Random Posts

Friday, June 24, 2016

Engines and Pawn Chains

     Who's better and by how much? The grandmaster says move A is the best, but a couple of different engine's suggest move B or move C is better. The grandmaster says move A is practically winning, but one engine says the position is equal while the other one says the grandmaster is wrong, his move is slightly inferior. Who do you believe? 
     Analyzing the position in this post taken from Forgacs vs. Tartakower, St. Petersburg, 1914 with different engines reminded me of the classic Abbott and Costello routine Who's On First? 
     
White to move

      It's clear black is operating on the Q-side while white is conducting operations on the K-side. The K-side attack is of greater importance since the enemy K is facing a potential mating attack and one would expect sacrifices to be necessary at some point. 
     While white's attack seems the more promising, his Pawn chain has all the lines closed so that his pieces cannot participate in the attack.  On the other hand black's P-chain is not weakened in any way, but notice that he has no pieces in the vicinity of his King with which to defend it. Therefore white has to find a way to break through and bring his pieces to bear on the black K. 
     Forgacs' moves 1.f5 and 2.g4 have been universally praised in the books as an example of a brilliant attack. However, analysis with Stockfish and Komodo reveals that Tartakower's defense was rather weak, but had he chosen a better defense the engine's didn't agree on the best continuation. What they did agree on was that white's best line was further preparing the f5 advance by first playing 1.g4. 
     Because some engines are only single core I limited all engines to one core for the comparison test; all detailed analysis was done with engines using three cores.  The engines were given 3 minutes to evaluate the position to see if any of them would discover Forgacs 1.f5 followed by 2.g4 and the subsequent attack on black's King which proved decisive. None did and they all wanted to play 1.g4 except Stockfish and Giraffe.

Engines selecting 1.g4. 
SmarThink (+/-) (0.72) 
Zappa (+/=) (0.63) 
Fritz 12 (+/=) (0.53) 
Komodo 8 (+/=) (0.46) 
Gull 3 (=) (0.39) 
Houdini 1.5 (=) (0.21) 
Rybka 2.3.2 (=) (0.20) 
Crafty (=) (0.02) 
Other first moves: Stockfish 7 1.Qg2 (+/=) (0.56) Giraffe 1.Bb1 (-/+) (-8.22) This is way wrong!

     Using Stockfish 7 and Komodo 8 I looked at the position after Forgacs played 1.f5 and both agreed that 1...exf5 was black's best reply, but after that things got really confusing. Both engines agreed that 2.g4 was best, evaluating the position at almost dead equal. But, after 3.Ng5 they could not agree on either the best more or the evaluation score! At least not in the hour or so I was fooling around with the position. 
     What can we make of this?! Was Forgacs' 1.f5 the best or was it the engine's 1.g4? After I played 1.g4, both engines gave different "best" lines and evaluation cores. I resorted to the old standby that I use in situations like this...a Shootout after both moves.

Results: 
First with Forgac's 1.f5 Both Stockfish and Komodo 8 scored +0 -1 =4 
Then with 1.g4 Stockfish scored +1 -0 =4 Komodo 8 scored +3 -0 =2 

Conclusions: 
1) Black has drawing chances after Forgacs' move, but after the engine choice white's winning chances are increased significantly. 
2) In some positions a laptop, a gaggle of engines and a couple of hours analysis is just not enough to determine the BEST lines or even how much better one side stands. 
3) What's practical in over the board play may turn out not to be the "best" according to the engines.

No comments:

Post a Comment