Author: Stephen Ham
Date: 09:54:50 03/01/04
Dear Readers, I've read that my fellow correspondence chess players often use the Deep Position Analysis feature of their CB engines. I'm not experienced in such matters but felt it was time for a test. So I selected a complex position from one of my correspondence games in progress. I had already sent my move the day before, so I felt that I knew what was best in my test position. First I set up the position in S8 and let it calculate in Infinite Analysis mode. The primary move at each depth (it went about 16-ply deep before I stopped it) was the move that I played in the game (it's also the most logical and natural move for humans). However, in move ordering, the second recommended line (this wasn't the send move in the recommended line, instead it was the second recommended line after the PV of the recommended line was displayed) was always a move that immediately dropped the a-pawn! While losing this pawn isn't fatal, and perhaps not even bad, it's certainly not a good move. The final evaluation is a clear advantage for White (me) in the main line. Then I set the same position up in Deep Position Analysis. I let it default to the factory settings for number of lines to be tested at 5'/ply. To my amazement, the pawn dropping move that was always second in the move ordering at Infinte Analysis mode never appeared as a candidate in the top 5-candidates. Also, while the same move was selected as best (mine), the evaluation showed that Black can probably equalize! So for a final test, I returned to Infinte Analysis mode and played each of my (human) responses to the best moves found by Deep Position Analysis and went 16-ply deep. The result was that White is again found to have an advantage, but this time S8 says it's a small one. Conclusion: there is a clear variance in move selection between Deep Position Analysis and Infinite Analysis. I think that Deep Position Analysis considered the best candidates, making it superior to the move ordering in Infinite Analysis mode. However, there's also a clear difference in the assessment of the same line, depending upon which analysis method is used. My human opinion is that the evaluation of both analysis methods was wrong and that the position is objectively better for my move with only a small edge. For the record, I've never liked S7's ability to evaluate positions (http://chesscafe.com/text/review365.pdf) and I don't think that S8 is any better. Has anybody else documented different move selection and different evaluations, depending upon which of the S8 analysis modes were selected? I suppose it's possible to find positions where the recommended line in Infinite Analysis mode differes from that from the Deep Posiion Analysis mode. Any thoughts, folks? All the best, Stephen
This page took 0.01 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.