Author: Ryan B.
Date: 12:36:54 12/25/05
Go up one level in this thread
On December 25, 2005 at 02:51:57, Thomas Gaksch wrote: >On December 24, 2005 at 23:45:10, Uri Blass wrote: > >>On December 24, 2005 at 19:08:15, Rolf Tueschen wrote: >> >>>On December 24, 2005 at 11:18:44, Uri Blass wrote: >>> >>>>On December 24, 2005 at 10:56:25, Thomas Gaksch wrote: >>>> >>>>>it is very difficult for me to describe this in english. >>>>>first of all i want to clarify, that i always thank fabien for his great program >>>>>fruit and everything he did for computer chess. an of course he is a great >>>>>programmer and a great person. no doubt about that. i have never said anything >>>>>else. i also have never said that i am the great programmer and that toga is >>>>>full of new ideas. i have also never said that fruit wouldn´t be as strong as it >>>>>is without toga. >>>>>i just read the comments from ryan, uri, tord (programmers) and others about >>>>>toga and i am very dissapointed from their statements. but i think there are >>>>>some facts. >>>>>first of all the most important fact is playing strength. >>>>>CEGT 40/40 Fruit 2.1 = 2713 ELO >>>>>CEGT 40/40 Toga II 1.1 = 2767 ELO >>>>>CEGT Blitz Fruit 2.1 = 2703 >>>>>CEGT Blitz Toga II 1.1 = 2767 >>>>>and if you critizise me that i only wrote 47 lines of code for this improvement >>>>>than i think that is not a negative point. not the quantity of lines is >>>>>important. >>>> >>>>I agree that it is not a negative point. >>>>I did not attack you in my posts and I only defended fabien against attacks that >>>>claimed that you are better programmer than him. >>>> >>>> >>>>>and if you critizise me that i invented nothing new than i can only say that it >>>>>is true. why should i invent something really new, if it is possible to improve >>>>>fruit with known techniques. >>>>>you say that it is so simple what i have done. thats partly true. but believe me >>>>>or not i invested a lot of time in testing und finding the right techniques >>>>>which improved the playing strenght so much. if everything is so simple and fast >>>>>to implement, why hasn´t done it fabien in fruit 2.0 or 2.1? i think fabien >>>>>hastn´t done it, because he didn´t believe that these techniques would increase >>>>>the playing strength so much. there is absolutely no doubt about it that it >>>>>would be easy for him to do that. and it is 100% clear that he never used toga >>>>>code in fruit. but i think i showed him the techniques which worked in fruit >>>>>very well. so he saved a little bit time in testing this things because he saw >>>>>in toga that it will work. >>>>>i also said that i would never release a clone if it is not better than his >>>>>original. so if ryan writes derogative about toga than i only can say "do it >>>>>better". but you release one beta after the other and no beta is really stronger >>>>>than toga 1.1. so you see it is not so simple to improve an existing engine like >>>>>everybody thinks. >>>> >>>>I am not sure if you are right here >>>> >>>>http://www.husvankempen.de/nunn/ratingall.html >>>> >>>>8 Toga II 1.1 2768 15 15 1319 60.8 % 2692 37.2 % >>>>12 Gambit Fruit 1.0 Beta 4bx 2743 39 39 229 51.5 % 2733 26.2 % >>>> >>>>The statistical error of Gambit Fruit 1.0 Beta 4bx is still too high to claim >>>>that it is not better than TogaII 1.1 >>>> >>>>Uri >>> >>>Good to see that you made this point, Uri. After reading Thomas one could think >>>that the difference found between Fruit and Toga of 50-60 points is a proven and >>>iron law. Also regarding the known fact that a newer version of a program is >>>_always_ "better". So - basically I cant understand this. If the author of the >>>original program would improve something the program would _also_ get better >>>results. What is the point of the whole debate? I mean the positive aspects. >>>(While we want to forget about the negative ones for good reasons.) Could you >>>give your personal view on the topic, please? >> >>The point is only that I contradict unproved statements. >>the claim of thomas was "no beta is really stronger than toga 1.1" >>It may be correct but it is still not clear. > >thats true uri. i only wanted to say, that it is not so simple to improve an >given source code by a margin of 20-30 elo. ryans starting point was toga 1.1 >with an elo of 2768. his goal should be 2790-2800 elo. i hope he will be >sucessfull now or in the future. that would be very good for the fruit project. > >> >>Previously I defended fabien against the claim that Thomas Gaksch is better >>programmer and showed that the claim that thomas had more success than fabien in >>improving fruit is wrong. > >it is good that you defend him because you are absolutely right. and i wanted to >clarify that these people do definitely not represent my opinion. > >> >>Uri My first goal with gambit fruit was not short term elo gain but long tem progress. I could get 2790-2800 elo with little work however I would rather fix what was going to limit future gain even at some short term loss. Notice that the last gambit fruit was still beta. If you do not have time to fully debug after each change it may be useful to look at what was taken out when I did Gambit Fruit. Even with some unsound chess knowledge added to make Gambit Fruit fun and lower NPS it is stronger that Toga at long time controls due to better search stability. Too bad very few ratings lists do long time controls with ponder on. Ryan
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.