Author: Uri Blass
Date: 06:33:17 11/02/04
Go up one level in this thread
On November 02, 2004 at 07:03:49, José Carlos wrote: >On November 02, 2004 at 04:54:08, Uri Blass wrote: > >>I am considering to release a new free version of movei. >> >>The latest version has a lot of options to change personality and my guess is >>that it is possible to get at least 50 elo only by changing evaluation and >>search parameters(It is probably possible to get more from movei than from >>programs when changing personalities were already tested more heavily like >>prodeo). >> >>The question is if to release new free version and hope that people will give me >>a personality that does significantly better or to give the version only to few >>beta testers that will need to test different personalities(if there are people >>who are interested). >> >>I am not interested in tournaments that include only one version of movei but >>games at different time control with different personalities from blitz to long >>time control may be productive for me. >> >>My question is how many testers are interested in testing different >>personalities in one of the following cases: >> >>1)If I send it only to beta testers. >>2)If I release it as a free version. >> >>please tell me if you are interested to test it in case 1 and in case 2 and I >>will decide if to release a free version or only give beta version to people who >>ask me based on the replies. >> >>Uri > > Averno has personalities option for some years now. So far, nobody has sent me >a new personality. AFAIK, no one has even tried changing one parameter. So for >last version, I decided to remove most personality options and hardcoded what I >considered best. > Thus, I suggest you to look for a beta tester. > > José C. Some points to consider: 1)it is possible that people simply did not know about the personalities options in averno. 2)It is possible that they decided not to test because they did not believe that they can get a big improvement from it. In the case of movei I start by telling everybody about the option to change personality and I also guess that testers may help me to get more than 50 elo improvement from changing personalities in movei. Of course I cannot be sure about it but I have reasons to believe in it and let me explain: There are parameters that I am almost sure that they are bad(for example the latest version does not increase the evaluation of passed pawns in the endgame when there is a possibility to do it). I simply found by games that not increasing the value of passed pawns in the endgame gave better results but it does not mean that the idea is bad. There are many other possibilities: a)it is possible that it is better to reduce the value of passed pawns in the opening and the middle game and not to increase the value of passed pawns in the endgame. b)It is also possible that bigger positional values cause some instability in the search and testers can prevent it by changing search parameters. c)It is also possible that I increased the value too much in the endgame but I do not think that the last option is the case because Fruit clearly increase it more than movei. d)It is possible that other parameters of evaluating pawns relative to the king need to be changed. passed pawns is only one example and movei also does not evaluate knight outposts by default but it does not mean that it is best to do it. 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.