|zanzibar: Looking in detail at the accuracy of Clarke's annotations:|
<Only a closer examination reveals White's cunning idea-12...Bxb4+ 13.Ke2!! The Bishop would then have nothing better than to go back to e7, after which White could either recover the pawn at once by Bxf6 and Bxh7+ or, still stronger, continue 14.Rad1 Qc7 15.Ng5 h6 16.Bh7+ Kh8 17.h4 with a tremendous attack. >
First of all, 14...Bd7 is likely better (giving equality when 14...Qc7 is +0.72/25 w sf8).
Also, the engine doesn't think 15.Ng5 is best for White.
But let's skip all that to evaluate the position where Clarke assesses White has a "tremendous attack".
(Black to move after 17.h2-h4)
click for larger view
r1b2r1k /1pq1bppB/p1n1pn1p/6N1/2N4P/P3P3/1BQ1KPP1/3R3R b - - 0 17
This is interesting to me because I often try to attack with similar "wild abandon". It certainly looks strong on the surface, but does White even have compensation for the pawn?
The engine evals the position as -0.81/26, showing that Black has adequate defenses.
In fact, Black seems to be able to immediately defang the attack with 17...Na5, "pinning" the c4-knight.
Then if 18.Ne5 Qxc2+ 19.Bxc2 and White's attack is much harder without the queen.
If 18.Nxa5 Qxa5 (eyeing a3), and it seems that White does best retreating with 19.Bd3 e5 (screening the f6-knight).
In fact, the engine likes to acknowledge White should retreat, and offers 18.Bd3 as perhaps best.
The point is the difficulty of assessing an attack. Black cedes White the lsq's, but White can't do anything on the dsq's. Nor can White force a file open to the king, who just sidesteps to the corner.
And without either rook or the QN getting into play, the king sits safe.