Author: Uri Blass
Date: 20:51:40 08/25/01
Go up one level in this thread
On August 25, 2001 at 22:38:02, Vincent Diepeveen wrote: >On August 24, 2001 at 10:56:16, Uri Blass wrote: > >>On August 24, 2001 at 10:50:25, Jeroen Noomen wrote: >> >>>On August 24, 2001 at 07:06:51, Uri Blass wrote: >>> >>>Hi Uri, >>> >>>IMO these statistics make no sense. I rather would like to see >>>Tiger being the first single program than having a better Elo >>>performance. >>> >>>Jeroen >> >>I understand but I think that it is still better to have a better prformance >>than having nothing. > >Your rating calculation sucks everywhere Uri, > >please put diep at 3500 rating and redo your calculation and you'll >see that Shredder has a higher TPR than tiger, simply because i played >shredder and tiger didn't. The calculation assumed that the tournament happens again and again and not assumed a known rating for every program. There is no known rating for every program but the result is logical and the result of Diep is typical result of Diep in tournaments. It is not the worst program but it is clearly weaker than the commercial programs and it could not win against spiddergirl in the last round. > >your TPR is simply too much dependant upon the rating you give other >engines, that's the whole problem! I give initial rating of 2300 to all engines and I calculate the rating based on the assumption that the results of the tournament happens again and again. > >Also take into account that there is a limited number of participants. >You don't pick opponents yourself. You GET them. > >If you draw the first round you sure don't get an easier schedule than >when you draw the last round, but you sure get a smaller TPR according >to your calculations. > >I don't see why diep would be weaker than crafty or easier than crafty >to play against in a tournament, but crafty definitely scored more >points than diep and crafty definitely is scaled higher, despite that >last 4 tournaments diep scored 3.5 out of 4 against crafty. Diep is clearly weaker than Crafty. looking only at results between Crafty and Diep without looking at results of both programs against other program is not the way to compare. Crafty did some results that are better than Diep results in the last tournament. Crafty beated Rebel when Diep lost against Rebel Crafty drew against Ferret when Diep lost against Ferret The only better result of Diep against the same opponents in the last tournament is the draw against Gromit. The results from previous tournaments are not important. I know that in the last WMCCC Crafty did a poor opening preperation(something that did not happen this time) Uri > >The whole rating issue sucks on the right and left. the only way >is to calculate it like in a round-robin tournament and you know that too! > >However that wouldn't serve your plans as in a round-robin tournament >the average rating is defined as being the average over *all* participants, >including your own rating (or whatever). > >Meaning that someone with a higher score is simply someone with a higher >TPR. > >6 out of 9 is better than 5 out of 9, like 4 out of 9 sucks completely >compared to 5 out of 9. > >till 4.5 out of 9 points are pretty easy to get IMHO, but above that >every half point is hard to get. Take further into account that the >Necchi book completely sucked everywhere, and you'll end up that shredder >definitely performed better than tiger. Necchi book does not suck everywhere. Necchi commercial book may be bad because shredder does not do good results in the ssdf games but the book for tournaments is a good book and the proof is the fact that shredder wins the world champion title every tournament. It may be possible to win this tournament with a bad book if your engine is clearly better than the other programs but it is not the case with shredder otherwise it could also win the ssdf games. 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.