The first position is from Liublinsky-Botvinnik, Moscow Championship, 1943. Black has weak P’s on the Q-side and White is threatening to trade R’s, so the question is, how did Botvinnik meet both of those threats?
The solution is 1…Rd4! which is what both Botvinnik and Houdini played. (Fritz 1...Rd4 0.88). The next position was a made up one with White to play.
The solution is 1.Rxc8+ Rxc8 2.g3! placing the B on the h3-c8 diagonal which is what Houdini suggested. Very good so far! (Fritz 1.Rxc8+ 0.92)
The next position is from Lputina-Minasian, Protvino Zonal, 1993. For the price of a P, White hit the dark squares and damaged Black’s Q-side. What did he play?
1.c5! threatens to seriously undermine Black’s influence on the dark squares after Bxa6. After 1…dxc5 (1…d5 2.Bxa6 bxa6 3.Qe2 Bc8 4.Ng1 and White prepares to evict Black’s only decent piece) 2.Bxa6 bxa6 3.Na4 and Black had to be careful. If 3…c4? 4.Nc5 h6 5.Qc3 is good for White. The game went 3…Be6 4.Nxc5 Bd5 with a good game for White. (Fritz 1.c5 0.08)
After 4 minutes Houdini suggested White was about 1/4 P ahead and recommended 1.c5 dxc5 2.Bxa6 bxa6 3.Na4. Upon actually arriving at the position after 2…bxa6 and thinking another 10 minutes, Houdini was having some difficulty arriving at a definite conclusion as it had been considering moves like 3.Na4, 3.Kb1 and 3.dxc5. Several times it changed its evaluation and preference and for a while even thought Black was slightly better. Finally after 15 minutes its opinion was that 3.Na4 was best after all and favored White by about 1/4 P.
I also was curious what it thought about 1.c5 dxc5 2.Bxa6 bxa6 3.Na4 c4? So gave it 2 minutes to ponder that. It agreed with the analysis that 4.Nc5 h6 5.Qc3 is good for White, and thought White was only about 1/4 P better in that case, too
And finally this position from Bykhovsky-Smirin, Beijing, 1991, Black to play:
In this fairly even position Black’s isolated P on e5 and White’s control of d5 suggests White stands a little better. However, Smirin was able to exploit White’s somewhat weakened K-side and at the same time deny White use of the e4 square and create an outpost of his own on e5 with 1…e4! After 6 minutes Houdini finally seems to have faltered in its evaluation
After about 10 minutes it was suggesting 1...Qf7 with an evaluation of -0.16. Its second choice was 1…b6 evaluated at -0.10 and its third choice was the correct 1...e4 which it evaluated at -0.09. At least it was correct in thinking the position was slightly in Black’s favor.
Just out of curiosity, I let Fritz and Fire 1.5 Xtreme take a gander at the position and Fire agreed with Houdini that 1…Qf7 was best and favored Black by about 1/4 P while Fritz 12 got it right when its suggestion was 1…e4 which it evaluated at 0.00!
Other engines: Spike 1…Qf7 (0.06), Ivanhoe 1…Qf7 (-0.17), Critter 1…Qf7 (-0.19) and Komodo 1…Qf7 (-0.20).
I don’t know what one can make of all this but one point seems to be that even though the engines do quite well at selecting positional moves their evaluations, at least in this little experiment, tended to show they did not appreciate the positional nuances to the same extent that the GM’s did. Houdini missed one of the positions and had some trouble making a decision in one position but all-in-all I think it qualifies as a positional genius. The only real difference seems to be the degree of optimism shown by the various engines.
The fact that Fritz got the last solution right prompted me to go back and let Fritz 12 look at all the positions. That’s why Fritz’ score has been attached to all of the puzzles, but without comment. Guess what? Fritz 12 is not the highest rated on engine rating lists and it has always come out second best when matched against Houdini in Blitz games but it got all four solutions right. That makes me wonder...exactly how good is Fritz 12 if you allow it to think for several minutes instead of making it play blitz all the time?!
When it comes to analyzing games, it really doesn’t matter much what engine we average players use so I have just wasted a couple of hours proving what we already knew, but it could have been worse…I could have been doing laundry or running the vacuum cleaner or watching television or any of a dozen other things that wouldn’t have been quite so enjoyable.