click for larger view

This is where Alekhine played 29...b4.

Stockfish says that much stronger is <29...Qh5 30.Nf2 Qe2 31.Nf1 Qe1 32.Rxa8 Rxa8 33.Qxe1 Nxe1>.

click for larger view

Which SF says is -6.35.

There is something very strange about this game. I suppose the best word is <ersatz>. Would any moden master play like Bogo did? Who would think of Bh1 followed by Rg2?

I can imagine some British pressman asking Bogo what his game plan would be for today's game against Alekhine:

Understand that our "full game analysis" has four modes:

• Blunders only
• Broad evaluation
• Intensive evaluation
• Annotate every move

As of right now, the default for analyzed games in the database is "Intensive", which means that when the eval is near 0.00 it takes a 0.50 difference to warrant a comment. As the eval strays from 0.00, the threshold increases. For instance, if somebody is winning by +7.50 and they missed an opportunity to turn it into +8.75, or dropped a pawn to achieve +6.45, we're not going to slap them down with a question mark.

In the case of move 29...b4, Stockfish does indeed say in the notes above <29...b4-+ -6.35 (40 ply) after 29...Qh5 30.Nf2 Qe2 31.Nf1 Qe1 32.Rxa8 Rxa8 33.Qxe1 Nxe1> agreeing that 29...Qh5 is stronger. (When it uses the word "after" like that, it is saying, "This is the evaluation of the position after 29...Qh5, even though ...Qh5 was not played.")

But is it really "much stronger?" If it was, our software should award a question-mark. The test is to analyze the position after 29...b4: — if the evaluation suddenly dropped to -2.85 or something, then yes, Alekhine was being too fancy for his own good. But if it remains in the neighborhood of -6.35 then no big mistake was committed.

So we look at analysis of move 30.? (This is very easy; just click on ...b4 and then use Olga's "ENGINE" link. You will NOT have to wait for the evaluation to complete; since we analyzed this game in advance our results are in the database for instant gratification.)

It shows that with what SF regards as best play the game would continue <1) -7.13 (40 ply) 30.Rxa8 Qxa8 31.Qc2 Ne1 32.Qb2 Nxg2 33.Bxg2 Qa3 34.Bf1 Ra8 35.c5 Qxb2> — which would imply ...b4 is even better for Black than the recommended 29...Qh5.

I believe this is an example of a horizon-effect in which a position seems to be much better/worse the longer it is analyzed. Moreover, the point is that from move 29 onward Alekhine had the game well in hand at this stage and therefore there is no reason to criticize his play. (If, however, he missed a checkmate, it would be noted.)

click for larger view

Analysis by Stockfish 030916 64:

1. -+ (-8.96): 29...Qh5 30.g4 fxg4 31.Nf1 g3 32.Qa3 Rab8 33.Nc3 Nf2 34.Rxg3 Nxh1 35.Rg2 Qxh4 36.Qe7 Rg8 37.f5 Bxf5 38.Rxb5 Rxb5 39.cxb5 Qe1 40.Qb4 Bh3 41.Nd5 Qd1 42.Qd2 Qxd2 43.Nxd2 Bxg2 44.Kxg2 Nxd5 45.Kxh1 Re8 46.Nc4 Rb8 47.Nd2 Nxe3 48.Nxe4 Nf5 49.Nc3 Nxd4 50.Kg2 Nxb5 51.Ne4 Re8 52.Nc5 Kh7 53.Kg3 Kg6 54.Kf4 Nd6

2. -+ (-8.09): 29...b4 30.Qa1 Rxa5 31.Qxa5 Qa8 32.Nb3 Qxa5 33.Nxa5 Ba4 34.Rd2 Ra8 35.Rxd3 Rxa5 36.Rd2 Bb3 37.Nf2 Ra1+ 38.Kh2 Bxc4 39.Rc2 Bd5 40.Rxc7 b3 41.Rc8+ Kh7 42.Rb8 Ra2 43.Kg1 b2 44.Bg2 Ra1+ 45.Kh2 b1Q 46.Rxb1 Rxb1 47.Bh3 Rb2 48.Kg1 Kg6 49.g4 Re2 50.h5+ Kf7 51.g5 Nxh5 52.gxh6 gxh6 53.Bxf5 Re1+ 54.Kg2 Rxe3 55.Ng4 Nxf4+ 56.Kg1 Rg3+ 57.Kf2 Rf3+ 58.Ke1 Ng2+ 59.Kd2 Rd3+ 60.Kc2 Ne1+ 61.Kc1 Rxd4 62.Nxh6+ Kf6 26.09.2017)

It looks like 29...Qh5 is slightly more efficient than 29...b4, but not much.

Stockfish 030916 64

click for larger view

Analysis by Komodo 10 64-bit:

1. -+ (-8.96): 29...Qh5 30.g4 fxg4 31.Nf1 g3 32.Qa3 Rab8 33.Nc3 Nf2 34.Rxg3 Nxh1 35.Rg2 Qxh4 36.Qe7 Rg8 37.f5 Bxf5 38.Rxb5 Rxb5 39.cxb5 Qe1 40.Qb4 Bh3 41.Nd5 Qd1 42.Qd2 Qxd2 43.Nxd2 Bxg2 44.Kxg2 Nxd5 45.Kxh1 Re8 46.Nc4 Rb8 47.Nd2 Nxe3 48.Nxe4 Nf5 49.Nc3 Nxd4 50.Kg2 Nxb5 51.Ne4 Re8 52.Nc5 Kh7 53.Kg3 Kg6 54.Kf4 Nd6