Search found 19595 matches

by hgm
Sun Aug 10, 2008 4:41 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: New testing thread
Replies: 243
Views: 45344

Re: BTW

bob wrote:You made a duplicate post. Are you having issues with very slow posting times? I've been trying to determine if it is just CCC or something in general on my end as I am seeing the same.
CCC is very slow today, also for me.
by hgm
Sun Aug 10, 2008 4:36 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: New testing thread
Replies: 243
Views: 45344

Re: 4 sets of data

However, would it be possible for either (a) you _do_ follow a specific discussion and post comments related to it or (b) if you choose to not follow context, then also choose to not make comments that have nothing to do with the discussion? You would have to be more specific. That you perhaps cann...
by hgm
Sun Aug 10, 2008 2:52 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: New testing thread
Replies: 243
Views: 45344

Re: Correlated data discussion

Another one: do not draw conclusions from _one_ partial result. The test has not even finished yet and here you go, drawing a conclusion. The way the test is run, the partial results can be highly misleading due to white/black bias. One does best by waiting until _all_ the data is in. You think tha...
by hgm
Sun Aug 10, 2008 2:52 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: New testing thread
Replies: 243
Views: 45344

Re: Correlated data discussion

Another one: do not draw conclusions from _one_ partial result. The test has not even finished yet and here you go, drawing a conclusion. The way the test is run, the partial results can be highly misleading due to white/black bias. One does best by waiting until _all_ the data is in. You think tha...
by hgm
Sun Aug 10, 2008 1:50 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Converting Word files to Windows help files
Replies: 2
Views: 1073

Re: Converting Word files to Windows help files

OK, thanks! This help compiler really works, and I was even able to fix the RTF source such that I only get 3 warnings. (They are all about too-long font names, which I did not specify at all, so I suppose that this is not my fault.)

Next WinBoard release will be with a complete up-to-date help!
by hgm
Sun Aug 10, 2008 10:55 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Testing: Optimizing for testing set
Replies: 2
Views: 887

Re: Testing: Optimizing for testing set

Yes, we are optimizing for the set. And no, that does not necessarily improve the engine if the set is too small or unrepresentative. Now the situation is not that bleak, as many improvements that give a better result on one position also will give better results on other positions, just because the...
by hgm
Sun Aug 10, 2008 9:44 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: New testing thread
Replies: 243
Views: 45344

Re: Intermediate results...

So far, just over 18,000 games have finished. I started this somewhere around 8pm or so, so another 8-10 hours will see the thing finished (one run of four). the interesting part will be the second, third and fourth runs to see if the results suddenly become stable enough to use. More as the things...
by hgm
Sun Aug 10, 2008 8:33 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: New testing thread
Replies: 243
Views: 45344

Re: Correlated data discussion

What we apparently disagree is about the potential that we will be able to re-use positions in a way that is independent. I highly doubt it can be done. Is this just a gut feeling, or is this doubt somehow based on mathematical calculation? The variances that would result from the different samplin...
by hgm
Sun Aug 10, 2008 7:37 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: New testing thread
Replies: 243
Views: 45344

Re: 4 sets of data

No, but if you'd just follow the discussion once in a while, you would see that reducing the number of games is good enough to test the hypothesis that a round-robin will stabilize the ratings more. Well, as I pointed out, even two games would be enough to 'test' the hypothesis that 1+1=2. Test wha...
by hgm
Sun Aug 10, 2008 7:07 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: New testing thread
Replies: 243
Views: 45344

Re: Correlated data discussion

Using the node count instead of time would mean you're not testing the time management code, right? I guess you'd then have to test that separately, which is really starting to make the testing complex. This would not be true if you use the node-based time-control mode of WinBoard 4.3.14. Then the ...