<19. Nxd4> (=-0.02)

<Engine preference>: <19. Rxd4> ( 0.69)

Displays an evaluation shift of 0.67, and notwithstanding the problems it presented Anand (who commented somewhat laconically that <19.Rxd4 is also very complicated.> - it is generally thought that Anand had prepared for <19. Rxd4>) over the board, it represents a <dubious move>.

<Weighting = <<0.5>> for a cumulative game weighting of <<0.5>>.>

<19…h5> (=-0.02)

Anand commented here that <19...Rg6!? 20.a4!? Maybe this cool computer-move is what Kramnik was planning on? (20.Bxd7+?! Kxd7 21.Nxe6+ Bd6! 22.Nf4 Rg5 is just much better for Black.; 20.Nxe6 fxe6 21.Rxd7 Kf8 22.Bd3 Be5 23.Bxg6 hxg6 24.Qc4 Ke8 25.Rh7 Bd4! this amazing move seems to hold the balance 26.Rxb7 Bxf2+ 27.Kf1 Qxb7 28.Qxe6+ Kf8 29.Qxf6+ Kg8 30.Qxg6+ Kh8 31.Qf6+ Kg8 32.Qg5+ Kh8, and despite the fact that White can check with his [queen] on whatever square he wants, no progress seems possible: 33.Qe5+ Kh7 34.Kxf2 Rf8+ 35.Kg1 Qb6+ 36.Kg2 Rf2+ 37.Kh3 Qh6+ 38.Kg4 Qg6+ with a perpetual.)>

<20. Nxe6> (=-0.02) <20…fxe6> (=-0.02)

<21. Rxd7> (=-0.02) <21…Kf8> (=-0.02)

<22. Qd3> (=-0.02)

Anand again: <Played instantly, and obviously this is what Kramnik intended when he sacrificed the piece on move 19. White covers g3 and threatens to invade on h7, so immediate action is required.>

<22…Rg7> (=-0.02)

Anand: < Here 22...Bc8? loses to 23.Rh7. And 22...f5? loses to 23.Qc3. After 22...Bxg3!? This seems to be just a forced draw 23.hxg3 h4! 24.Rd6 (24.Kf1? hxg3 25.fxg3 Rg5! wins for black.) 24...Qc5 25.b4 Qe5 26.Rd8+ Rxd8 27.Qxd8+ Kg7 28.Qe7+ Kh6 29.Qf8+ Rg7 30.Qh8+ Rh7 31.Qf8+ with a perpetual.>

<23. Rxg7> (=-0.02) <23…Kxg7> (=-0.02)

<24. gxf4> (=-0.02) <24…Rd8> (=-0.02)

<25. Qe2> ( 0.46)

<Engine preference>: <25. Qb3> (=-0.02)

Anand: <Also possible was 25.Qb3!? Kh6 26.a4 (26.Kf1? Bd5 27.Bc4 Bxc4+ 28.Qxc4 Rd2 and f2 collapses.) 26...Rg8+ 27.Kf1 Rg2 28.Qe3 Qxe3 29.fxe3 Rxh2 and the black h-pawn is just as dangerous as the white a- and b-pawns.>

<25…Kh6> ( 0.46)

Anand: <A very interesting and unbalanced position has arisen. White has two extra, and passed, pawns, but his kingside is permanently damaged and open to a direct mating attack.>

<26. Kf1> ( 0.46) <26…Rg8> ( 0.46)

<27. a4> ( 0.78)

<Engine preference>: <27. Rc1> ( 0.46)

Anand: <A good multi-purpose move, making sure the bishop and the pawn defend each other in the upcoming complications. Wrong was 27.f5? Bg2+! 28.Ke1 Bc6! 29.Qd2+ Kh7 30.Bxc6 Qxc6 31.Ke2 Qb5+ 32.Kf3 Rg4 33.Re1 Qc6+ 34.Ke2 Qc4+ 35.Kf3 (35.Kd1 Rd4 36.fxe6 Rxd2+ 37.Kxd2 Qb4+ 38.Kd1 Qd4+) 35...Rd4 and wins.>

<27…Bg2+> ( 0.78)

<28. Ke1> ( 0.78) <28…Bh3> ( 0.78)

<29. Ra3> ( 1.79)

<Engine preferences>: <29. Rd1> ( 0.78); <29. Kd2> ( 1.43).

But for a later mistake by White on move 31, this would have been the losing move by White, as it shifts the game evaluation to over 1.40 in favor of Black. It therefore constitutes a blunder.

<The move <29. Ra3> is weighted at <<2.0>>, representing a blunder. Cumulative game weighting = <<2.5>>.>

Anand found this to be a fatal mistake: <Unfortunately for White, his [previous] move allows this possibility. He would have been [better] off ignoring it by playing 29.Rd1! Bf5!? An amazing move. White is ok, but would you be able to calmly play Qf1 or h3, as he is practically in a state of zugzwang, a [primary] point being 30.Qe3? Rg1+ 31.Bf1 Qa6!>

<29…Rg1+> ( 1.79)

<30. Kd2> ( 1.79) <30…Qd4+> ( 1.84)

<31. Kc2> ( 1.84) <31…Bg4> ( 0.95)

<Engine preference>: <31…Bf5+> ( 1.84). The move <31…Bg4> is a blunder as it (temporarily) concedes the win, although the nominal evaluation shift is only 0.89, reducing the game evaluation to below 1.40, from a winning advantage to a moderate advantage.

<Weighting for <31…Bg4> is therefore <<2.0>>, representing a blunder. Cumulative game weighting = <<4.5>>>

Anand gave this a <?> commenting that <Despite being 75 minutes ahead on the clock, by now I had caught up. I wanted to provoke f3. The correct way to continue was 31...Bf5+! 32.Rd3! a) 32.Kb3 Rc1! 33.a5 (33.Ra2 Bc2+! 34.Qxc2 Rxc2 35.Kxc2 Qxf2+ 36.Kb3 Qe3+ 37.Kc2 Qxf4 38.a5 Qxh2 ) 33...Qd5+! This move wins, but there are a lot of incredibly difficult moves in the winning line. 34.Bc4 a1) 34.Ka4 Bc2+ 35.Kb4 (35.b3 Bxb3+ 36.Rxb3 Qd4+ 37.Rb4 Qa1+) 35...Qd6+ 36.Kc4 Bd1+; a2) 34.Kb4 Qc5+ 35.Ka4 (35.Kb3 Bc2+ 36.Ka2 Qd5+ 37.Bc4 Qh1 38.Qxc2 Rxc2) 35...Bc2+ 36.b3 Bxb3+; 34...Qb7+ 35.Ka4 (35.Bb5 Bc2+ 36.Ka2 Qh1 mates.) 35...Rc2!! This [quiet] move is the point 36.Ba6 Qd7+ 37.Qb5 Rc4+ 38.Kb3 Qd3+ 39.Ka2 Qb1+ 40.Kb3 Rc2 41.Ra2 Be4!! 42.Bb7 Qd1 43.Ka3 Bxb7 44.Qxb7 Rc4 45.b3 Qd6+ 46.Kb2 Qd2+ 47.Ka3 Qxa5+ 48.Kb2 Qc3+ 49.Ka3 Rc5 and finally all white resistance is broken.; b) 32.Bd3 Rg2!? 33.Bxf5 Rxf2 34.Bd3 Rxe2+ 35.Bxe2 Qe4+ 36.Bd3 Qxf4 37.a5 Qxh2+ 38.Kb1 h4 39.a6 Qg1+ 40.Ka2 Qa7 and despite the mess, blacks pawns should prevail.; 32...Rg4!? This could be best 33.Kb3 Bxd3 34.Qxd3 Qxf2 keeps winning chances for black.>

<32. f3> ( 1.90) <the losing move>

<Engine preference>: <32.Rd3> ( 0.95). The move played (<32. f3>) conceded only 0.95 centipawns but constitutes a blunder as it was the losing move, shifting the position evaluation to over 1.40 in Black’s favor.

<Weighting is therefore <<2.0>>, representing a blunder. Cumulative game weighting = <<6.5>>>

<32…Bf5+> ( 1.90)

<33. Bd3> (#21)

<Engine preferences>: <33. Kb3> ( 1.90); <33. Rd3> ( 2.85)

The evaluation shift is unquantifiable as the forced mate replaces the numerical evaluation. Although the move actually played is a blunder in the traditional sense as it would have terminated the game almost immediately, it is not considered to be a blunder for the purposes of the method employed for this project as Black’s position was already lost, and remained so for the rest of the game. Accordingly, no weight is attached.

<33…Bh3> (-4.02 )

<Engine preference> <33…Bxd3+> (#21). The evaluation shift due to White’s move is also unquantifiable as it concedes a forced mate. However as it does not concede White’s forced win, it is not considered a mistake for the purposes of this project and as such there is no weighting attached to <33…Bh3>.

<34. a5> (-4.02 ) <34…Rg2> (-4.02 )

<35. a6> (-6.30 )

<Engine preference>: <35. Qxg2> (-4.02 )

<35…Rxe2+> (-6.30 )

<36. Bxe2> (-6.30 ) <35…Bf5+> (-6.10 )

<37. Kb3> (-11.36 )

<Engine preferences>: <37. Kc1> (-6.10 ); <37. Bd3> (-8.53 )

Anand: < Or 37.Bd3 Bxd3+ 38.Rxd3 Qc4+ and wins. Or 37.Kc1 Qxf4+ 38.Kd1 Qd4+ 39.Kc1 Qe5! and wins.>

<37…Qe3+> (-11.36 )

<38. Ka2> (-15.00 )

<First three engine preferences>: <38. Kb4> (-11.36 ); <38. Kc4> (-11.36 ); <38. Ka4> (-11.57 )

<38…Qxe2> (-15.00 )

<39. a7> (-18.82 )

<First three engine preferences>: <39. Rc3> (-15.00 ); <39. Ra4> (-15.68 ); <39. Kb3> (-15.97 )

<39…Qc4+> (-18.41 )

<40. Ka1> (-23.03 )

<Engine preference>: <40.Rb3> (-18.41 )

<40…Qf1+> (-21.83 )

<Engine preference>: <40…Qc2> (-23.03 )

<41. Ka2> (-21.83 ) <41…Bb1+> (-21.83 )

White resigns

<ANALYSIS>

<Note> The fluctuations generated in the relatively low (16 minimum) ply forward slide were smoothed out as far as possible in the return slide. The corrected evaluations extracted from the return slide are used in this analysis, as they are considered more reliable than the raw evaluations generated on the initial forward slide. All moves have been evaluated on forward and return slide for completeness.

<Evaluation range>:

Between <-#21> (equal to infinity) applying to <33.Bd3> - representing a forced mate for Black - and < 0.76> in respect of the move group <<15…Bd6 16. Rd1 Rg8 17. g3 Rg4> representing a significant/moderate advantage for White.

<The largest evaluation shifts>:

- for White was #21 between <32…Bf5+> ( 1.90) and <33. Bd3>

- for Black was from -#21 between <33. Bd3> and <33…Bh3> ( 4.02)

<Computer statistics>:

• 93.9% of the ply in this game (77/82) coincided with engine preferences 1, 2 or 3

• 86.6% of the ply in the game (71/82) coincided with engine preferences 1 or 2

• 58.5% of the ply in the game (48/82) coincided with the engine’s first preference

• 92.7% of Kramnik’s moves (38/41) coincided with the engine preferences 1, 2 and 3

• 95.1% of Anand’s moves (39/41) coincided engine preferences 1, 2 and 3

• 80.5% of Kramnik’s moves (33/41) coincided with engine preferences 1 and 2

• 92.7% of Anand’s moves (38/41) coincided with engine preferences 1 and 2

• 53.7% of Kramnik’s moves (22/41) coincided with the engine’s first preference

• 63.4% (26/41) of Anand’s moves coincided with the engines first preference.

<The engine evaluation of the final position>:

1 <blunder> by Anand:

<31…Bg4> ( 0.95)

2 <blunders> by Kramnik>:

First Kramnik blunder:

<29. Ra3> ( 1.79)

Second Kramnik blunder:

<32. f3> ( 1.90)

1 Kramnik dubious move:

<19. Nxd4> (=-0.02)

RESULTS:

<Using method A, the game is weighted at <<6>>, representing <<1 blunder>> by Anand and <<2 blunders>> by Kramnik.>

<Under weighting method B, weighting is <<6.5>>, representing <<1 blunder>> by Anand and <<2 blunders>> and <<1 dubious move>> by Kramnik.>

click for larger view

Analysis by Komodo32 3 32bit (depth=26):

1. ± (0.71): 18.b4 Ke7 19.a4 Re4 20.Qd3 Ne5 21.Nxe5 Rxe5 22.Bb2 Rxb5 23.axb5 e5 24.Rxa8 Bxa8 25.f3 Bd5 26.Ra1 Qb7 27.Kf2 h5 28.Kg2 Qc7 29.Rc1 Qb7 30.Ba3 Kf8 31.Rc5 Qa8 32.Bc1 Bxc5 33.bxc5 Qa1

2. = (0.25): 18.Nd2 Ke7 19.Bxd7 Rag8 20.Nc4 Qa6 21.Ba4 Qxa4 22.Nxd6 Kxd6 23.f3 Rh4 24.b3 Qc6 25.Ba3+ Kd7 26.Rdc1 Qxf3 27.Qxf3 Bxf3 28.Kf2 Rxh2+ 29.Kxf3 Rh3 30.Rg1 e5 31.Bb4 f5 32.Rac1 Rg4 33.Bc5 e4+

3. = (0.24): 18.a4 Qc5 19.Ne1 Re4 20.Qd3 Qh5 21.f3 Re5 22.Bd2 Rb8 23.b4 Rxb5 24.Qxb5 Qxb5 25.axb5 Bd5 26.Kg2 Ne5 27.Ra6 Nc4 28.Bh6 e5 29.Nd3 Rxb5 30.Re1 Bb7 31.Raa1 Bd5 32.Ra4 Bb8 33.Rea1 e4

4. = (0.17): 18.Bf4 Bxf4 19.Rxd4 Kf8 20.Bxd7 Rd8 21.Rad1 Rxd7 22.Rxd7 Bxg3 23.hxg3 Rxg3+ 24.Kh2 Bxf3 25.Qe3 Rg2+ 26.Kh3 Qxe3 27.fxe3 Rxb2 28.R1d2 Rb4 29.Kg3 Bc6 30.R7d4 Rb1 31.e4 Rg1+ 32.Kf2 Rg4 33.Ke3 f5

5. = (-0.05): 18.Qd3 0-0-0 19.Qe2 Kb8 20.a4 Qc5 21.Ba6 d3 22.Qxd3 Bxg3 23.Bxb7 Qxf2+ 24.Kh1 Bc7 25.Rd2 Qb6 26.h3 Rg3 27.Be4 f5 28.a5 Rxh3+ 29.Kg2 Rg8+ 30.Kxh3 Rg3+ 31.Kh4 Rg4+ 32.Kh3 Rg3+

At lower depth (20 or so plys), three options: Bf4, b4 and a4 appear to be approximately equal. <visyanbraindoctor> gives these three options evaluations of .69, .76 and .70. Going deeper (to a depth of 26), Komodo gives these three options evaluations of .17, .71 and .24. Kramnik's actual game move of 18 Bf4 is at least a "questionable move" if one analyses it in any depth.

However Black's bishop is placed powerfully on the long diagonal h1-a8 whereas White's bishop is doing little.

Black is therefore playing with an extra bishop.