Hi Bill,
OK, then, maybe I didn't really understand what the expression "straw man" means.
<<However, take some intermediate position from a game, insert it into the solver as the starting position, erase the TT. The solver's decision will be true for this starting position, but this decision may not be dependable for the original deal.>>
Oh, OK, I'd just say "of course not". From any winnable starting position, to be sure, any number of unwinnable intermediate positions might be derived. Except if they are unwinnable, they aren't "intermediate", they're "terminal". So we shouldn't be talking about "intermediate" positions at all, we should be talking about "subsequent" or "derived" positions. One task of the solver is to figure out whether a "subsequent" position is "intermediate" or "terminal".
But one thing that is useful about analyzing "intermediate" positions from a solution is to test for false impossibles, because false impossibles, especially with the four-freecell game, are so scarce as to be very hard to find. But the intermediate positions on the path to many solutions often, some of them, skate very close to the edge of unwinnability, and hence are much more prone to be falsely declared impossible than an original position. I have a variation in FcPro that takes a solution, works its way back through the solution one move at a time, and tests the solver at each position. I was able to identify some false impossibles from Shlomi's solver that way. In fact, that gives me an idea. I should resurrect that technique to apply it to his solver again, to see whether or not his "atomic move" technique really eliminates the false impossibles.
Best regards, --------------Adrian
Received on Wed Jul 03 2002 - 20:07:41 IDT