Author: Robin Smith
Date: 13:36:30 06/10/03
Go up one level in this thread
Hi Mark, Thank you for responding to my post. > i would say that onw of your seven points is relevant to a discussion of >analysis functions- specifically hpw the engine works on EPD files. All seven points I made are directly relevant to chess analysis. I listed them in my personal order of importance for analysis, so I think #'s 1-4 are all at least as important analysis features as EPD file analysis (which is also very important). If you think only EPD is relevant to analysis then you don't appreciate various other possible (and powerful) analysis method possibilities. >most of your ideas center on how to conduct engine- engine matches on single >chip pcs. Right. Engine matches can be a VERY powerful analysis method. Have you ever seen a position which all chess engine evaluations think favors Black, but engine tournament results strongly (and consistently) favor White? I have. >although to run the gaunlet without changing colors could be viewed as a form of >stochastic analysis. assuming the engines learn, though, there will be alot of >randomness in the results. ( because the best attempt still leads to an desrable >outcome for one side, a poorer variation might be elected, as an example) Once you run an engine-engine match, looking at the data stochastically is only one method, and probably one of the most error prone due to the randomness factor you mention. If you understand chess you can also look at the games themselves. Looking at these games helps the human analyst learn about the position. Humans are still better than computers at learning! There are a number of other things you can do with engine matches as well. But the present Fritz GUI limitation of not being able to start from arbitrary positions makes this very awkward. >do you feel that there are chessengines which are suitable for stochastic and >NOT positional analysis? if so, what are you using, crafty? tiger? cap data? The best engines for both stochastic as well as positional analysis are usually the same (strongest you can get) engines. I mostly use Chessmaster, Fritz, Hiarcs, Junior, Shredder, Tiger, and Yace. On rare occaisions I might also use another enigine. I don't use cap data at all. Cap data would be more usefull if it did not mix results from multiple engines into the same database, and short of that at the very least if it recorded which engine did the analysis, for how long, and on what hardware for each position. Mixing multiple versions of multiple engines into one database makes the cap data as it stands only good for looking for blunders. Cap is a great concept with a flawed implimentation and unfortuanately for more than blunderchecking I think cap is nearly worthless. Best regards, Robin > >best regards, marc > > > >On June 09, 2003 at 20:08:13, Robin Smith wrote: > >>On June 09, 2003 at 14:40:36, margolies,marc wrote: >> >>>stu, >>>if you want to change the analysis function, might you please give us some ideas >>>about how you use it now. what would make it better and not merely different. >>>have you tried training/tutorial modules by any company, do you like them? do >>>you use the coach feature in cb gui and how is this different than smart talking >>>sounds? would a speach-to-text feature in the coach be what you really want? >>>cheers, marc >> >>I have several ideas about how to improve analysis features: >> >>1) Allow arbitrarily long opening books and/or tournaments from arbitrary >>starting positions (tournament style shootouts, not just engine vs itself >>shootouts). >>2) Allow "run the gauntlet" shootouts where the side running the guantlet always >>gets the same color instead of alternating. >>3) Allow more sophisticated game searches >> a) for positions with a certain piece configuration regardless of where it >>occurs on the board and >> 2) material imbalances ala Chess Assistant >>4) Fix the winboard adapter so it doesn't mess up winboard programs. >>5) Allow import/export analysis of EPD files >>6) Allow multiple instances of the same engine to kibitz simultaneously (locked >>on different positions). >>7) Allow multiple instances of the same engine to play in tournaments, each with >>different engine parameters. >> >>Robin
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.