Author: Robert Hyatt
Date: 18:38:06 06/25/98
Go up one level in this thread
On June 25, 1998 at 20:40:00, Vincent Diepeveen wrote: > >On June 25, 1998 at 13:31:30, Bruce Moreland wrote: > >> >>On June 25, 1998 at 12:15:09, Robert Hyatt wrote: >> >>>I pointed out to Vincent last week here that he was not doing well playing >>>against Crafty. He pointed out (factually, by the way) that Crafty has a >>>hardware advantage with the 4 processor ALR. But he's going to have to take >>>a stand on one side of the fence or the other. Either speed is important, and >>>his note about me having faster hardware does explain why crafty wins most of >>>the games played on ICC, or speed is not important and doesn't explain crafty's >>>win/lose ratio vs the various DiepX programs on ICC. >> >>This is a fascinating point. > >>Some people might argue that speed doesn't matter, but I've yet to see one of >>them show up at a WMCCC on a crusty old 486 because they don't think it's worth >>the bother of copying their software to a faster machine. > >>I'll take faster hardware any time, on ICC it's like a shot of adrenaline. I >>get the idea that it matters less as the time controls slow down, but my >>experience is based upon pretty small hardware advantages. > >>bruce > >Relax, hardware doesn't matter that much anymore when getting 11/12 ply. >At icc however diep searches between 6 ply and 8 ply, then >hardware simply kills. Especially for a program which hasn't been adjusted >to blitz (meaning that at low depths its preferences are a little odd sometimes, >as parameters aren't adjusted at all). > >6 ply simply means 3 moves. Then there is this selectivity thing of mine >which takes care that at 6 ply it still misses some simple things in order >to get that huge depth. > your analysis fails. What about the many 60 0 and 30 30 games we have played? those are *not* blitz speeds. And it doesn't change a thing in the outcome, although I can do a search and post the results if anyone wants the real numbers for inspection. >So if you give Diep's selective 12 ply search in opening/middlegame >Bob, or bruce, then i give you 14 ply fullwidth for free. S.E. you get for >free too. give a frog a pocket, he carries a gun, and never worries about snakes again, either. But he doesn't have the pocket, and he lives in a swamp, and he ends up "soup du jour" regularly... > >Most human plans are btw around 6 moves, so this also indicates that >with 12 ply the search is not the problem anymore. Things like evaluation >become way way more important at those depths. > >Of course there ARE things you miss with a 12 ply search which you DO >find with 14 ply search. No doubt that 14 ply is more relaxed than 12, >but it is so rare that a deeper search brings new tactics. > >If you just play blitz like is happening on the internet, then i can imagine >why some people developed 'insights' that just searching deeper is the >way to go, especially when combining this with a program that doesn't >do tactics in q-search (like not doing checks nor evaluating pins). > >Even at 4 processors crafty gets at blitz 'just' 8/9 ply. > >That's just 5 moves. So matethreads are nearly always missed (as nullmove >last x plies gives cutoff if you don't do check, so you actually only see >6 plies of threads, and 6/7 ply for pins). I don't see enough mate threats overlooked to worry about. When I lose more than one game out of 10 to a human, I start looking, but it generally is not an overlooked mate threat, it is more likely a positional mistake. > >This probably leads to those insights which are so widely written down in >CCC and RGCC by the author. > >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.