1. e4 e5 2. Nf3 Nc6 3. Bb5 a6 4. Ba4 Nf6 5. Qe2 Be7 6. c3 b5 7. Bb3 d6 8. a4 Bb7 9. O-O O-O 10. d3 h6 11. Re1 Nh7 12. d4 Ng5 13. Bxg5 hxg5 14. d5 Na7 15. Rd1 g4 16. Ne1 Bc8 17. axb5 Nxb5 18. Bc4 Bd7 19. Nc2 Qc8 20. Nd2 g6 21. Nb4 Bg5 22. Ra5 Qb7 23. Nb3 Kg7 24. Rda1 f5 25. Nxa6
'White wins a pawn and completely occupies the Q-side. It seems that Black has failed to create any counter chances, but that is only an illusion. Now the game becomes unexpectedly sharp'—Karpov.
Black's only chance. He sacrifices a piece and gets an ominous attack due to the fact that White's pieces are tied down'—Karpov.
Markland gives this an exclamation mark. Karpov remarks that this is best and provides an analysis of an alternative variation: 27. Bxd7 Rxf2 28. Qxf2 Be3 29. Qxe3 Qxe3+ 30. Kh1 Rh8 31. Bxg4 Qg3 32. Bh3 Rxh3 33. gxh3 Qf3+ 34. Kg1 Qe3+ 35. Kf1 Qf3+ and draws. If 36. Ke1 e3! .
'Black's attack is over and best here would be to resign'—Markland. However, this may not be at all correct and possibly Karpov should even has chosen his alternative at move 27.
'Now I had to think again.'—Karpov.
Karpov was concerned about being mated playing this move but considered 'Black has an excellent position' after the alternative 30. Kf1 Rf2+ 31. Ke1 cxb6 (in fact it's even worse than that; the engine gives 31... Bxb5 and mate in 8).
The modern engine again sees more: Karpov had considered the possibility of a mating attack starting with 30... Rh8 and he was going to rely on 31. Bxd7, but after 31... Bf4 32. h3 Rxh3+ 33. gxh3 g3 (precise move order is important here; Karpov had only considered 33... Rh2+ 34. Kg1 g3 35. Re1 winning for White) 34. Kg1 (the threat was Be3 and Rh2 mate) 34... g2 35. h4 Be3+ 36. Kh2 g1=Q+ and Black mates next move.
If Black had gone 30... Rh8 White needed to play 31. Bxe2 when, as Karpov had foreseen, 31... g3 32. h3 Bxh3 33. gxh3 Rxh3+ 34. Kg2 Rh2+ 35. Kxg3 Bf4+ 36. Kg4 Rg2+ 37. Kh3 Rh2+ 38. Kg4 is a perpetual check.
Karpov simply remarks: 'White is a rook ahead.' However it appears that things are still not that simple.
Probably an error. The engine prefers 34. Na6 assessing it at about -0.5.
It would seem that only here did Moles go wrong. If instead 34... Bxd5 White will still have great difficulty unravelling his pieces. Analysing this position is above my pay grade but the engine is confident that White is at least -2.0.
Here's another little surprise. We are following the game score as given in the two books. However many modern websites and databases (e.g. the ICU) give the end of the game as 37... Rb3 (which would give away a Rook) 38. Ne3 Rb5 39. g3 Rc5 40. Kg2 Rb5 41. Rc1. A hat tip to the Kibitzers at Chessgames.com.
'David's instinct regarding this game was absolutely right and as for his paygrade—he should ask for a raise.'—Doyle.
Annotator(s): David McAlister (see reference); Tony Doyle (via email, December 25, 2016).
Source(s): K. J. O'Connell, D. N. L. Levy, and J. B. Adams, The Complete Games of World Champion Anatoly Karpov (Batsford 1976) pp. 54-55 (game 0509); P. R. Markland, The Best of Karpov (Oxford Chess, 1975).
Reference(s): David McAlister, Karpov-Moles, Groningen 1967-8, IRLchess, October 29, 2014.