Author: Vasik Rajlich
Date: 01:35:24 02/17/06
Go up one level in this thread
On February 16, 2006 at 21:32:48, Rolf Tueschen wrote: >On February 16, 2006 at 21:07:02, Chessfun wrote: > >>On February 16, 2006 at 19:24:41, Rolf Tueschen wrote: >> >>>On February 16, 2006 at 18:37:18, Uri Blass wrote: >>> >>>>On February 16, 2006 at 18:28:31, Rolf Tueschen wrote: >>>> >>>>>On February 16, 2006 at 17:58:27, Chessfun wrote: >>>>> >>>>>>As this table shows other engines like Spike 1.0 and SmarThink are far closer to >>>>>>Fruit than Rybka. As for Toga the numbers speak for themselves. >>>>>> >>>>>>http://kd.lab.nig.ac.jp/chess/CCRL-4040/engine-distance-table-all.shtml >>>>>> >>>>>>Sarah. >>>>> >>>>> >>>>>I doubt that these results have a bigger meaning but just an impression: it >>>>>looks as if Hiarcs has bought something from Chess Tiger. >>>> >>>>I think that simply not enough moves were played. >>>> >>>>98 moves are not enough. >>>> >>>>Uri >>> >>>I saw it here >>> >>>http://kd.lab.nig.ac.jp/chess/CCRL-4040/eval-difference-table-best-limited.shtml >>> >>>but you are right, this is strange. Allegedly it is 150 games but only 98 moves >>>counted... So therefore all in all I concluded that the tables have no value. >>>Also the background of the tables isnt discussed at all. Why this and what >>>exactly... That isnt sound. I see we two are agreeing here, Uri. >> >> >>Rolf, >> >>You should read more before posting something like that. It isn't strange at all >>the pair you made an assumption about have played only a couple of games. >>http://kd.lab.nig.ac.jp/chess/CCRL-4040/pairwise-results-all.shtml >>Plus you didn't look at the link I provided and read what was written. Then you >>make an assumption about a difference table after I posted distance. >> >>Sarah > > >Excuse me, but I didnt talk anything about you. I was just not impressed enough >by these tables on that page. That's all. Of course I can also assume that there >are some theories behind that sort of tables. But I say that there should be >more explained in such a table. It is simply something you expect if you see >such a table. Nothing against those who made these tables. They did what they >could - I think. But still one should speak out what one thinks. Criticism isnt >something negative. Critic clarifies. Perhaps you also consider that I'm not >known for doing tests myself. That's correct. But still I once had studied such >problems and the graphic presentation. And from these memories I post something >from time to time. In my case you should forget all what you normally assumed >when you read something from younger people. I do never write something against >a certain "name" or such; I dont know someone, with some exceptions. It's only >the topic that interests me. - What I liked in these tables is the colored >details. Very interesting. > >Let me just summarize my critic for you, Sarah. If you want to present your >results in such a table and you create different colors and almost 90% of the >cases fall into a specific category (here greenish), then such a presentation >makes no sense. Because, ask yourself, what is the conclusion? And if everything >is almost the same, why then so many categories? Know what I mean? A graphical >picture should always clarify a context that is otherwise a bit difficult to >understand. But this here is so simple that you dont need all the details, a >short statement would have been better. > >And at last Vince made a good add-on which leads me to the conviction that such >presentations make _here_ no sense at all, because what you get ISNT really what >it seems. The results were completely different with little changes in the >program code. And I didnt even know this when I wrote my own critic. But somehow >I could smell it, that here something was pretended (with all good intentions) >which doesnt stand. What this experiment measures is the extent to which the evaluation concepts of the programs overlap. Indeed, a drastic change in evaluation concept could be accomplished with a very small change to the code. Likewise, you could code a very similar (or identical) evaluation concept in totally different code. Vas
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.