Author: Robert Hyatt
Date: 22:43:08 11/21/02
Go up one level in this thread
On November 21, 2002 at 22:52:50, Vincent Diepeveen wrote: >On November 21, 2002 at 22:35:09, Robert Hyatt wrote: > >>On November 21, 2002 at 21:55:49, Vincent Diepeveen wrote: >> >>>On November 21, 2002 at 16:14:17, Robert Hyatt wrote: >>> >>>Bob, do you play at tournaments with programs getting a fixed >>>depth against each other or do you play with a clock? >> >>DO you _ever_ read? He is comparing _his_ program. Not his program >>vs another program. That makes your point pointless. If you know what >>I mean. > >comparing program A versus program A' with only difference an algorithmic >change then it's not fair that program A receives 20 times more time >for a move than program A'. Of course it isn't fair. However, he didn't do that, so what is the point? his verification search numbers (nodes) were smaller using R=3 with verification than using plain R=2, so the above comment of yours leaves me wondering what you are talking about? The two programs are searching about the same speed since they are the same except for the minor search detail of the verification search. That isn't going to affect NPS. Nor total time if the node counts stay close with both tests as they did... His games seemed to be real timed games of course, not depth vs depth. > >I bet conclusion will then be that alfabeta is useless and minimax is >better. > Not for me. But I'm not ready to write something off completely until I try it to be sure it doesn't work for me. >> >>> >>>I gladly play with diep at a fixed depth against crafty of course. >>> >>>You outsearch me by 2 or 3 ply (commercial programs 3-4 ply). >>> >>>If you give me like 15 times a move what you need a move, >>>then of course i appreciate the fair offer and take it for >>>the coming cct4 tournament in every game. I will not cheat >>>there. I will play with the default diep version if you >>>do with crafty too. We can appoint a fixed depth of 12 ply. >>> >>>That's fine with me. >>> >> >> >>Fine by me. I'll tune my extensions a bit however. Just name the time >>and place. fixed depth=12 plies. To show you how stupid such a comment >>is... Might take me a few hours to search 12 plies however... >> >> >> >> >> >> >> >>>>On November 20, 2002 at 19:02:49, Gian-Carlo Pascutto wrote: >>>> >>>>>On November 20, 2002 at 18:54:30, Omid David Tabibi wrote: >>>>> >>>>>>>Could you please compare (Adptv + small quiesc) vs (Vrfd +small quiesc) ? >>>>> >>>>>When I have more time. >>>>> >>>>>If you want more data, I expect others will post results >>>>>from their programs as well. Maybe those are more encouraging... >>>>> >>>>>>BTW, please allocate more time for each position. The deeper you go, the >greater will be the advantage of verified null-move (see Figure 4 of my >>>>>>article). >>>>> >>>>>Compared to R=2! But it scales inferior to R=3. So I don't expect >>>>>more time to give it an advantage compared to Heinz Adaptive Nullmove. >>>>> >>>>>>Or you might want to conduct a test to a fixed depth of 10 plies, and then >>>>>>compare the total node count and number of solved positions. >>>>> >>>>>Fixed depth tests are nonsense. I play games with a clock, not with >>>>>a fixed amount of plies. >>>>> >>>>>-- >>>>>GCP >>>> >>>> >>>>Actually they are _not_ "nonsense". They are a perfectly useful metric for >>>>comparing >>>>things. Fixed time tests are just as useful in some ways, and just as >>>>nonsensical in other >>>>ways. >>>> >>>>Fixed depth works fine unless you somehow believe that one program is doing way >>>>more >>>>work per node than the other, so that the tree sizes for a fixed depth don't >>>>compare very >>>>well. Otherwise it is perfectly ok and has been used for 25 years in testing >>>>parallel >>>>chess engines and reporting results. It avoids the problem I had in the DTS >>>>paper, of >>>>being unable to produce an exact node count (for one instance).
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.