Rybka and Fritz evaluate this as drawn in 3s. Crafty misevaluates this as won even at 60s. Take engine opening and endgame analysis with a grain of salt.
strange computer evaluation
Sort:
likesforests wrote:
Rybka and Fritz evaluate this as drawn in 3s. Crafty misevaluates this as won even at 60s. Take engine opening and endgame analysis with a grain of salt.
don't even need an engine for the final position, just a tamilov endgame database.
Yup, 6-man tables solve the final position. Unfortunately, 6-man tables don't seem to make Crafty any better at evaluating the initial position.
Yes, looks that computers, even with nalimov tables, are still poor at endings.
I remember once I played here with a cheater a blitz game. I was two pawns down, but it was an ending with opposite colors bishops. Soon I reached a drawing position and I began moving my bishop back and forth. And I offered my opponent a draw. The poor cheater however continued to move on and on for another 40-50 moves :)) After the game, I took an engine and saw that the end position was evaluated as a decisive advantage for my opponent (+4 or something). Although a human would have evaluated it on the spot as a dead draw.
Two of the neat features of Aquarium (an engine UI) are "Randomizer" (play out a position 20x in 2 minutes) and IDeA ("partially" analyzes to depth 30-50 quickly). These tools only work with Rybka 3 and often reveal drawn endings that pure analysis cannot. But they require the human to participate (minimally) in creating the analysis.
Here is a position from a game I played:
To me it looks that this position is a dead draw. In the game I chose one of the many possibilities of forcing the draw.
The surprise came after the game, when chess.com engine evaluates the position as winning for White. My move 1. Kc7 is labeled as a mistake, instead the computer suggests this line:
And says that "white has decisive advantage".
Really?? I see that the game is a draw, where does the computer find that decisive advantage?
What do you think?