Author: Bob Durrett
Date: 09:38:03 09/23/03
Go up one level in this thread
On September 23, 2003 at 01:41:24, William Penn wrote: >On September 22, 2003 at 22:00:58, Bob Durrett wrote: > >>On September 22, 2003 at 16:12:09, Sally Weltrop wrote: >> >>>On September 22, 2003 at 13:49:28, Dann Corbit wrote: >>> >>>>On September 22, 2003 at 10:14:48, Sally Weltrop wrote: >>>> >>>>>From what I have seen H9 justs mauls Shredder. A lot of X's Shredder doesn't >>>>>even see it coming until it is too late. Looks like H9 evaluation is more in >>>>>line with what u could expect in real life if a position is played properly. >>>>> >>>>>Hats off to Mark and his hard work in making a fine engine. Ohh I forgot one >>>>>other thing. I used 128 MB hash for HIARCS 9 when it's optimized for 192 so it >>>>>still did good. >>>>> >>>>>This engine WILL be #1 in the future SSDF lists. >>>> >>>>You have made your prediction of a 40/2 contest based on a handful of blitz >>>>games. >>>> >>>>It may be stronger, but your test environment cannot predict SSDF outcomes >>>>because it is grotesquely different. >>> >>>Not that much different. if HIARCS8 played Shredder 7.04 to a standstill and >>>this ver (9) is doing a number on it ... I think it's logical to say it'll b >>>much better. Is HIARCS9 going to do worse with more time? probably Shredder will >>>get stronger, maybe HIARCS9 will too. Just wait & see >>>> >>>>IMO-YMMV >> >>I am curious to know whether or not HIARCS jumps quickly to a "high" ply depth >>the way DJ seems to do [on my single processor machine.] Perhaps it uses some >>innovative/novel pruning in the beginning? >> >>Perhaps a chess engine which does very well at extremely fast games might do so >>because it's programmer made it immediately cull out "obviously weak" moves in >>the initial searching. Some of the moves culled out may be considered later if >>more time is allowed? Perhaps the HIARCS programmer could answer this. >> >>The fact that HIARCS does well at blitz is not insignificant. Many internet >>games are blitz games. However . . . >> >>My use for chess engines is analysis, where each move is considered for several >>or many minutes. If anybody has run such very slow games/analysis with HIARCS, >>I would like to know the results. If HIARCS becomes one of the top engines for >>slow chess, I will run to the store and purchase it post haste and start doing >>comparisons [during game analysis] of it with the other top engines. >> >>This re-opens "Pandora's Box" regarding the issue of whether or not any chess >>engines are deliberately programmed to play differently in fast games when >>compared to the way they are programmed to play in slow games. In other words, >>the search [& other] algorithms used might be different for the fast and slow >>cases. Position evaluation might use a simplified algorithm for speed chess >>simply to save time. I have not seen a definitive answer to this question >>although I recall Hyatt saying something about this sort of thing for Crafty. >>If algorithm choices are built into the engine software as a function of the >>speed setting, then the results in speed chess may be a poor indicator of >>performance in slow chess. >> >>Bob D. > >One of the best indicators of strength in "slow" play or analysis is >correspondence chess (CC). I used Chessmaster for a long time. It was very >competitive through version 6000. Then I switched to the various Fritz chess >engines for several years with excellent results, usually winning my section in >serious master level national/international CC competition. Then I just recently >purchased Shredder 7 and have been giving the 7.04 engine a good tryout, but am >only getting draws with it so far in master level CC play. It seems dull. I'm >not seeing the flashes of tactical brilliance (quick wins) Fritz engines can >produce. Perhaps part of the problem is lack of documentation. Wouldn't the >"Normal" style be better than the "Aggressive" style setting for the Shredder >7.04 engine? I would think so, just based on those word connotations, but who >really knows? We're lucky just to know an acceptable word for a translation >between languages; connotations are usually lost. >WP My tentative observation about Shredder 7.0.4 is that all moves beyond the first are garbage. Fritz, on the other hand, often gives several good moves in a line. When I get a line from Shredder, I immediately truncate all but the first move in the line, knowing that the rest of the moves are untrustworthy. May I suggest: Don't use a chess engine to find only the best move. Also have it find the best several and then check out the second or third best, because deeper analysis may show that they are really best. Someone else here suggested having several engines running simultaneously on your computer evaluating the same position. I have not tried that because the hash table sizes must be relatively small. I run five engines but do it sequentially, using large hash tables. If one engine doesn't find the killer move you're looking for, then maybe another engine will. There is something inherently very dissatisfying about letting a chess engine find your moves for you in real time. Doing so reduces you to the role of a secretary, a data entry person, or a clerk. All you are doing in that case is watching your chess engine(s) play a game. Maybe that would be fun for computer chess enthusiasts, but it's not playing chess yourself. It is apparently true that a large percentage of the correspondence chess players do use chess engines to help them find good moves. This is a fact of life, and one must "deal with it." It would be just plain stupid to put yourself at a competitive disadvantage. On the other hand, there is a compromise which may be satisfying. That is to analyze the position and find the moves yourself. Only after having done that would you use the chess engine to verify that you were not about to play an exceptionally stupid move. I no longer play correspondence chess due to old age and general senility. However, it's a great way to use one's leisure time and computers can make it more interesting. IMHO. : ) Bob D.
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.