Author: Robert Hyatt
Date: 17:54:05 06/19/03
Go up one level in this thread
On June 19, 2003 at 18:19:58, Tom Kerrigan wrote: >On June 19, 2003 at 18:01:57, Robert Hyatt wrote: > >>On June 19, 2003 at 17:17:15, Tom Kerrigan wrote: >> >>>On June 19, 2003 at 16:43:52, Robert Hyatt wrote: >>> >>>>I take the comment as what I said. The original position 2 had 8 pieces. He >>>>complained about the evaluation. Even taking three pawns off it _still_ was >>>>wrong. >>> >>>What in the WORLD are you talking about?? What article are you referring to? I'm >>>going to this URL, which is at the top of this thread: >>> >>>http://www.clubkasparov.ru/521772350.html?462691585533321 >>> >>>What do you mean, "original position 2"? Has the article changed? >> >>"original position 2": the position given in the article. >> >>"modified position 2": the position you get if you "remove three pawns >>as suggested by the author." >> >>Was that too hard to understand??? >> >> >> >>>? Because the >>>current position 2 has 7 pieces (not 8) and if you take the pawns off (for what >>>reason?), you're left with KB vs. KB, which you'd think Junior and Fritz would >>>evaluate as a draw, and the couple sentences of discussion about it don't talk >>>about databases at ALL. >>> >>>-Tom >> >> >>On the other hand, you are correct. I meant position 1. Not position 2. >> >>So in the above replace 2 with 1 in both places. And we are back to the >>point. Removing three of the pawns produces a bogus evaluation. Because >>the tables were _not_ working. He _should_ have known that before writing >>an article and publishing it. >> >>If you remove three pawns, it will be in the 5 piece tables. If he really >>had them. He implied he did. But he didn't. And that leaves out the fact >>that _some_ programs can get this one right without tables, even with the >>original 8 pieces on the board. Mine can, for example. >> >>There is little excuse for saying something doesn't work, when it doesn't >>work because of a user error. Publishing it is even worse. > >I think it's not optimal. It seems like he should have thought twice when his >5-man databases didn't give him the solution to a 5-man position. One can argue >that his complaints are legitimate because the programs gave him the impression >that the databases were installed when they weren't. > >I've always agreed with that point though. > >Now that you've dropped your assertion that the guy is an idiot because he >wasn't counting kings as pieces in his 5-man positions, it seems like this has >been discussed to conclusion. > >-Tom That was only _one_ implication in my post. What he meant isn't extremely clear. Complaining about a bogus eval before reaching 5 pieces is wrong for the reasons I have given before. Complaining about a bogus eval after reaching 5 pieces is also wrong because something is _obviously_ broken. But there are other things. The title is about "computers are dumb at endings" although not exactly in those words. That is _also_ wrong. I get position 1 right without tables. Others probably do also. There is a version of crafty that will get the totally blocked positions right, although I don't run with that code enabled at the present, even though it seems to work well. So the base premise is also flawed. There are _other_ flaws. Whether he thought position one should access a table or not is only one point out of several. I don't know what made it _the_ point in my post...
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.