Author: blass uri
Date: 11:50:48 06/08/99
Go up one level in this thread
On June 08, 1999 at 09:19:33, Dave Gomboc wrote: >On June 08, 1999 at 03:30:37, blass uri wrote: > >> >>On June 08, 1999 at 00:47:02, Dave Gomboc wrote: >> >>>On June 07, 1999 at 21:22:45, Mike Castañuela wrote: >>> >>>>On June 07, 1999 at 19:08:42, Dave Gomboc wrote: >>>> >>>> >>>> >>>>>On June 07, 1999 at 16:41:56, Dann Corbit wrote: >>>>> >>>>>>On June 07, 1999 at 16:35:24, blass uri wrote: >>>>>>[snip] >>>>>>>In part of these 7% of the cases(including your 3 examples) underpromotion is >>>>>>>needed only to win faster but is not important practically for the result of the >>>>>>>game. >>>>>>The same would be true of a program that takes a bishop to go up +3 instead of >>>>>>taking a rook to go up +5. Which program would you rather have? >>>>> >>>>>The one that knows how to get to the position where it can win either way. >>>>> >>>>>Dave >>>> >>>>But if one way (that can be unique to the victory path) >>>>of those all ways is not feasible, then what? >>>>Too, underpromotions weigh in esthetic terms, not only practical. >>> >>>If? Well, then you miss the win. Tough one. But how realistic is this "if"? >>>Not very. However, if accounting for underpromotions would slow down Junior by >>>a few percent in every game it plays, that's not an "if", that's a definite >>>relative penalty, in every single game. >>> >>>It's a chess engine. Aesthetics are for user interfaces, not engines. ;-) >>> >>>Dave >> >>I think that there is no reason to slow Junior down by a few percent to consider >>underpromotion. >> >>You can use 0.1% of your time to compare between search with underpromotion and >>without underpromotion and if there is a difference then to use an evaluation >>function that allow underpromotions. >> >>In this case you lose less than 0.1% of your time and I believe that you play >>slightly better. >> >>Uri > >You need to identify the need for underpromotion while the actual underpromotion >is out at the leaf nodes. Otherwise you will choose a different variation and >not require it. Checking for it only at the root isn't terribly useful in the >context of playing a game. > >Dave I believe that in significant part of the cases seeing the underpromotion 1-4 plies before it happens could save Junior from stupid mistakes. 3r3r/2QPqpkp/p5p1/1pN5/2p5/8/PP3PPP/3R1K2 b - - 0 1 This is a position from Junior's game against Nimzo99 in enrique's tournament Junior could play Qf6 insteqad of Rhe8. I do not know if it is a good example because I am not sure if Qf6 could save Junior. Uri
This page took 0 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.