ChessUSA.com TalkChess.com
Hosted by Your Move Chess & Games
 
 FAQFAQ   SearchSearch   MemberlistMemberlist   UsergroupsUsergroups   RegisterRegister 
 ProfileProfile   Log in to check your private messagesLog in to check your private messages   Log inLog in 

tuning via maximizing likelihood
Post new topic    TalkChess.com Forum Index -> Computer Chess Club: Programming and Technical Discussions Flat
View previous topic :: View next topic  
Author Message
Michel Van den Bergh



Joined: 28 Sep 2008
Posts: 1906

PostPost subject: Re: tuning via maximizing likelihood    Posted: Thu Oct 12, 2017 7:20 am Reply to topic Reply with quote

jdart wrote:
Quote:
r * log( logistic(score) ) + (1 - r) log( 1 - logistic(score) )


This may have been covered, at least indirectly, but the issue I see with this function and draws is that the value of the loss is not zero when the predicted score is 0.5 and the actual score is 0.5. The loss function for draws does have its minimum value in that case. But it is not zero.

Each draw will increase the value of the objective even if it is predicted correctly, and this increase is more than what most loss and win positions contribute (because many will be more or less accurately predicted).

--Jon


This was the objective function discussed in the OP but it is not really how maximum likelihood is supposed to work in the presence of draws. One needs to choose a draw model. Luckily there are reasonable draw models available which are used by rating tools. This is discussed in subsequent posts.

A draw model involves at least one extra parameter which is a proxy for the draw ratio (draw_elo in the Bayes Elo model) but perhaps it may involve more parameters.

Whether or not the text book approach yields better results in actual games than more adhoc approaches is of course impossible to know in advance.
_________________
Ideas=science. Simplification=engineering.
Without ideas there is nothing to simplify.
Back to top
View user's profile Send private message
Display posts from previous:   
Subject Author Date/Time
tuning via maximizing likelihood Daniel Shawul Wed Oct 04, 2017 4:54 pm
      Re: tuning via maximizing likelihood Álvaro Begué Wed Oct 04, 2017 5:38 pm
            Re: tuning via maximizing likelihood Daniel Shawul Wed Oct 04, 2017 5:45 pm
                  Re: tuning via maximizing likelihood Álvaro Begué Wed Oct 04, 2017 5:51 pm
                        Re: tuning via maximizing likelihood Daniel Shawul Wed Oct 04, 2017 6:00 pm
                              Re: tuning via maximizing likelihood Álvaro Begué Wed Oct 04, 2017 6:20 pm
                                    Re: tuning via maximizing likelihood Daniel Shawul Wed Oct 04, 2017 8:15 pm
                                          Re: tuning via maximizing likelihood Álvaro Begué Wed Oct 04, 2017 11:18 pm
                                    Re: tuning via maximizing likelihood Rein Halbersma Thu Oct 05, 2017 7:52 am
                                          Re: tuning via maximizing likelihood Michel Van den Bergh Thu Oct 05, 2017 9:13 am
                                                Re: tuning via maximizing likelihood Michel Van den Bergh Thu Oct 05, 2017 10:20 am
                                                      Re: tuning via maximizing likelihood Rein Halbersma Thu Oct 05, 2017 11:07 am
                                                            Re: tuning via maximizing likelihood Michel Van den Bergh Thu Oct 05, 2017 11:25 am
                                                                  Re: tuning via maximizing likelihood Michel Van den Bergh Thu Oct 05, 2017 11:45 am
                                          Re: tuning via maximizing likelihood Daniel Shawul Thu Oct 05, 2017 3:26 pm
                                                Re: tuning via maximizing likelihood Álvaro Begué Thu Oct 05, 2017 3:33 pm
                        Re: tuning via maximizing likelihood Rein Halbersma Thu Oct 05, 2017 7:22 am
                  Re: tuning via maximizing likelihood Jon Dart Wed Oct 11, 2017 11:43 pm
                        Re: tuning via maximizing likelihood Rein Halbersma Thu Oct 12, 2017 6:59 am
                        Re: tuning via maximizing likelihood Michel Van den Bergh Thu Oct 12, 2017 7:20 am
                        Re: tuning via maximizing likelihood Daniel Shawul Thu Oct 12, 2017 4:56 pm
                              Re: tuning via maximizing likelihood Jon Dart Thu Oct 12, 2017 7:55 pm
                                    Re: tuning via maximizing likelihood Daniel Shawul Thu Oct 12, 2017 9:12 pm
            Re: tuning via maximizing likelihood Rein Halbersma Thu Oct 05, 2017 7:15 am
                  Re: tuning via maximizing likelihood Michel Van den Bergh Thu Oct 05, 2017 2:45 pm
                        Re: tuning via maximizing likelihood Álvaro Begué Thu Oct 05, 2017 2:53 pm
                              Re: tuning via maximizing likelihood Michel Van den Bergh Thu Oct 05, 2017 3:40 pm
                                    Re: tuning via maximizing likelihood Álvaro Begué Thu Oct 05, 2017 3:44 pm
                                          Re: tuning via maximizing likelihood Rein Halbersma Thu Oct 05, 2017 8:36 pm
                                                Re: tuning via maximizing likelihood Michel Van den Bergh Fri Oct 06, 2017 6:57 am
                                                      Re: tuning via maximizing likelihood Michel Van den Bergh Fri Oct 06, 2017 8:19 am
                                                            Re: tuning via maximizing likelihood Michel Van den Bergh Fri Oct 06, 2017 8:42 am
                                                                  Re: tuning via maximizing likelihood Michel Van den Bergh Fri Oct 06, 2017 9:36 am
                                                                        Re: tuning via maximizing likelihood Michel Van den Bergh Fri Oct 06, 2017 1:44 pm
                                                                              Re: tuning via maximizing likelihood Daniel Shawul Fri Oct 06, 2017 5:37 pm
                                                                        Re: tuning via maximizing likelihood Rein Halbersma Sat Oct 07, 2017 12:35 pm
                                                            Re: tuning via maximizing likelihood Rein Halbersma Sat Oct 07, 2017 12:15 pm
                                                            Re: tuning via maximizing likelihood Rein Halbersma Sun Oct 08, 2017 7:36 pm
                                                      Re: tuning via maximizing likelihood Rein Halbersma Sat Oct 07, 2017 12:22 pm
                                                      Re: tuning via maximizing likelihood Rein Halbersma Sat Oct 07, 2017 12:41 pm
                                                            Re: tuning via maximizing likelihood Jon Dart Sat Oct 07, 2017 2:34 pm
      Re: tuning via maximizing likelihood Jon Dart Wed Oct 04, 2017 10:45 pm
            Re: tuning via maximizing likelihood Daniel Shawul Thu Oct 05, 2017 12:47 am
                  Re: tuning via maximizing likelihood Jon Dart Thu Oct 05, 2017 1:46 am
                  Re: tuning via maximizing likelihood Rein Halbersma Thu Oct 05, 2017 11:14 am
                  Re: tuning via maximizing likelihood Daniel Shawul Fri Oct 06, 2017 5:23 pm
                        Re: tuning via maximizing likelihood Michel Van den Bergh Fri Oct 06, 2017 7:16 pm
                              Re: tuning via maximizing likelihood Daniel Shawul Fri Oct 06, 2017 7:36 pm
                                    Re: tuning via maximizing likelihood Michel Van den Bergh Sat Oct 07, 2017 3:26 am
                                          Re: tuning via maximizing likelihood Daniel Shawul Sat Oct 07, 2017 2:07 pm
                                                Re: tuning via maximizing likelihood Michel Van den Bergh Tue Oct 10, 2017 6:43 am
                              Re: tuning via maximizing likelihood Evert Glebbeek Fri Oct 06, 2017 10:32 pm
                        Re: tuning via maximizing likelihood Daniel Shawul Sat Oct 07, 2017 3:21 pm
                              maximum a posteriori Daniel Shawul Sat Oct 07, 2017 3:34 pm
                                    Re: maximum a posteriori Michel Van den Bergh Wed Oct 11, 2017 9:41 am
                                          Re: maximum a posteriori Daniel Shawul Thu Oct 12, 2017 5:04 pm
                                                Re: maximum a posteriori Michel Van den Bergh Fri Oct 13, 2017 8:04 am
                                                      Re: maximum a posteriori Daniel Shawul Fri Oct 13, 2017 6:40 pm
                                                            Re: maximum a posteriori Daniel Shawul Fri Oct 13, 2017 7:46 pm
                                                                  Re: maximum a posteriori Michel Van den Bergh Sat Oct 14, 2017 5:44 am
                                                                        Re: maximum a posteriori Daniel Shawul Sat Oct 14, 2017 4:11 pm
                                                                              Re: maximum a posteriori Evert Glebbeek Sat Oct 14, 2017 4:56 pm
                                                                              Re: maximum a posteriori Daniel Shawul Sat Oct 14, 2017 9:01 pm
                                                                              Re: maximum a posteriori Michel Van den Bergh Tue Oct 17, 2017 6:19 am
                                                                                    Re: maximum a posteriori Álvaro Begué Tue Oct 17, 2017 11:25 am
                                                                                    Re: maximum a posteriori Daniel Shawul Tue Oct 17, 2017 5:07 pm
                                                                                          Re: maximum a posteriori Daniel Shawul Tue Oct 17, 2017 7:01 pm
                                                                              Re: maximum a posteriori Daniel Shawul Tue Oct 17, 2017 5:43 pm
                                                                                    Re: maximum a posteriori Álvaro Begué Tue Oct 17, 2017 6:42 pm
                                                                                          Re: maximum a posteriori Daniel Shawul Tue Oct 17, 2017 7:18 pm
                                                                                          Re: maximum a posteriori Álvaro Begué Tue Oct 17, 2017 7:37 pm
                                                                                          Re: maximum a posteriori Daniel Shawul Tue Oct 17, 2017 7:54 pm
                                                                                          Re: maximum a posteriori Álvaro Begué Tue Oct 17, 2017 9:06 pm
                                                                                          Re: maximum a posteriori Daniel Shawul Tue Oct 17, 2017 9:30 pm
                                                                                          Re: maximum a posteriori Álvaro Begué Tue Oct 17, 2017 11:15 pm
                                                                                    Re: maximum a posteriori Daniel Shawul Wed Oct 18, 2017 4:11 pm
                                                                                    Re: maximum a posteriori Evert Glebbeek Wed Oct 18, 2017 8:33 pm
                                                                                          Re: maximum a posteriori Daniel Shawul Wed Oct 18, 2017 9:14 pm
Post new topic    TalkChess.com Forum Index -> Computer Chess Club: Programming and Technical Discussions

 
Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum




Powered by phpBB © 2001, 2005 phpBB Group
Enhanced with Moby Threads