On Thursday I returned home from an evening out around 11:30pm. The puzzle of the day on chessgames changes at midnight on the East Coast, so at a few minutes past midnight I checked the new puzzle of the day.

I try hard to solve these puzzles, and the ones later in the week are harder than the Monday puzzles. So the Friday puzzle I knew was going to be hard.

I tried to solve it but could not. I looked at 29...Qb2 and 29...Nb2 but could not figure out the win. Now I clicked the link to the game and observed the solution, 29...Qb2 and noted White's response, 30.Bc2.

I wonder what Rybka says about this, I thought. I downloaded the pgn file for the game and loaded the position into Rybka. I set Rybka to "Ultra slow" mode and set it to look at the best 2 moves.

Rybka usually verifies these puzzles in seconds, I am afraid to say. In this case, I saw that White's response, Bc2, was scored as roughly -1.0, and steadily increased as the game progressed. But there was another move, 29...Qc2, that was scoring -0.25.

Hmmm I thought. Sliding forward into the 30.Qc2 line, I saw that if Black played 30...Qxa1, White could play 31.Bxa3 and amazingly hold, with a score of 0.00.

I looked at the other possibility for Black, 30...Rxf4, and it was scoring -0.2, with White responses 31.Bxf4 and 31.Qxd1 as possibilities. Could White hold here?

I went to the chessgames site and posted:
29...Qb2 30.Qc2! Now who can show the win for Black? Hint - if you play 30...Qxa1 I play 31.Bxa3.

I went back to the 30...Rxf4 line and picked 31.Bxf4 as the likely continuation and watched the variations.

Tariqov prompted me to finish looking at the 30...Qxa1 line, only because his icon was rather sinister looking and his bio says that he is under 14 years old: Line A: 30...Qxa1
31.Bxa3 Rxa3

(31...Rxf4
32.Bd6+ Kc8 (32...Kb6 33.Qc5+ leads to mate; 32...Kd8 33.gxf4 fav. white) 33.gxf4 b6 34.Qxd1 Kb7
35.Qc2 fav. white)

32.Rxf8 Nc3 33.Qc1 (or g6) and now:

a) 33...Qxb1 34.Qxa3 Qe4+ 35.Kh2 Nxe2 36.Qd6+ Kb6 37.Qc5+ per.chk

b) 33...Nxb1 34.g6 Qxa2 35.Qc5 Qxe2 36.Qd6+ per.chk

c) 33...Nxe2 34.Qxa3 (...) 35.Qd6+ per.chk

d) 33...Ra8 fav. white after 34.Rxa8

nolanryan challenged me to solve the 29...Qb2 30.Qc2 Rxf4 line. I watched in disappointment as my first choice, 31.Bxf4, promptly fell apart under computer analysis. That left the unusual looking Qxd1, which appears to hold. The17thpawn and pittpanther prompted me to finish the analysis: Line B: 30...Rxf4 31.Qxd1 and now: a) 31...Rxf2 32.Kxf2 Qxa1 33.Qd3 Black can make no progress

b) 31...Qxa1 32.Bxf4 b5 33.Qd3 b4 34.Bd2 and Black is stuck

c) 31...Rxd4 32.Nxd4 Qxa1 (32...Qxd4 33.Qxd4 Nxd4 34.g6 +1.09/18) 33.Nb5+ Kb8,c8,or d8 34.Qc2 favors white )

Life Master AJ zeroed in on my a) line (as well as patzer2) as being the critical line in this variation. Dozens of hours of computer time have been unable to produce a single line that wins for Black.

Many other kibbitzers on the puzzle of the day site helped here. It appears that 29...Qb2 has been refuted by 30.Qc2! which is an amazing move. What do you think?