Castling long

[Lyudmil Tsvetkov]

Thank you, Lyudmil.
I also felt surprised me.

Remember if clikea to points above the " A8 black rook" has more games other engines.
Total 5 games whith "O.O.O", on the same page.

Thank you, Jose.
5 games in all, many Sicilians, and among those 5 games only one engine managed to win with long castling - Bouquet. Overall, not very much an Alekhine-style showing, but still...

[Lyudmil Tsvetkov]

Just to add a further related comment.

I think that in order to be able to play successfully with long castling, an engine should have a good grasp of piece control of center. If a pawn on one of the center-most 4 squares is due some 40 centipawns, a pawn in the extended center some 20 cps, then a minor piece controlling some of the 4 center-most squares might be due some 10cps (5 cps for a bishop controlling on an x-ray, still important), 7cps for rook control of those squares (and within that would be counted the square immediately in front of a pawn with a rook behind it, if central), etc. I think this is very important, and I can not imagine how an engine can play a sophisticated game with a long castling without such a knowledge.

The Sicilian would be an example of the importance of this. Besides, for the Sicilian, other considerations might be of utmost importance, like adequate minor piece control in front of backward pawns. A minor controlling the square in front of a backward pawn could be worth some 15 cps (not at all exagerrated), a rook control might be worth some 10cps, etc. Also x-rays might be important.

For semibackward pawns, where backwardness would depend by definition on piece control in front of such squares, all pieces could get somewhat lower values. I simply can not imagine how an engine could play a good Sicilian game without adequate knowledge of this. And for me, this is one of the reasons why most modern engines still have difficulties with the Sicilian (at least some lines).

Just my few cents.

[Don]

I think you are approaching this all wrong. You are starting with an assumption (that chess engines are badly broken) and then trying to "bend" the program to your will, to play the way you think they should.

So you have all this "stuff" below to "fix" the programs to do this "right" and you have not considered the possibility that you have it wrong.

You should begin with an attempt to determine if your assertion is correct. You have humans castling long about 11% of time and computers, who are several hundred ELO stronger doing it about 3 or 4% of the time. So why are you so sure the computers are the ones that are wrong? Maybe they are, but you have not mentioned this possibility.

If Larry and I wanted to check this out we would probably start by trying to increase the long side castling frequency by changing the weights. We could easily make it castling on the long side more often with a few minor adjustments and then we would see if we can make that happen without making the program weaker.

Do you have several examples of the computer castling short when it is clear that it should castle long? We could trying running those positions at great depths to see if the program "wakes up."

I have had numerous people show me positions where Komodo did something wrong, but I don't remember a single case where it was reported to me that Komodo should have castled long. It's pretty rare that Komodo errors with respect to castling as far as I know.

It would be great to know if Komodo's statistics are the same as other strong programs, so maybe I will run some games just for that.

Just to add a further related comment.

I think that in order to be able to play successfully with long castling, an engine should have a good grasp of piece control of center. If a pawn on one of the center-most 4 squares is due some 40 centipawns, a pawn in the extended center some 20 cps, then a minor piece controlling some of the 4 center-most squares might be due some 10cps (5 cps for a bishop controlling on an x-ray, still important), 7cps for rook control of those squares (and within that would be counted the square immediately in front of a pawn with a rook behind it, if central), etc. I think this is very important, and I can not imagine how an engine can play a sophisticated game with a long castling without such a knowledge.

The Sicilian would be an example of the importance of this. Besides, for the Sicilian, other considerations might be of utmost importance, like adequate minor piece control in front of backward pawns. A minor controlling the square in front of a backward pawn could be worth some 15 cps (not at all exagerrated), a rook control might be worth some 10cps, etc. Also x-rays might be important.

For semibackward pawns, where backwardness would depend by definition on piece control in front of such squares, all pieces could get somewhat lower values. I simply can not imagine how an engine could play a good Sicilian game without adequate knowledge of this. And for me, this is one of the reasons why most modern engines still have difficulties with the Sicilian (at least some lines).

Just my few cents.

[Dirt]

I'm not certain, but I would guess long castling is less drawish than short castling is. It probably also depresses the players result, at least slightly. If you're the best player in the world that might work out for you, but mostly it won't.

[Don]

I am running the stats on Houdini 3 and Houdini 3 castles long 2.85 percent of the time in the games I played.

For reference Komodo castles 3.6 percent of the time. For Komodo I used a sample of 19,817 games and for Houdini I have 7,170 games.

I seriously doubt this has anything to do with anything other that style. I'll try stockfish 3 now.

Don

[Sven]

Castling statistics will have a lot to do with the book.

Komodo has a wide but very shallow book of exactly 5 moves - with about 9% of the openings having castle short in them. Only 2 position have long castling out of over 35,000 openings.

I think this first part is easy to explain: you need exactly one move of the d-pawn and three moves of pieces (QBN) to enable queenside castling within five moves from the initial position, while you have one extra "arbitrary" move to enable kingside castling, and you also have two choices for the necessary pawn move (e- or g-pawn). That means that it is highly unlikely for a 10 ply move sequence to contain a long castling, simply based on the rules. It would already look quite different for a 6, 7, 8 moves book.

Sven

[Don]

Castling statistics will have a lot to do with the book.

Komodo has a wide but very shallow book of exactly 5 moves - with about 9% of the openings having castle short in them. Only 2 position have long castling out of over 35,000 openings.

I think this first part is easy to explain: you need exactly one move of the d-pawn and three moves of pieces (QBN) to enable queenside castling within five moves from the initial position, while you have one extra "arbitrary" move to enable kingside castling, and you also have two choices for the necessary pawn move (e- or g-pawn). That means that it is highly unlikely for a 10 ply move sequence to contain a long castling, simply based on the rules. It would already look quite different for a 6, 7, 8 moves book.

Sven

It may be that a much deeper search will play castle long more often. Even though current chess programs can easily do 20 or 25 ply searches my test games are more like 12 or 13 ply - and even 20 ply or more will not necessarily see castling long if there is a lot going on tactically that require the programs attention. I think current programs with aggressive LMR and pruning tend to be a little greedy with respect to counting beans, preferring to have their reward now instead of later.

[Lyudmil Tsvetkov]

I think you are approaching this all wrong. You are starting with an assumption (that chess engines are badly broken) and then trying to "bend" the program to your will, to play the way you think they should.

So you have all this "stuff" below to "fix" the programs to do this "right" and you have not considered the possibility that you have it wrong.

You should begin with an attempt to determine if your assertion is correct. You have humans castling long about 11% of time and computers, who are several hundred ELO stronger doing it about 3 or 4% of the time. So why are you so sure the computers are the ones that are wrong? Maybe they are, but you have not mentioned this possibility.

If Larry and I wanted to check this out we would probably start by trying to increase the long side castling frequency by changing the weights. We could easily make it castling on the long side more often with a few minor adjustments and then we would see if we can make that happen without making the program weaker.

Do you have several examples of the computer castling short when it is clear that it should castle long? We could trying running those positions at great depths to see if the program "wakes up."

I have had numerous people show me positions where Komodo did something wrong, but I don't remember a single case where it was reported to me that Komodo should have castled long. It's pretty rare that Komodo errors with respect to castling as far as I know.

It would be great to know if Komodo's statistics are the same as other strong programs, so maybe I will run some games just for that.

Hi Don.

It would be difficult to bend an engine to one's will, even if one would like to.

I do not hate engines, I sympathize with engines, those are just some suggestions, many of which might be counter-productive in a specific environment.

Unfortunately, I do not own Komodo, so I can not be of any help concerning its play.

I did not do any extensive tests, I myself am not convinced that a perfect play would suppose, say, a 7% long castling rate (instead of the 5% Komodo displays, 3% Houdini, etc.). You should blame it all on Alekhine and Kasparov, especially Alekhine, that castled long so many times to convincing victories achieved with ease, in every possible opening.

I did not investigate specific examples of where an engine could have failed because of avoiding to castle long, but I am sure there are numerous. It just does not make sense not to be able to castle long more frequently, when you have to make just a limited amount of 'spare moves' to do so.

You might try Komodo on the position in the initial post to see if Komodo castles short or long.

Please, do not take my messages as a crusade on engines. They are quite the opposite, sympathetic postings with humble suggestions, based on the assumption that some of them might work. No one makes you run all those tests, but if you are curious, it is not my fault

[Lyudmil Tsvetkov]

Hi Don again.
Here just a small further message just to keep alive the discussion (this is a discussion board, is not it?). Please, do not take it as any kind of infringement upon your status, or other programmers' status, war on engines, etc. It is just to keep the discussion going (btw., thank you very much for your statistics, I appreciated them enormously, and they are quite an useful amount of information).

Below are 3 games of Alekhine from his peak period.

[pgn][MLNrOfMoves "32"]
[MLFlags "000000"]
[Site "Santiago"]
[Date "1927.??.??"]
[White "Alekhine A"]
[Black "Rodrigo Flores"]
[Result "1-0"]
[ECO "D35"]

1. d4 Nf6 2. c4 e6 3. Nc3 d5 4. Bg5 Nbd7 5. e3 c6 6. cxd5 exd5 7. Bd3 Be7 8.
Nge2 O-O 9. Ng3 Re8 10. Qc2 Nf8 11. h4 b6 12. Nf5 Bxf5 13. Bxf5 c5 14. O-O-O c4
15. Kb1 b5 16. e4 g6 17. e5 Nh5 18. Bg4 Bxg5 19. hxg5 Nf4 20. g3 Nd3 21. Bf3 Ne6
22. Nxd5 Nxd4 23. Nf6+ Qxf6 24. gxf6 Nxc2 25. Bxa8 Rxa8 26. Kxc2 Nxf2 27. Rhf1
Nxd1 28. Rxd1 h5 29. Rd7 Re8 30. Re7 Rd8 31. Rxa7 Rd5 32. Ra8+ Kh7 33. e6 1-0

[Site "Barcelona"]
[Date "1928.??.??"]
[White "Alekhine A"]
[Black "Domenech R"]
[Result "1-0"]
[ECO "D00"]
[MLNrOfMoves "34"]
[MLFlags "000000"]

1. d4 Nf6 2. Nf3 g6 3. Nc3 d5 4. Bf4 Bg7 5. e3 c6 6. Bd3 Nbd7 7. e4 dxe4 8. Nxe4
Nxe4 9. Bxe4 Nf6 10. Bd3 Bg4 11. c3 Nh5 12. Be3 e5 13. Be2 exd4 14. Nxd4 Bxe2
15. Qxe2 O-O 16. O-O-O Qa5 17. Kb1 Nf6 18. h4 Rfe8 19. Qf3 Nd5 20. Bc1 Qc7 21.
h5 a5 22. Qh3 Bxd4 23. Rxd4 b5 24. Rg4 Re6 25. Rxg6+ Rxg6 26. hxg6 fxg6 27. Qe6+
Kg7 28. Bh6+ Kh8 29. Re1 Rg8 30. Qe4 Qf7 31. Qd4+ Qf6 32. Re5 Nc7 33. Bg5 Qg7
34. Qd6 Rf8 35. Re7 1-0

[Site "Europe"]
[Date "1930.??.??"]
[White "Eljaschev"]
[Black "Alekhine Alexander"]
[Result "0-1"]
[ECO "C01"]
[MLNrOfMoves "20"]
[MLFlags "000000"]

1. e4 e6 2. d4 d5 3. exd5 exd5 4. Bd3 Bd6 5. Nf3 Nc6 6. O-O Bg4 7. c3 Nge7 8.
Re1 Qd7 9. Nbd2 O-O-O 10. Qa4 Rde8 11. b4 Kb8 12. b5 Nd8 13. Ne5 Bxe5 14. Rxe5
Ng6 15. Rxe8 Rxe8 16. Nf1 Re1 17. Bb2 Qe8 18. f3 Qe3+ 19. Kh1 Rxa1 20. Bxa1 Qxd3
0-1
[/pgn]

And some diagrams. One could check what her engine would do: castle long or short. (btw., the development of some games is quite interesting)

[d]r2qrnk1/p3bppp/1p3n2/2pp1BB1/3P3P/2N1P3/PPQ2PP1/R3K2R w KQ - 0 14

Alekhine here castles long with white. I am sure many engines would find this 'scary'.

[d]r2q1rk1/pp3pbp/2p3p1/7n/3N4/2P1B3/PP2QPPP/R3K2R w KQ - 0 16

Alekhine castles long with white. Looks a bit scary, too.

[d]r3k2r/pppqnppp/2nb4/3p4/3P2b1/2PB1N2/PP1N1PPP/R1BQR1K1 b kq - 0 9

Alekhine castles long with black. I do not know what most engines would prefer.

For anyone interested more in long castling, please see Alekhine games (there are an awful lot, and much better examples, but I do not have the opportunity to investigate deeper).

Also some notes on Alekhine style from Wikipedia:

Statistical ranking systems differ sharply in their views of Alekhine. "Warriors of the Mind" rates him only the 18th strongest player of all time and comments that victories over players such as Bogoljubov and Euwe are not a strong basis for an "all time" ranking.[81] But the website "Chessmetrics" ranks him between the fourth and eighth best of all time, depending on the lengths of the peak periods being compared, and concludes that at his absolute peak he was a little stronger than Emanuel Lasker and Capablanca, although a little weaker than Botvinnik.[82] Jeff Sonas, the author of the website "Chessmetrics", rates Alekhine as the sixth highest peak strength, relative to other players of the same era, of all-time on the basis of comparable ratings.[83] He also assesses Alekhine's victory at the tournament of San Remo in 1930 as the sixth best performance ever in tournaments.[83] In his 1978 book The Rating of Chessplayers, Past and Present, Arpad Elo gave retrospective ratings to players based on their performance over the best five-year span of their career. He concluded that Alekhine was the joint fifth strongest player of those surveyed (tied with Paul Morphy and Vasily Smyslov), behind Capablanca, Botvinnik, Emanuel Lasker and Mikhail Tal.[84]

Alekhine's peak period was in the early 1930s, when he won almost every tournament he played, sometimes by huge margins. Afterward, his play declined, and he never won a top-class tournament after 1934. After Alekhine regained his world title in 1937, there were several new contenders, all of whom would have been serious challengers.[10]

Alekhine was one of the greatest attacking players and could apparently produce combinations at will. What set him apart from most other attacking players was his ability to see the potential for an attack and prepare for it in positions where others saw nothing. Rudolf Spielmann, a master tactician who produced many brilliancies, said, "I can see the combinations as well as Alekhine, but I cannot get to the same positions."[10] Dr. Max Euwe said, "Alekhine is a poet who creates a work of art out of something that would hardly inspire another man to send home a picture post-card."[85] An explanation offered by Réti was, "he beats his opponents by analysing simple and apparently harmless sequences of moves in order to see whether at some time or another at the end of it an original possibility, and therefore one difficult to see, might be hidden."[86] John Nunn commented that "Alekhine had a special ability to provoke complications without taking excessive risks",[87] and Edward Winter called him "the supreme genius of the complicated position."[88] Some of Alekhine's combinations are so complex that even modern champions and contenders disagree in their analyses of them.[89]

Nevertheless, Garry Kasparov said that Alekhine's attacking play was based on solid positional foundations,[89] and Harry Golombek went further, saying that "Alekhine was the most versatile of all chess geniuses, being equally at home in every style of play and in all phases of the game."[90] Reuben Fine, a serious contender for the world championship in the late 1930s, wrote in the 1950s that Alekhine's collection of best games was one of the three most beautiful that he knew,[10] and Golombek was equally impressed.[90]

Alekhine's games have a higher percentage of wins than those of any other World Champion, and his drawn games are on average among the longest of all champions'.[91] His desire to win extended beyond formal chess competition. When Fine beat him in some casual games in 1933, Alekhine demanded a match for a small stake. And in table tennis, which Alekhine played enthusiastically but badly, he would often crush the ball when he lost.[10]

Bobby Fischer, in a 1964 article, ranked Alekhine as one of the ten greatest players in history.[92] Fischer, who was famous for the clarity of his play, wrote of Alekhine:


Alekhine has never been a hero of mine, and I've never cared for his style of play. There's nothing light or breezy about it; it worked for him, but it could scarcely work for anyone else. He played gigantic conceptions, full of outrageous and unprecedented ideas. ... [H]e had great imagination; he could see more deeply into a situation than any other player in chess history. ... It was in the most complicated positions that Alekhine found his grandest concepts.[92]

Alekhine's style had a profound influence on Kasparov, who said: "Alexander Alekhine is the first luminary among the others who are still having the greatest influence on me. I like his universality, his approach to the game, his chess ideas. I am sure that the future belongs to Alekhine chess."[93] In 2012, Levon Aronian said that he considers Alekhine the greatest chess player of all time.[94

[lucasart]

To post just another small message to cool my necessities in the heat.

Two of the best chess players in history, Kasparov and Alekhine, used to castle long much more frequently than other players: by Kasparov the percentage of games with long castling might be somewhere 15-20, while I suppose (without strict measurements done) that Alekhine would even reach some 30%, an unprecedented score. Other prominent and less prominent chess players would castle long much less frequently, unless, of course, they are inveterate white Sicilian players. I would also notice that, in general, engines would prefer castling short in the great variety of cases, much unlike Kasparov and Alekhine.

I would ask myself what could the reason for such a behaviour be? Were Kasparov and Alekhine weak players, or maybe they had bigger imagination than others? Are engines always correct when they choose to castle short instead of allowing themselves an option of long castling, or maybe they are missing some valuable unexplored alternatives? It is obvious that castling long would immediately expose the king to bigger danger, but maybe there might be some trade-offs for that in the sense of good activity in the center and good attacking chances on the king side.

For the time being I am very much undecided on this issue. I could suppose, though, that most engines, just like most humans, do not make the most of castling long. Maybe they have some larger than necessary penalties for long castling, or maybe lacking a bit in imagination Or maybe the engine side is right after all, with deeper search... Difficult to say, I would intuitively go for the first option. Interesting to know your opinion on the matter, does anyone by chance have some data (when this is possible to collect, for example, with very short books and castling still available) on which engines would exhibit a higher castling long rate than the average, and if those engines would be among the leading pack?

Below is a game of Alekhine on the theme.

[pgn][Site "Mexico City (08)"]
[Date "1932.??.??"]
[White "Alekhine Alexander"]
[Black "Brunner J"]
[Result "1-0"]
[ECO "D36"]
[MLNrOfMoves "25"]
[MLFlags "100000"]

1. d4 Nf6 2. c4 e6 3. Nc3 d5 4. Bg5 Nbd7 5. cxd5 exd5 6. e3 c6 7. Bd3 Be7 8. Qc2
O-O 9. Nge2 h6 10. h4 Re8 11. Bf4 a6 12. O-O-O c5 13. g4 cxd4 14. Nxd4 Nc5 15.
g5 Nxd3+ 16. Rxd3 Nh5 17. gxh6 Nxf4 18. exf4 g6 19. h5 Kh7 20. hxg6+ fxg6 21.
Rg3 Rg8 22. Rhg1 Qb6 23. Nxd5 Qxd4 24. Rxg6 Qc5 25. Rg7+ Kxh6 26. Rh7# 1-0
[/pgn]

[d]r1bq1rk1/pp1nbpp1/2p2n1p/3p2B1/3P4/2NBP3/PPQ1NPPP/R3K2R w KQ - 0 10

It is interesting to know how many engines would play now 10.h4, like Alekhine, deciding to castle long in the aftermath.

I very much hope you are not bored by this, I think it has some relevance.

Best, Lyudmil

Castle long or not: that's a tricky question that I often ask myself when playing over the board.

As far as DiscoCheck is concerned, what I can tell you is that:

* it does not distinguish in the evaluation if it is long castled or short castled. in fact it does not even know that. instead it looks at the caracteristics of the position (castling is only one way to reach a position that you could otherwise reach with artificial castling in many moves, but the eval would be the same).

* the eval is neutral with respect to side of castling:
- B1 and G1 are both optimal squares for the King in the opening, as far as the piece on square table (PST) is concerned.
- more generally PST for King is filewise symmetric (both opening and endgame).
- king safety is square agnostic, as far as piece attack goes, and adds more weight to the total attacks on the king based on min(distance to B1, distance to G1). again OO and OOO are equal here.

Yet DiscoCheck plays OOO somewhat less often than OO too. One reason is the loss of tempo Kb1, and another reason is the vulnerability of the King on the queen side.

In fact, I think that DiscoCheck chooses OOO too easily, and it often backfires. Just after reading your post, I played only one game, and that pattern happenned: DiscoCheck plays OOO and falls prey to a king attack:
[pgn]
[White "Discocheck"]
[Black "Fruit-051103"]
[ECO "B82"]
[Opening "Sicilian"]
[Variation "Scheveningen, Tal Variation, 8...e5"]
[TimeControl "300+3"]
[Result "0-1"]
[PlyCount "89"]

1. Nc3 c5 2. e4 d6 3. Nf3 Nf6 4. d4 cxd4 5. Nxd4 a6 6. Be3 e6 7. Qf3
{(Qd1-f3 Bf8-e7 O-O-O Qd8-c7 Kc1-b1 O-O g2-g4 Nb8-c6 Nd4xc6 Qc7xc6 g4-g5
Nf6-d7 h2-h4 Nd7-e5 Qf3-h5 Bc8-d7 f2-f4 g7-g6) +0.44/18 13} e5 {(e6-e5
Nd4-b3 Bf8-e7 h2-h3 O-O Bf1-e2 Nb8-c6 Qf3-g3 Kg8-h8 O-O-O Qd8-c7 Kc1-b1
Bc8-e6 h3-h4) -0.39/14 8} 8. Nf5 {(Nd4-f5 Bc8xf5 Qf3xf5 Nb8-c6 O-O-O Bf8-e7
Bf1-e2 O-O Kc1-b1 Qd8-c8 Qf5xc8 Ra8xc8 Rh1-f1 b7-b5 f2-f4 Rf8-d8 a2-a3
Rd8-d7 h2-h3) +0.68/19 13} g6 {(g7-g6 Nf5-h6 Bc8-e6 Be3-g5 Bf8-g7 Bf1-d3
Nb8-c6 O-O Ra8-c8 Qf3-g3 Nc6-b4 Ra1-d1 Rc8-c5 Rf1-e1 Nb4xd3 Rd1xd3)
-0.24/15 13} 9. Nh6 {(Nf5-h6 Bc8-e6 Be3-g5 Bf8-e7 O-O-O Nb8-c6 Kc1-b1
Nc6-d4 Qf3-d3 Ke8-f8 f2-f4 Nf6-g8 Nc3-e2 Be7xg5 f4xg5 Nd4xe2 Bf1xe2 Qd8xg5
Qd3xd6+ Kf8-g7) +0.48/19 12} Be6 {(Bc8-e6 Be3-g5 Bf8-g7 Bf1-d3 Nb8-c6 O-O
Ra8-c8 Rf1-e1 Nc6-d4 Qf3-g3 Rc8-c6 Qg3-h4 Rc6-c5 Nc3-d5 Be6xd5 e4xd5
Rc5xd5) -0.25/15 14} 10. Bg5 {(Be3-g5 Bf8-e7 Bf1-d3 Nf6-h5 Bg5-e3 Nh5-f4
g2-g3 Nf4xd3+ c2xd3 Be7-f8 O-O Bf8xh6 Be3xh6 Nb8-c6 Ra1-d1 Nc6-d4 Qf3-e3
Be6-h3 Bh6-g5) +0.43/18 12} Be7 {(Bf8-e7 Bf1-d3 Nf6-h5 Bg5-e3 Nh5-f4 O-O-O
Be7-g5 g2-g3 Bg5xh6 g3xf4 Nb8-c6 Kc1-b1 O-O Rh1-g1) -0.34/15 14} 11. O-O-O
{(O-O-O Nb8-c6 Bg5-e3 Be7-f8 Qf3-e2 Nc6-d4 Qe2-e1 b7-b5 Bf1-d3 b5-b4 Nc3-e2
Qd8-a5 f2-f3 Ra8-c8 Qe1-h4 Bf8xh6 Be3xh6) +0.41/17 12} Nc6 {(Nb8-c6 Nc3-e2
Nf6-d7 Bg5-e3 Nd7-f6 Be3-g5) 0.00/14 11} 12. Kb1 {(Kc1-b1 Nc6-d4 Qf3-d3
Ke8-f8 h2-h4 Kf8-g7 Nc3-e2 Qd8-a5 a2-a3 Nd4-b5 g2-g3 Qa5-a4 f2-f3 Qa4-c4
Ne2-c1 Qc4xd3 Nc1xd3) +0.37/17 11} Nd4 {(Nc6-d4 Qf3-d3 Ra8-c8 Bg5-e3 Be7-f8
Qd3-d2 b7-b5 Be3xd4 e5xd4 Nc3-d5 g6-g5 Qd2xg5 Nf6xe4 Qg5xd8+ Ke8xd8 f2-f3
Be6xd5 f3xe4) +0.29/14 11} 13. Qd3 {(Qf3-d3 Ke8-f8 Nc3-e2 Nd4-c6 Qd3-e3
Kf8-g7 h2-h4 Rh8-e8 Ne2-c3 Qd8-a5 Bf1-d3 Nc6-d4 f2-f3 Kg7-h8 g2-g4 b7-b5
h4-h5 Ra8-c8) +0.26/18 11} Rc8 {(Ra8-c8 Bg5-e3 Be7-f8 Qd3-d2 b7-b5 Be3xd4
e5xd4 Nc3-d5 Rc8-c5 Nd5xf6+ Qd8xf6 f2-f3 Bf8-g7 Nh6-g4 Qf6-h4 Bf1-d3)
+0.12/13 11} 14. h4 {(h2-h4 Nf6-g4 Bg5xe7 Qd8xe7 Nh6xg4 Be6xg4 f2-f3 Bg4-e6
Qd3-d2 O-O h4-h5 g6-g5 Bf1-d3 f7-f5 e4xf5 Be6xf5) +0.16/16 11} Qb6 {(Qd8-b6
Bg5-d2 Ke8-f8 Bd2-e3 Kf8-g7 Qd3-d2 Rh8-e8 f2-f3 Qb6-a5 a2-a3 Qa5-b6 Be3xd4
e5xd4) +0.30/12 9} 15. f4 {(f2-f4 Qb6-a5 a2-a3 Nf6-g4 Nh6xg4 Be6xg4 Rd1-d2
f7-f6 f4xe5 d6xe5 Bg5-e3 O-O Be3xd4 e5xd4 Qd3xd4) +0.37/15 11} Rc6 {(Rc8-c6
f4xe5 d6xe5 Bg5-d2 Nf6-g4 Qd3-g3 Ng4xh6 Bd2xh6 f7-f6 a2-a3 Nd4xc2 Kb1xc2
Be6-b3+ Kc2-b1 Rc6xc3 Qg3xc3 Bb3xd1 Qc3-c8+ Be7-d8) +0.34/12 14} 16. g3
{(g2-g3 Ke8-f8 Bf1-h3 Be6-c4 Qd3-d2 Kf8-g7 f4-f5 Nd4-f3 Qd2-c1 Rh8-e8 f5xg6
h7xg6 Bg5-e3 Nf3-d4) +0.61/14 10} Bc4 {(Be6-c4 Qd3-e3 Nd4-b5 Qe3xb6 Nb5xc3+
b2xc3 Rc6xb6+ Kb1-a1 Bc4xf1 Rh1xf1 Nf6xe4 f4xe5 Ne4xg5 h4xg5 Rh8-f8 Rf1-e1
Ke8-d7 e5xd6 Rb6xd6) +0.72/13 10} 17. Qd2 {(Qd3-d2 Bc4xa2+ Kb1xa2 Nf6xe4
Nc3xe4 Rc6xc2 Qd2xc2 Nd4xc2 Bf1-c4 Qb6-c6 Bc4xf7+ Ke8-f8 Bg5xe7+ Kf8xe7
Ne4-c3 Nc2-b4+ Ka2-b3 Qc6-b6 Nc3-a4 Qb6-b5 Na4-c3) +0.36/18 11} Nxe4
{(Nf6xe4 Nc3xe4 Bc4xa2+ Kb1xa2 Rc6xc2 Qd2xc2 Nd4xc2 Ne4-f6+ Be7xf6 Bg5xf6
Nc2-b4+ Ka2-a3 Rh8-f8 f4xe5 Qb6-a5+ Ka3-b3 Nb4-d5 Bf1-h3 Qa5-b4+ Kb3-a2
Qb4-a4+ Ka2-b1 d6xe5 Rh1-e1) +1.61/12 12} 18. Nxe4 {(Nc3xe4 Bc4xa2+ Kb1xa2
Rc6xc2 Qd2xc2 Nd4xc2 Bf1-c4 Qb6-c6 Bc4xf7+ Ke8-d7 Ne4-c3 Nc2-e3 Rd1-e1
Be7xg5 h4xg5 Rh8-f8 f4xe5 d6xe5) +0.08/17 5} Bxa2+ {(Bc4xa2+ Kb1xa2 Rc6xc2
Qd2xc2 Nd4xc2 Ne4-f6+ Be7xf6 Bg5xf6 Nc2-b4+ Ka2-b1 Qb6-c6 Rh1-h2 Qc6-a4
Rd1xd6 Qa4-a2+ Kb1-c1 Qa2-a1+ Kc1-d2 Qa1xb2+ Kd2-e3 Nb4-d5+ Rd6xd5 e5xf4+
g3xf4 Qb2xf6 Rd5-e5+ Ke8-d7 Rh2-d2+ Kd7-c7) +1.66/13 11} 19. Kxa2 {(Kb1xa2
Rc6xc2 Qd2xc2 Nd4xc2 Bf1-c4 d6-d5 Bg5xe7 Ke8xe7 Bc4xd5 Rh8-c8 Ka2-b1
Nc2-a3+ Kb1-a1 Na3-c2+ Ka1-b1) -0.25/17 10} Rxc2 {(Rc6xc2 Qd2xc2 Nd4xc2
Ne4-f6+ Be7xf6 Bg5xf6 Nc2-b4+ Ka2-b1 Qb6-c6 Rh1-h2 Qc6-a4 Rd1xd6 Qa4-a2+
Kb1-c1 Qa2-a1+ Kc1-d2 Qa1xb2+ Kd2-d1 Qb2-b1+ Kd1-e2 Qb1-c2+ Ke2-f3 e5-e4+
Kf3-e3 Qc2-b3+ Ke3xe4 Qb3xg3) +1.54/14 11} 20. Qxc2 {(Qd2xc2 Nd4xc2 Bf1-c4
d6-d5 Bg5xe7 Ke8xe7 Bc4xd5 Rh8-c8 Ka2-b1 Nc2-a3+ Kb1-a1 Na3-c2+ Ka1-b1)
-0.25/16 9} Nxc2 {(Nd4xc2 Rd1-c1 Nc2-b4+ Ka2-b1 Qb6-a5 Rc1-c8+ Ke8-d7
Rc8xh8 Qa5-a2+ Kb1-c1 Qa2-a1+ Kc1-d2 Qa1xb2+ Kd2-e1 Qb2-b1+ Ke1-f2 Qb1xe4
Bf1-h3+ f7-f5 Rh8xh7 Qe4xh1 Rh7xe7+ Kd7-c6) +1.51/10 2} 21. Bc4 {(Bf1-c4
d6-d5 Bg5xe7 Ke8xe7 Bc4xd5 Rh8-c8 Ka2-b1 Nc2-a3+ Kb1-a1 Na3-c2+ Ka1-b1)
-0.25/15 9} d5 {(d6-d5 Bc4xd5 f7-f6 Bg5xf6 Be7xf6 Nh6-g4 Bf6-e7 f4xe5
Nc2-d4 Ng4-f6+ Be7xf6 Ne4xf6+ Ke8-d8 Rh1-f1 Kd8-e7) +1.58/12 9} 22. Bxe7
{(Bg5xe7 Ke8xe7 Bc4xd5 e5xf4 Bd5xf7 f4xg3 Ne4xg3 Rh8-c8 Ka2-b1 Nc2-a3+
Kb1-a1 Na3-c2+ Ka1-b1) -0.25/15 9} Kxe7 {(Ke8xe7 Rd1xd5 Nc2-b4+ Ka2-b1
Nb4xd5 Bc4xd5 f7-f5 Nh6-f7 Rh8-f8 Ne4-g5 Qb6-d4 Rh1-e1 Qd4-d3+ Kb1-a1
e5-e4) +3.08/11 6} 23. Rxd5 {(Rd1xd5 Nc2-b4+ Ka2-b1 Nb4xd5 Bc4xd5 Rh8-d8
Bd5xf7 e5xf4 g3xf4 Qb6-c6 Ne4-f2 Qc6-g2 Rh1-e1+ Ke7-f8 Nf2-g4 Qg2-g3
Re1-e2) -2.04/15 9} Nb4+ {(Nc2-b4+ Ka2-b1 Nb4xd5 Bc4xd5 f7-f5 Ne4-c3 Rh8-c8
Rh1-c1 e5-e4 g3-g4 Ke7-f8 g4xf5 Kf8-g7 f5xg6 Kg7xh6 g6xh7) +3.04/13 4} 24.
Kb1 {(Ka2-b1 Nb4xd5 Bc4xd5 f7-f5 Ne4-c3 Rh8-c8 Rh1-c1 e5-e4 g3-g4 Ke7-f8
g4xf5 Kf8-g7 Nh6-g8 Rc8xg8 Bd5xg8 Kg7xg8 Nc3xe4) -2.00/17 9} Nxd5 {(Nb4xd5
Bc4xd5 Rh8-d8 Ne4-c3 Qb6-f2 Kb1-a1 e5xf4 g3xf4 b7-b5 Nh6xf7 Rd8xd5 Nc3xd5+
Ke7xf7 h4-h5 g6xh5 Rh1xh5) +2.96/13 2} 25. Bxd5 {(Bc4xd5 Rh8-d8 Ne4-c3
Qb6-f2 Bd5xf7 Qf2xg3 f4xe5 Qg3-d3+ Kb1-a1 Qd3-d2 Nc3-d5+ Ke7-f8 Rh1-f1
Qd2xh6 Bf7xg6+ Kf8-g8 Nd5-f6+ Kg8-h8 Bg6-e4) -2.68/16 4} Rd8 {(Rh8-d8
Ne4-c3 Qb6-e3 Rh1-d1 Qe3xg3 f4xe5 Qg3xe5 Kb1-a2 Qe5-g7 Nh6xf7 Rd8xd5
Nc3xd5+ Ke7xf7 Rd1-e1 Qg7-d4 Re1-e7+ Kf7-f8) +3.22/14 7} 26. Bxf7 {(Bd5xf7
e5xf4 g3xf4 Qb6-c6 Rh1-e1 Rd8-d1+ Re1xd1 Qc6xe4+ Kb1-c1 Qe4xf4+ Kc1-c2
Qf4xh6 Bf7-d5 Qh6xh4 Bd5xb7 Qh4-c4+ Kc2-b1 h7-h5 Bb7-d5) -2.95/16 5} exf4
{(e5xf4 Rh1-e1 Ke7-f8 g3xf4 Kf8-g7 Nh6-g4 Qb6-a5 Re1-c1 Kg7xf7 Ng4-e5+
Kf7-e7 h4-h5 g6xh5) +4.69/12 5} 27. gxf4 {(g3xf4 Qb6-c6 Rh1-e1 Rd8-d1+
Re1xd1 Qc6xe4+ Kb1-c1 Qe4xf4+ Kc1-c2 Qf4xh6 Bf7-d5 Qh6xh4 Bd5xb7 Qh4-c4+
Kc2-b1 h7-h5 Bb7-d5 Qc4-e2) -4.03/16 13} Qb4 {(Qb6-b4 Ne4-c3 Ke7-f8 Rh1-c1
Kf8-g7 h4-h5 Kg7xh6 h5xg6 h7xg6 f4-f5 g6xf5 Bf7-e6 f5-f4 Rc1-h1+ Kh6-g5
Be6-d5 Rd8-e8 Rh1-g1+ Kg5-f5) +5.68/15 10} 28. Nc3 {(Ne4-c3 Ke7-f8 Rh1-c1
Kf8-g7 f4-f5 Kg7xh6 f5xg6 h7xg6 Rc1-c2 Qb4xh4 Rc2-e2 Kh6-g7 Bf7-e6 Qh4-f6
Kb1-a2 Rd8-e8 Be6-d5 Re8xe2 Nc3xe2) -5.08/19 8} Kf8 {(Ke7-f8 Rh1-c1 Kf8-g7
h4-h5 Kg7xh6 h5xg6 h7xg6 f4-f5 g6xf5 Bf7-e6 f5-f4 Rc1-h1+ Kh6-g5 Rh1-g1+
Kg5-h5 Be6-d5 Qb4-e7 Rg1-h1+ Kh5-g6 Rh1-g1+ Kg6-f5) +5.63/15 5} 29. Rc1
{(Rh1-c1 Kf8-g7 f4-f5 Kg7xh6 f5xg6 h7xg6 Rc1-c2 Qb4xh4 Kb1-a2 Qh4-f4 Bf7-d5
b7-b5 Bd5-b3 g6-g5 Nc3-d5 Qf4-e5 Nd5-b4 Qe5-d6 Nb4-c6) -5.23/19 8} Kg7
{(Kf8-g7 f4-f5 Kg7xh6 f5xg6 h7xg6 Bf7-e6 Qb4xh4 Be6-b3 Qh4-f2 Rc1-d1
Rd8xd1+ Nc3xd1 Qf2-f5+ Kb1-a2 g6-g5 Nd1-c3 g5-g4 Bb3-d5 b7-b6) +5.75/15 7}
30. f5 {(f4-f5 Kg7xh6 f5xg6 h7xg6 Rc1-c2 Qb4xh4 Kb1-a2 g6-g5 Rc2-g2 g5-g4
Bf7-d5 Qh4-h5 Bd5-e4 Rd8-d4 b2-b3 Qh5-h3 Rg2-c2 b7-b5 Ka2-b2) -5.38/19 8}
Kxh6 {(Kg7xh6 f5xg6 h7xg6 Bf7-e6 Qb4xh4 Be6-b3 g6-g5 Bb3-c2 Rd8-d4 Rc1-f1
Kh6-g7 Bc2-f5 g5-g4 Bf5-e6 b7-b6 Rf1-f7+ Kg7-g6) +5.78/15 6} 31. fxg6
{(f5xg6 h7xg6 Bf7-a2 Qb4xh4 Kb1-a1 Rd8-d2 Ba2-b1 Qh4-d4 Bb1-c2 Kh6-g7
Nc3-b1 Rd2-g2 Rc1-d1 Qd4-c4 Nb1-a3 Qc4-e6 Ka1-b1 Rg2-e2 Rd1-d8 Qe6-e7
Bc2-d1) -5.56/20 7} hxg6 {(h7xg6 Bf7-e6 Qb4xh4 Be6-b3 g6-g5 Kb1-a2 Qh4-b4
Rc1-h1+ Kh6-g7 Rh1-g1 g5-g4 Bb3-e6 Rd8-d4 Ka2-a1 a6-a5 Ka1-b1 a5-a4)
+6.10/15 8} 32. Ba2 {(Bf7-a2) -5.55/19 7} Qxh4 {(Qb4xh4 Ba2-b3 g6-g5 Bb3-c2
Rd8-d4 Rc1-g1 Kh6-g7 Kb1-a2 b7-b5 Rg1-d1 Rd4xd1 Nc3xd1 g5-g4 Nd1-e3 Qh4-f2
Ne3-f5+ Kg7-f6 Bc2-e4) +6.20/16 7} 33. Be6 {(Ba2-e6 Rd8-d2 Be6-b3 Qh4-h2
Bb3-c2 b7-b5 b2-b3 Qh2-e5 Nc3-a2 Qe5-d4 Rc1-h1+ Kh6-g5 Rh1-f1 a6-a5 Na2-c1
a5-a4 Rf1-f3 Rd2-g2 b3xa4) -5.86/18 7} g5 {(g6-g5 Be6-f5 Rd8-d4 Bf5-c2
g5-g4 Bc2-e4 Qh4-e7 Rc1-e1 g4-g3 Re1-h1+ Kh6-g5 Be4-f3 Qe7-f7 Nc3-e4+
Kg5-f4 Rh1-f1 Qf7-e6 Bf3-g2+ Kf4-g4) +6.44/15 6} 34. Rf1 {(Rc1-f1 Rd8-d2
Be6-b3 Qh4-h2 Kb1-a1 Qh2-g2 Rf1-f6+ Kh6-g7 Rf6-f7+ Kg7-g6 Bb3-a2 g5-g4
Rf7-f4 Kg6-h5 Nc3-e4 Rd2-d1+ Ba2-b1 Qg2-c2 Ne4-c3 Rd1-e1 Rf4-f7) -5.98/18
7} Rd2 {(Rd8-d2 Be6-b3 Qh4-h2 Kb1-a1 Qh2-g2 Rf1-f6+ Kh6-g7 Rf6-f5 Kg7-g6
Rf5-d5 Rd2xb2 Rd5-d6+ Kg6-f5 Bb3-d5 Qg2-h2 Bd5-e4+ Kf5-e5) +6.86/14 6} 35.
Bb3 {(Be6-b3 g5-g4 Bb3-d5 g4-g3 Rf1-h1 Rd2-h2 Rh1xh2 Qh4xh2 Bd5xb7 g3-g2
Bb7xg2 Qh2xg2 Kb1-a2 Kh6-g5 Ka2-b3 Qg2-b7+ Kb3-c2 Qb7-d7) -6.15/18 10} Qd4
{(Qh4-d4 Rf1-h1+ Kh6-g7 Rh1-e1 g5-g4 Re1-e7+ Kg7-h6 Kb1-c1 g4-g3 Re7-e6+
Kh6-g7 Re6-e4 Qd4-d7 Re4-e3 g3-g2 Re3-g3+ Kg7-h6 Nc3-e4 Rd2-e2) +7.15/14
11} 36. Rh1+ {(Rf1-h1+ Kh6-g7 Rh1-h5 Qd4-g1+ Kb1-a2 Kg7-g6 Rh5-h8 Qg1-c1
Nc3-d1 Rd2xd1 Bb3xd1 Qc1xd1 Rh8-g8+ Kg6-f6 b2-b3 Qd1-d2+ Ka2-a3 Qd2-a5+
Ka3-b2 Qa5-e5+ Kb2-c2 Kf6-f7 Rg8-a8 g5-g4) -7.14/17 8} Kg7 {(Kh6-g7 Rh1-e1
g5-g4 Re1-e7+ Kg7-h6 Re7-e6+ Kh6-h5 Kb1-c1 Rd2-h2 Kc1-b1 Qd4-d3+ Kb1-a1
Qd3-d2 Nc3-d1 Qd2-a5+ Bb3-a2 g4-g3 Re6-e7) +7.40/15 8} 37. Rh5 {(Rh1-h5
Qd4-g1+ Kb1-a2 Kg7-g6 Rh5-h8 Qg1-c1 Nc3-d1 Rd2xd1 Bb3xd1 Qc1xd1 Rh8-g8+
Kg6-f6 b2-b3 g5-g4 Rg8-f8+ Kf6-g7 Rf8-c8 Qd1-d7 Rc8-c1 g4-g3) -7.40/17 6}
Qd3+ {(Qd4-d3+ Kb1-a1 Qd3-f1+ Ka1-a2 Qf1-g2 Ka2-a1 Kg7-g6 Rh5-h8 Rd2xb2
Rh8-g8+ Kg6-h7 Rg8xg5 Qg2-d2 Nc3-b1 Qd2-c1 Bb3-a2 Rb2xa2+ Ka1xa2 Qc1xg5
Nb1-c3) +9.00/15 9} 38. Ka1 {(Kb1-a1 Qd3-f1+ Ka1-a2 Qf1-c1 Nc3-d1 Kg7-g6
Rh5-h1 g5-g4 Rh1-g1 Rd2xb2+ Nd1xb2 Qc1xg1 Bb3-d5 g4-g3 Bd5xb7 g3-g2 Nb2-d3
Qg1-d4 Bb7xg2 Qd4xd3 Bg2-b7 Qd3-e2+ Ka2-a1) -7.92/21 6} Qf1+ {(Qd3-f1+
Ka1-a2 Qf1-g2 Nc3-d1 Rd2-d3 Nd1-c3 Rd3xc3 Rh5xg5+ Qg2xg5 b2xc3 Qg5-d2+
Ka2-b1 Qd2xc3 Bb3-c2 b7-b5 Kb1-c1 Qc3-d4 Bc2-b3 a6-a5 Kc1-c2) +9.48/16 7}
39. Ka2 {(Ka1-a2 Qf1-c1 Nc3-d1 Rd2-e2 Bb3-a4 Re2-e5 Nd1-c3 Re5-a5 Rh5-h2
Qc1-f4 Rh2-c2 b7-b5 b2-b3 b5xa4 Nc3xa4 g5-g4 Ka2-b2 g4-g3 Rc2-g2 Ra5-d5
Rg2-c2) -8.54/21 6} Qg2 {(Qf1-g2 Nc3-d1 Rd2-d3 Nd1-c3 Rd3xc3 Rh5xg5+ Qg2xg5
b2xc3 Qg5-d2+ Ka2-b1 Qd2xc3 Bb3-c2 b7-b5 Kb1-c1 Qc3-d4 Kc1-b1 a6-a5 Kb1-a2
a5-a4) +9.52/16 7} 40. Ka1 {(Ka2-a1 Rd2-d8 Nc3-e4 Qg2xe4 Rh5xg5+ Kg7-h6
Rg5-g1 Rd8-d2 Rg1-c1 Qe4-g2 Rc1-b1 Qg2-f2 Ka1-a2 Kh6-g6 Bb3-e6 Qf2-d4
Be6-b3 Kg6-g5 Ka2-a1 Qd4-b6) -8.28/19 6} Kg6 {(Kg7-g6 Rh5-h8 Rd2xb2 Bb3-d5
Qg2-d2 Bd5-e4+ Kg6-f6 Rh8-f8+ Kf6-e6 Be4-d5+ Ke6-e7 Rf8-f7+ Ke7-e8 Nc3-b1
Qd2-c1 Bd5-a2 Rb2xa2+ Ka1xa2 Qc1-c2+ Ka2-a1 Ke8xf7 Nb1-a3) +10.99/15 5} 41.
Rh8 {(Rh5-h8 Rd2xb2 Bb3-d5 Qg2-d2 Bd5-e4+ Kg6-f7 Rh8-h7+ Kf7-g8 Rh7-c7
Rb2-b4 Rc7-c8+ Kg8-f7 Rc8-c7+ Kf7-e6 Be4-d5+ Ke6-d6 Bd5-b3 Kd6xc7 Nc3-d5+
Kc7-c6 Nd5xb4+ Qd2xb4 Ka1-a2 g5-g4 Ka2-b2) -10.97/19 9} Rxb2 {(Rd2xb2
Bb3-d5 Qg2-d2 Bd5-e4+ Kg6-f6 Rh8-f8+ Kf6-e6 Be4-d5+ Ke6-e7 Rf8-f7+ Ke7-e8
Nc3-b1 Qd2-c1 Bd5-e4 Ke8xf7 Be4-d5+ Kf7-f6 Bd5-e4 g5-g4) +13.17/14 5} 42.
Bd5 {(Bb3-d5 Qg2-d2 Nc3-b1 Qd2xd5 Ka1xb2 Qd5-e5+ Kb2-c2 Qe5xh8 Kc2-d3
Qh8-h3+ Kd3-d4 Qh3-g2 Kd4-c3 Qg2-c6+ Kc3-d3 Qc6-d5+ Kd3-c2 g5-g4 Nb1-d2
g4-g3 Kc2-c3) -12.56/18 8} Qd2 {(Qg2-d2 Nc3-b1 Qd2-b4 Bd5-e4+ Kg6-f6
Rh8-f8+ Kf6-e7 Rf8-b8 Qb4-d4 Rb8-c8 Rb2-c2+ Rc8-c3 Rc2xc3 Nb1xc3 Qd4xc3+
Ka1-a2 Qc3-c4+ Ka2-b2 Qc4-b4+ Kb2-c2 Qb4xe4+ Kc2-d2) +13.52/13 5} 43. Nb1
{(Nc3-b1 Rb2-a2+ Bd5xa2 Qd2-d4+ Nb1-c3 Qd4xc3+ Ka1-b1 Qc3xh8 Kb1-c2 g5-g4
Ba2-c4 g4-g3 Bc4-d3+ Kg6-g5 Bd3-e4 Kg5-f4 Be4-d5 Qh8-d4 Bd5-h1 Qd4-f2+
Kc2-d1 Qf2-g1+ Kd1-d2 Qg1xh1 Kd2-e2) -14.53/17 6} Qb4 {(Qd2-b4 Bd5-e4+
Kg6-f6 Rh8-f8+ Kf6-e7 Rf8-b8 Qb4-d4 Rb8-c8 Rb2-c2+ Rc8-c3 Rc2xc3 Nb1xc3
Qd4xc3+ Ka1-a2 Qc3-c4+ Ka2-b1 Qc4xe4+ Kb1-c1 Qe4-e1+ Kc1-b2 g5-g4)
+15.55/13 4} 44. Be4+ {(Bd5-e4+ Qb4xe4 Rh8-g8+ Kg6-h7 Ka1xb2 Qe4-g2+ Kb2-c1
Kh7xg8 Nb1-d2 g5-g4 Kc1-c2 Qg2-c6+ Kc2-d3 g4-g3 Kd3-e3 g3-g2 Nd2-f3 Qc6-c3+
Ke3-e2 Qc3xf3+ Ke2xf3) -11.93/17 6} Kf6 {(Kg6-f6 Rh8-f8+ Kf6-e7 Rf8-b8
Qb4-d4 Rb8-c8 Rb2-c2+ Rc8-c3 Rc2xc3 Nb1xc3 Qd4xc3+ Ka1-a2 Qc3-c4+ Ka2-b2
Qc4-b4+ Kb2-a2 Qb4xe4 Ka2-b2 g5-g4) +15.72/14 7} 45. Rh5 {(Rh8-f8+ Kf6-e7
Rf8-b8 Qb4-d4 Rb8-c8 Rb2-b3+ Nb1-c3 Qd4-d2 Nc3-a4 Rb3-a3+ Ka1-b1 Qd2-b4+
Na4-b2 Qb4xe4+ Rc8-c2 Ra3-c3 Kb1-a1 Rc3xc2 Ka1-b1 Rc2xb2+ Kb1xb2 g5-g4)
-M427/17 8 White resigns} *
[/pgn]

I think you should listen to the wisdom of Don on this one. Alekhine is no authority at 3200 ELO level...