I was analysing an opening position with multiple lines displayed in
Shredder and Rybka and observed some curious jumps in the evaluation
function depending on the parity of side to move. I ran Shredder for
about 24 hours and Rybka for 4 days. The position is from the English
opening A33 :

r1bq1rk1/pp1p1ppp/2n1pn2/8/1bPN4/2N3P1/PP2PPBP/R1BQ1RK1 b - - 0 8

With black to move. Conventional wisdom according to the references I
have to hand are that
8. ... a6 is most often played but not especially successful
8. ... Qe7 is the strongest continuation
8. ... d5 and 8. ... Qa5 are roughly equal in merit
8. ... Qb6 has a poor history (but only 3 games)

Qe7 is the clear winner. But the most curious lines are
Re8, Qa5, Be7 which flap around by 10cp with each ply.

score in cp
22 21
ply ply line
22 22 8. ... Qe7 9. Bg5 h6 10. Bxf6 Qxf6 11. e3 Qe5
28 24 8. ... Qb6 9. Nb3 Bxc3 10. bxc3 Qa6 11. Qd3 b6
30 8. ... Re8 9. Ndb5 d5 10. cxd5 Nxd5 11. Nxd5 exd5 12. Be3
Be6
41 10. Bg5 dxc4 11. Qa4 Qa6
28 31 8. ... Bc5 9. Nb3 Be7 10. e4 a6 11. Qe2 d6
42 32 8. ... Qa5 9. Ncb5 d5 10. a3 Be7 11. Nb3 Qd8
35 35 8. ... Ng4 9. e3 f5 10. Bd2 Be7 11. Rc1 Rb8
34 36 8. ... d5 9. cxd5 exd5 10. Be3 Bxc3 11. bxc3 Bg4
37 36 8. ... Bd6 9. Qd3 Ne5 10. Qc2 Nxc4 11. Ncb5 a6
45 37 8. ... Be7 9. Bf4 Nh5 10. Be3 f5 11. Qd2 Nf6
42 42 8. ... Qc7 9. Ndb5 Qb8 10. Bf4 e5 11. Bg5 a6
45 42 8. ... h6 9. Bf4 Nh5 10. Be3 Nf6 11. Qa4 Qe7
39 42 8. ... a6 9. Bf4 Re8 10. Nxc6 bxc6 11. Bd6 Qa5

One surprise is that Re8, Bc5, Ng4 and even Bd6 look playable and a6
which was historically the most common move is not even in the top
ten. d5 which releases the tension is also well down the list. It
would be interesting to hear comments on these lines by experts in
opening theory and/or the results of deep analysis of this position on
other engines. I'd be grateful for some guidance on for example why
8. ... Bd6 isn't penalised more.

Best regards,
Martin Brown