Author: Pete R.
Date: 15:47:56 07/15/00
Go up one level in this thread
On July 15, 2000 at 17:40:00, Ralf Elvsén wrote: >I have looked at most of Crafty's evaluation. Let me first say that >this is no Crafty bashing. I like the program. That said, the chess >content encoded is very crude. It is heuristics which may or may not >apply to the position in question. None of the little pieces of evalutation >does a better job than an average clubplayer. It is in combination >with the search it becomes so powerful. Of course I realize it is very >hard to balance such amount of code and I would fail miserably myself >if I tried. I am *not* complaining. This may be so, but how much greater is a GM's knowledge really? Being realistic and not worshipping the talented, that is. ;) They also have a lot of simple rules of thumb used in combination with good visualization ability. What they really have is the intelligence to apply these rules appropriately given a particular position. They can look and say, I might be able to force a complex of weak squares on my opponent, and then start calculating to see if it can be achieved. A computer will see the same position in the future, but may not choose this as best to shoot for because the weighting of this future position is drowned out by other terms in the eval function. This begs the question of whether a single eval function can be tuned to meet all circumstances. Perhaps certain terms should be ignored depending on the type of position, which of course the program would have to determine. I don't know really, but it seems possible that someone could hit upon such a good tuning that existing hardware is sufficient to do the job. Or maybe it requires a lot more eval parameters, like DB. Maybe the DB eval could be distilled into simpler functions, assuming the details are published. But that's the underlying question, do you need a vastly more complex eval function or just a bit more with perfect tuning? >I think we need many more plies >before the program can "teach" us what a position is really about, or >alternatively a much more precise (and slower) evalutaion. I agree, but my feeling is that more plies may not prevent the "stupidity" shown in some of these anticomputer games. So we are back to the evaluation, and how much added complexity will achieve over simply better tuning (there must be some overlap there of course). >Let me take this opportunity to ask someone who knows: Has Ed Schröder >chosen to add "everything" in Rebel's evalution. I might have misunderstood >some of his statements. He seemed to be disappointed that more chess >knowledge could give a weaker program in comp-comp games. I.e. will he >add knowledge regardless of the implications against computers in order >to have a program which humans can enjoy more? Perhaps he should simply have two entirely separate evaluation functions, one for humans and one for computers. Perhaps he already has this in a sense with the anti-GM setting? In any case I think the kind of person who looks at the SSDF list is the kind of person you would find here in CCC. I might be wrong but I think most here are more impressed to see how well the programs fare against the top humans at standard time control. The differences on the SSDF rating list are not significant in my mind, particularly since they can be due to book lines, etc. Ultimately I think we want to see any program consistently make moves that the top human GMs would approve of, instead of falling for the same old anticomputer strategies, or making senseless moves. We want to see a program smart enough to take early countermeasures against such strategies, and just in general play good positional chess. That's the program I will trust to evaluate a position properly.
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.