Author: Vincent Diepeveen
Date: 15:31:55 07/19/99
Go up one level in this thread
On July 19, 1999 at 17:31:25, pete wrote: >On July 19, 1999 at 10:05:27, Vincent Diepeveen wrote: > >>On July 18, 1999 at 22:52:23, K1 wrote: >> >>>I have started such a site for winboard programs, strictly for fun. URL is at : >>>http://www.geocities.com/Colosseum/Base/5978/ >>>Regards. >> >>Those old times: if we all would be still playing 40 moves in 5 minutes >>at a P75 with 32 mb RAM, so actual speed is >>a P37.5 >> >>Let's calculate a little: >> >>a P133 is exactly 3 times slower for DIEP as a Pentium pro200. >>a PII450 is a little bit more than 2 times faster than a PRO200. >> >>So a PII450 is about 4 x 3 x (133/37.5) times faster = 45 times faster. >> >>Let's see: 40 moves in 300 seconds = 7.5 seconds a move. >>That's 7.5 / 45 = 0.167 seconds a move at a PII450. >> >>Gotta be a 1 ply search a move tournament... >>...and winner is a prog searching 2 ply i bet? >>...hopefully you have good old genius somewhere, cuz >>it loves those tournaments... >> >>I still didn't take into account >>the small hashsize which caused huge >>swapping playing the engines against each other. >> >>-------------------------------------------------------------- >>distinct winboard programmes + 3 older versions (#30 to #32, to make total >>number of competitors = 32) >> >>Knock out system (alphabetical order for #1 to #29! Drawn game replayed until >>winner emerged) >>Time control : 40 moves/5mins >>Hardware/OS : Toshiba Libretto, Pentium 75mHz, 32mb ram, Windows 95 >>All games played using Tim Mann’s Winboard ver 4.0.2 > >I also have one of those slow computers a P133 ; 16 MB RAM; and I sometimes like >to watch games between two computers ( similar to watching TV ) > >And as the other one is P233MMX or PII300 I know only one prog that can keep up >with the ones on the faster computer; and this _is_ genius . >I noticed you seem to hate genius ; how come ? And why do people think it is >dumb ? The moves it makes are usually sensible ; not always the best ones but >never real bad ones , which is not always true of the best progs now. Right you already indicate it. I don't hate richard lang. In contradiction. I have big respect for what Richard has done: being for many years the unbeaten number one, defending year after year his titles with success. What i like about genius is that at a slow computer it can kick the hell out of faster hardware. However, this is because it's not only a brilliant program, but rather because it's always producing a move from a certain level. Really brilliant things we can't expect from genius, as the knowledge is simply not in it. Every move is of level x. What i find very weird is next position: white Kh1 black Kh8 Bf6,h4,h5,h6,h7 dead drawn of course. Of course hardly anyprogram sees that. That's not the thing what i find interesting. If you change the selective search depth, then restart genius, then it produces different scores. *completely* different scores. I would love to know how the selective search of genius works, but i have a feeling it has to do with alfa and beta bigtime! >It is simple truth that on very slow hardware genius is still one of the if not >_the_ absolute best ; and it seems to be also true that genius profits much les >than other progs from faster hardware . >Now can you explain why everyone thinks genius is dumb ? from a naive point of >view one could also state the complete opposite. In contradiction, the proof you have for that it is dumb you explain as it being smart! >and there is one field where genius still seems to rule ; despite of all the >tablebase stuff ; and this is complex endings; again something not sounding dumb >:) Weird. Even DIEP's bad endgame is enuf to crush genius in the endgame. Genius endgame is outdated in endgame i guess (like my DIEP, but i'm working on its endgame right now!). >now everyone also seems to agree that genius is basically the same since genius3 >; and this is reproducable as genius3 scores the same as genius 6 ; so , maybe >it is just a prog with no serious development in the last years ? It has become tactical stronger last few years. Genius was some years ago tactical one of the strongest programs, never missing a trick where it could fall for. That's why it's still so strong in blitz and at slower hardware it simply outperforms anything. Genius is having a little bit more pawnstructure code than fritz, but it's all very basical knowledge. Very simple knowledge. If you debug gnuchess, take some bugs out of its evaluation (some gross errors there, i heart from others in search they exist too). speed up gnuchess a lot, and it should already kick ass of genius, as it's having way more knowledge. So genius strength is mainly derived from tactics and never making a real bad move. Against well tested programs having more knowledge and getting above 10 ply it doesn't make a chance nowadays. It definitely will not in the future too. It's branching factor is from the old days that progs used to search fullwidth. Kind of natural induction! >This is a sensible question and I hope you won't feel offended. >Pete Questions are always ok. I'm very objective in that. I can't garantuee an answer in public always though. If i miss any question, then email it to me: diep@xs4all.nl I checked last few days CCC a little, but after wasting a lot of time at writing stuff and the slow internet and meeting Dave here and, who's simply not sensible to objective arguments why MTD sucks bigtime for nowadays chessprograms, i feel i'm gone real soon again for a while. At CCC except for some posts from Bob i hardly see any interesting post here. Bunch of beginners posting, and the pro's keeping their mouth shut. If i would have the guts to quote what some said about MTD, i mean *all of them*. I spoke with most dudes about MTD in past, and no one of all those dudes is writing about its experiences with it! Idem for FHR, probcut etc. Greetings, Vincent
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.