Author: Uri Blass
Date: 09:01:00 08/17/05
Go up one level in this thread
On August 17, 2005 at 09:52:05, Vasik Rajlich wrote: >On August 17, 2005 at 05:36:21, Uri Blass wrote: > >>On August 17, 2005 at 05:07:05, Vasik Rajlich wrote: >> >>>On August 16, 2005 at 17:26:28, Uri Blass wrote: >>> >>>>On August 16, 2005 at 17:16:33, Paolo Casaschi wrote: >>>> >>>>>>I know that usually when program improve they improve in all time controls. >>>>>>I do not know of evaluation changes or search changes that make programs weaker >>>>>>at blitz but stronger at long time control. >>>>>> >>>>>>In thoery it can happen but I need to see a proof for it and I believe that >>>>>>fabien mainly test in blitz time control(he can correct me if I am wrong) >>>>>>because usually productive changes in blitz of adding knowledge to the >>>>>>evaluation are also productive at long time control. >>>>> >>>>>Do you have any proof or evidence that there is some correlation between blitz >>>>>strenght and slower speed strenght? >>>>>If you dont, then we can only compare assumptions and I tend to agree with Bob >>>>>Hyatt since the same non-correlation is evident with humans and because common >>>>>sense... >>>>> >>>>>--Paolo >>>> >>>>Of course there is correlation. >>>> >>>>Look at every rating list at long time control and you can see Shredder,Fruit >>>>Fritz,Junior,Hiarcs at the top of the list. >>>> >>>>Now look at rating list at blitz. >>>>What do you see? >>>> >>>>Surprise for you >>>>Again the same programs. >>>> >>>>You do not believe it? >>>>Here are 2 rating lists one for blitz and another one for longer time control. >>>> >>>> >>>>http://www.husvankempen.de/nunn/rangliste.html >>>>http://www.miko42.de/turniere/blitzturniere/blitzrangliste.html >>>> >>>> >>>>Uri >>> >>>1 minute/move on one processor is a sort of blitz, compared to 2 or 3 >>>minutes/move on 8 processors. >>> >>>It makes sense to me that highly selective search works better at shorter time >>>controls. It also seems that we have some data which starts to confirm this. >>> >>>For example, in this WCC, you've got three programs (Crafty, Zappa and Fruit) >>>who use (apparently) a plain search, and two who are (apparently) much more >>>selective (Shredder and Junior). Look who is overperforming and underperforming, >>>compared to rating lists from faster time controls. >>> >>>Another issue is the books. Shredder for example has a "big" book, which is >>>maybe good for blitz & SSDF but awful for this sort of time control. >>> >>>It seems that some of the work of the professionals is going to be wasted as >>>hardware moves forward. >>> >>>Vas >> >>I do not know why do you think that fruit does not use selective search. >>I think that history based pruning is type of selection. >> >>I also do not know what type of search use programs without source code and I >>have no idea why do you think that zappa use plain search when Junior and >>Shredder use selective search. >> >>You may suspect that it is the case for shredder because it searches deeper and >>sometimes has problems to solve relatively simple positions but I do not know >>what is the basis for this opinion about Junior when Junior does not have plies >>in the same meaning of other programs and the fact that Junior is fast searcher >>means that it does not waste much time in selecting moves. >> >>The problem of Junior in the only game that it lost(against Crafty) was not >>selective search but bad evaluation >> >>Uri > >Fruit is relatively non-selective. Basically it does null move and a very safe >form of fail-low reducing backed up by history table. Minimal extensions >-recaptures only along PV, no passed pawns, nothing fractional. No futility >pruning. Even no transposition table pruning along PV. If it is non selective it cannot outsearch the opponents but I think that it usually outsearch everything except shredder(Junior has different definition to ply so I do not count it) > >As for the non-open source engines, I think it's pretty certain based on various >author comments. Amir has discussed Junior many times in the CCC archives - they >have some sort of a scheme of costing 2 plies for unpromising moves, and tons of >extensions. Yes but I am not sure if it is right to call it selective based on it. It searchs unpromosing moves to reduced depth and fruit does the same by history based pruning. 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.