Author: Dann Corbit
Date: 15:26:30 09/14/00
Go up one level in this thread
On September 14, 2000 at 18:20:20, Ricardo Gibert wrote: >On September 14, 2000 at 17:17:23, Dann Corbit wrote: > >>On September 14, 2000 at 16:52:38, Ricardo Gibert wrote: >>[SNIP] >>>There is an easy win with 1.Nf6. Of course 1.Qxh6+ mates by force, but I think >>>that is irrelevant. I'm always satisfied by an easy win (unless there is an easy >>>mate that costs me little additional effort). >>> >>>If you think programs should feature a mate finder mode for problemists, that's >>>fine, but in a regular chess game, that is not really practical. I do not view >>>1.Nf6 as a failure to solve the position. A human opponent would probably resign >>>on the spot against 1.Nf6, so who cares? >> >>Your analysis is correct: >> >>[D]r1bq1r2/ppp2pkp/3p2nn/2bN2NQ/2B1P3/8/PPp3PP/R1B2R1K w - - acd 12; acn >>-290540850; ce 657; pv Nf6 Qxf6 Rxf6 Kxf6 Nxh7+ Ke7 Bxh6 Rh8 Bg5+ Kd7 Bxf7 Rxh7 >>Qxh7 Ne5; pm Nf6; bm Qxh6+; id "BWTC.0134"; >> >>But does not answer the question at hand: How hard it is to search 16 plies >>forward. >> >>I picked these two positions because they are interesting *to me* but most of >>the other positions (in my challenge above) will not be nearly so easy to see >>any sort of solution. >> >>Everyone seems to want to dance out of the way of the real question which I was >>addressing: >>"How hard is it to search 16 plies?" >> >>So far the only answers I have received are "Who cares -- these are easy mates?" >>Which (in fact) I obviously knew -- see the thread title! >> >>The question I want answered is, "How hard is it to search 16 plies in a busy >>position?" >> >>If someone does not want to try the checkmates, then ignore them and try the >>others. > >You're failing to give an important issue suffient weght. An important issue is >what constitutes a success. In the example given, 1.Nf6 fills the bill more than >adquately. The move 1.Nf6 constitutes a successful solution IMO. Finding the >fastest mate is a significantly different problem. It is a specialty of interest >to problemists, but not to a practical player. A program like Chest is better at >finding mates, but nobody _should_ suppose that capability makes for a stronger >program. > >I can just picture someone tweaking their program so that it can score "better" >on these type of positions and thereby unwittingly making their program weaker, >hence the motive for my post. > >As far as how hard it is to search 16 plies, I don't think the example you give >is a good one. I can see how a program might prune away from its search a line >leading to mate, because it has already found a line that is quite strong. In a >sense, it is getting "misled" by the presence of a perfectly strong move. It >would be better to use an example where there is only one continuation that >leads to a win rather than several. > >I don't want a program to waste its time looking for a "better" winning line. How do you know when it is time to stop searching. Doesn't your program ever fail low? >Particularly when the "better" winning line is likely to not be the main >continuation. I think this is the case most of the time in a programs search. Any of the positions for which you find checkmates, or even "a really good move" before ply 16, just remove them from the test set and pretend they don't exist.
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.