4 22.Qxc4 Rac8 Black’s rook will prove superior to White’s knight and two pawns. 19.Be2? Nxe3 20.Qg3 Nxc2 21.Rc1 c4 is even worse for White. His best option seems to be 19.Qg3 Nf6 (19….c4? 20.Rxg4 ) 20.Be2 and Black is a little bit better after 20….Re8, 20….Ne4, or 20….c4.

This is a good illustration of the difficulties human annotators labor under -- Keres was very aware of the possibility of …c6-c5 and examined it at multiple moves. But it slipped his mind here, causing him to misevaluate the position. I think he also slightly overestimated his game, and thought …Nf6-g4 was an inferior move, so he was not really considering the possibility that Black might be better.

click for larger view

18.Rf1-f4!

<Indirectly defending the pawn on e3, since now 18….Nxe3 would be answered by 19.Re1 Ng4 20.Rxg4, and 19.Rxg4 is threatened in several lines.>

18….Ng4-e5

<After 18….h6 White has the simple reply 19.Qg3 which, besides other things, also threatens 20.Rxg4.

On the other hand, intriguing complications could have occurred after 18….Rf6. The safest reply is certainly 19.Qg3 (19….Rh6 20.h3, or 19….Rg6 20.Rxg4, but also very interesting is 19.h3 Nxe3. If now 20.Re1?, then 20….g5! 21.Qxg5+ Rg6 22.Qh4 Nxg2 and Black wins. White must therefore play differently, for example 20.Rf3 which seems to yield him the better game.

It is true that with the text-move Black avoids all these complications, but now his attack has also come to an end, and White threatens to obtain a marked positional advantage by utilizing the weakness of the central dark squares.>

19.Qh4-g3 20.Ra8-a7

<Of course not 20….Ng6 21.Rxf5 and White retains the extra pawn.>

20.Ra1-f1 Ra7-f7 21.Nb3-d4!

<Now Black is practically forced to capture on d3, because it is no longer possible to protect the f5-pawn by other means. But after this exchange the central dark squares become even weaker, which gives White a clear positional advantage. It now becomes clear that, instead of the hazardous 16….Ng4, it would have been better for Black to try to equalize by 16….Re8, which he is no longer able to do in the game.>

Shredder thinks the game would have remained quite even after 21.…Qg6 whether White continued 21.Qh3 Nxd3 or 21.Qf2 Ng4. After 21….Nxd3 the position tilts slightly in White’s favor.

21….Ne5xd3 22.c2xd3 c6-c5 23.Nd4-f3

<Initially White had planned to play 23.Nb3 at this point, which would also have given him a clear positional advantage after a subsequent d3-d4. But, at the last moment, White for some reason decided to retreat the knight in the other direction. After an excellent middlegame White, under the influence of slight time pressure, starts to play weaker and weaker. Finally he loses all his advantage, and is even the one who gets into difficulties.>

23….Qd6-b6

<The main drawback of White’s previous move was the fact that it removed White’s last piece from the queenside and left it completely without protection, and it is there that Black now directs his attack. Removing the queen from d6 is necessary, since 23….Re7 would after 24.Rxf5 Qxg3 25.Rxf8+ Kxf8 26.hxg3 Rxe3 27.Kf2! have given White a clearly better endgame [27….Rxd3 28.Ke2], and also 23….Re8 24.Re4 Qd8 25.Rxe8+ Qxe8 26. Ne5 followed by d3-d4 is favorable for White.>

24.Rf4-h4

<An important gain of tempo. After 24.b3 Black can play 24….c4 and thus eliminate the tension in the center.>

This awkward placement of the rook seems unnecessary — Shredder thinks White is slightly better after the simple and logical 24.Ne5 Rb7 25.R4f2 Qb6 26.d4.

24….h7-h6

click for larger view

25.Nf3-e5?

<Now White has serious difficulties defending his queenside pawns. In addition, White has to consider his rook on h4, as it is not taking part in the game. The correct continuation was instead 25.Qe5!, centralizing the queen and simultaneously protecting the pawn on b2. In so doing, White could have secured his advantage, while after the text-move, on the other hand, Black gets some awkward threats.>

Shredder thinks that Ne5 and Qe5 are both fine.

25….Rf7-f6 26.d3-d4?

<A bad mistake, which not only loses a pawn but in addition enables Black to create a number of tactical threats. In time-trouble White was unable to fully assess the complications occurring after 26.b3 g5. But in actual fact they would have been by no means dangerous for him; after 27.Rhf4 followed by R4f2 Black would only have achieved an unnecessary weakening of his own king position, without getting anything in return.

After the weak text-move the scales quickly tilt in Black’s favor.>

26….cxd4

<After 26….Qxb2 White can play 27.dxc5, since 27…f4 fails to 28.Rhxf4 Qxe5 29.Rxf6.> 27….Re8 28.Nd3 Qxa2.>

Shredder slightly prefers 26….Qxb2 to Botvinnik’s move, continuing 27.dxc5 Re8 28.Nd3 Qxa2 29.Rd4 Rfe6 with an about -0.65 evaluation.

27.Rh4xd4

27….Qb6xb2 28.Rd4xd5?

<This capture should have immediately placed White in a lost position. Here 28.a4 had to be played; White would then, despite the minus pawn, have obtained quite decent counterplay. But in time trouble both sides play very inaccurately.>

Black is considerably better after 28.a4 f4 29.Rdxf4 Rxf4 30.exf4 Qd4+ 31.Rf2 Qd1+ 32.Rf1 Qxa4. Shredder prefers going into an ending with 28.Qf2 Qxf2+ 29.Rxf2 with what seem like very good chances of picking up the d-pawn and holding on after 29….Rd8 30.Rfd2.

28….Bc8-e6?

<Also not totally clear in its outcome was the capture 28….Qxa2, since after 29.Rc5! White woudl get very dangerous counterplay owing to the threat of 30.Rc7.

However, 28….f4! 29.exf4 Rxf4 would have won on the spot. If now 30.Rxf4, then naturally 30….Qc1+ 31.Kf2 Rxf4+ 32.Nf3 Be6; and if 30.Qxf4 Rxf4 31.Rxf4, then 31….Bb7 32.Rd8+ Kh7 and White cannot prevent further material losses. Finally 30.Nf3 Qxa2 also leads to a clear win for Black, as he has a better position and a strong extra pawn.

Instead White could also try 29.Qe1 (29.Qf2? fxe3), but in this case Black can, for instance, obtain a winning position with the continuation 29….fxe3 30.Rxf6 Rxf6 31.Rd8+ Kh7, since 32.Rxc8 Qxe5 or 32.Nd3 Qb6 or 32.Nc4 Qf2+ 33.Qxf2 exf2+, followed by …Bb7, are all insufficient to save White. Besides this, after 29.Qe1 Black could also continue 29….Qxa2, and retain a better position and an extra pawn.>

click for larger view

29.Rd5-d4?

<During the course of the game, White was so occupied with investigating the consequences of 28….f4! that, in time trouble, he was unable to react directly to the weaker move made in the game.

Here 29.Rd7! Bxd7 30.Nxd7 Qxa2 31.Nxf8, followed by Qe5, had to be played, and it is unlikely that Black would be able to win the resulting endgame. However, if Black would instead answer 29.Rd7 by playing 29….g5, then there would follow 30.Rd4 with the threat of 31.h4, assuring White of counterplay rich in possibilities, as 30….f4 is not dangerous on account of 31.Rdxf4.>

Shredder thinks 29.Rd7 is considerably weaker than the game move.

29….Kg8-h7?

<Now Black had to play 29….Qxa2 if he wanted to retain any prospect of an advantage. White should in either case play 30.Nd7 Bxd7 31.Rxd7 and have some drawing chances in the following endgame, or 30.h4 followed by a later Qf4, strengthening his grip on the central dark squares.>

30.Ne5-d7

<Once again White was taken by surprise by Black’s previous move, and makes the first move that comes into his mind.

Apparently simpler was 30.Qf2, retaining the pawn without exchanging the knight. However, 30.h4! looks stronger, with the intention of playing 31.Rf2. If thereafter 30….f4, then 31.Rfxf4 Rxf4 (31….Qa1+ 32.Kh2 Rxf4 33.Rxf4) 32.Qg6+ Kh8 33.Rxf4 with a sure draw; if however 30….Qxa2, then 31.Rc1! and White’s attack more than compensates for the sacrificed pawn.>

Shredder agrees that White is in decent shape after 30….Qxa2 31.Rc1. But he is also OK after the text move, and material is equal.

30….Be6xd7 31.Rd4xd7 Rf6-g6 32.Qg3-f2 Qb2-e5 33.Rd7-d4

33….Rf8-b8 34.Qf2-f4

<Naturally not 34.Qxf5? Qxe3+ 35.Qf2 Rb2! and Black wins.>

34….Qe5-e6 35.Rd4-d2 Rb8-b5 36.h2-h3 Rb5-e5 37.Kg1-h2

<Simpler was 37.Rf3 Rf6 38.Qd4 and Black’s attack has come to an end.>

37….Rg6-f6 38.Rf1-d1?

<The decisive mistake, which loses a pawn and leads to an endgame where White has only problematic changes of a draw.

A natural continuation, especially in time trouble, was to play 38.Rf3 Re4 39.Qg3 and, although White’s pieces are rather constricted at the moment, he should not have any difficulties holding the draw.>

38….Re5-e4

<Naturally not the immediate 38….Rxe3 39.Rd6 with equality.

After the text-move White’s best chance was to play 39.Qd6, trying to obtain an exchange of queens. With the queens on the board, Black threatens to generate a strong attack against the enemy king.>

39.Qf4-b8 Re4xe3 40.Rd2-d8 Qe6-e5+

<Although the following rook exchange is winning for Black, it still requires overcoming large technical difficulties. Simpler therefore was 40….f4, since 41.Rh8+ Kg6 was by no means dangerous. Furthermore, Black always retained the possibility of offering an exchange of queens, and thus transposing to the endgame that occurs in the game.>

Shredder thinks both players erred on their 40th move. After 40.Rd8, Botvinnik could have continued 40….Qxa2! with the continuation 41.R1d7 Qe6 42.Rh8+ Kg6 43.Rhd8 Qe5+ 44.Qxe5 Rxe5 45.Rg8 Rf7 46.Rd6+ Kh7 47.Rgh8 f4 48.Rdxa6 f3! 49.gxf3 Re2+ 50.Kg1 Rb7 51.Ra1 Rbb2 with decent winning chances. Of course no player could reasonably be expected to calculate this in time trouble on the last move of the time control. Best for White seems to be something noncommittal like 40.Rc1 or 40.Rb2.

41.Qb8xe5 Re3xe5

And at this point White sealed 41.Rd1-d2, which brings us back to where we started.

42...f4 43 R8-d7 Ra5 44 Rf2 Rff5 45 Re7 Ra4 46 Rd2 Ra3 47 Rf2 h5 48 Rd7 Kh6 49 Rc2 Rfa5 50 Rc8 Kh7 51 Rdd8 Rf5 52 Rc2 Rf6 53 Rf2 -.98/40

Apparently the ...f3 lines disappear at higher depths. I forced one through at move 47 in this line, and this was the resulting position, with Black to move

click for larger view

-.38/42