Author: Robert Hyatt
Date: 10:11:08 01/20/04
Go up one level in this thread
On January 20, 2004 at 11:20:31, Mike Hood wrote: >On January 19, 2004 at 12:31:17, Robert Hyatt wrote: > >>On January 19, 2004 at 04:57:01, William Penn wrote: >> >>>On January 18, 2004 at 23:19:26, Robert Hyatt wrote: >>> >>>>On January 18, 2004 at 23:10:55, Bob Durrett wrote: >>>> >>>>>On January 18, 2004 at 22:39:23, Robert Hyatt wrote: >>>>> >>>>>>On January 18, 2004 at 05:53:49, martin fierz wrote: >>>>>> >>>>>>>i just received my brand-new shredder 8. when using it for analysis in chessbase >>>>>>>(which is in fact the only thing i use it for), it often gives PVs which are >>>>>>>completely ridiculous - the first few moves are ok, then one side blunders a >>>>>>>piece according to the PV, but the evaluation of that line doesn't show it. >>>>>>>looks like there is a PV bug in shredder 8? is there any fix for this? i find it >>>>>>>very annoying... >>>>>>> >>>>>>>cheers >>>>>>> martin >>>>>> >>>>>> >>>>>>I have heard that Shredder (and others) try to reconstruct the PV by probing the >>>>>>hash table at the end of the search. This simply does not work with any degree >>>>>>of accuracy. IE suppose you search and reach position A while searching the PV. >>>>>> Later, at very shallow depths, you reach position A again and >>>>>>overwrite it with different "best moves" depending on the depth remaining, >>>>>>extensions triggered, etc. Now when you try to recover the PV from the hash >>>>>>table, you get the right position A, but the wrong best move. And then the PV >>>>>>looks funny. It doesn't happen every time, but if the PV is reconstructed >>>>>>enough this way, it happens often enough. I tried this _years_ ago and ran into >>>>>>the same problem. Never saw it in debugging. Saw it regularly when kibitzing >>>>>>PVs on ICC. :) >>>>>> >>>>>>I now do it the correct way, backing the PV up along with the score... >>>>>> >>>>>>Remember that this is speculation since I have never seen Shredder's source. But >>>>>>recovering the PV in this way is simply going to produce errors, and there is >>>>>>nothing that can be done about it. The first move and score will be correct, of >>>>>>course. But beyond that, who knows, and the farther out, the greater the >>>>>>probability of a bogus move. >>>>> >>>>>There was much discussion here about this problem with Shredder 7.04. You would >>>>>have thought that this would have been corrected it for version 8.0 but maybe >>>>>we'll have to wait until Shredder 8.04. : ( >>>>> >>>>>Bob D. >>>> >>>> >>>>He might not want to worry with it. It isn't a tough change, but it is a >>>>change... >>> >>>I have seen it suggested that he may prefer to leave it that way to hide his >>>exact search algorithm, so that others can't copy it. It might be something >>>special, unique, that nobody else knows about (yet). Possible? Or just verbiage? >>>WP >> >>Certainly a possibility. Showing the _real_ PV tends to expose part of your >>search strategy. Showing partial, incomplete and sometimes bogus PVs makes that >>harder to understand. :) > >I reported this bug to Chessbase when I first saw it in Shredder 7.04. This was >the example I used: > >[D] 3r2k1/1pp2ppp/3r4/1P1npq2/3n4/3PPP2/1BQ1BP1P/2RR3K w - - 0 1 > >1.exd4 Nf4 2.Rg1 Rh6 3.Rg4 Qh5 4.h4 Qxh4+ 5.Rxh4 Rxh4+ 6.Kg1 Rd6 7.Qxc7 Rg6+ >8.Kf1 Rh1# >² (0.70) Depth: 12/34 00:00:50 6470kN > >Although the eval is only 0.70, the PV leads to mate. > >The reply was, in my own words, "We know about this problem, but Shredder 7.04 >is still the world's strongest chess program". Or, to paraphrase more freely, >"Who cares what the analysis says as long as we're the best?" Crafty can produce _that_ problem, because it doesn't detect check/checkmate in the quiescence search. Yet when I display the PV, I do notice that Rh1 is a check and that the opponent has no legal moves so it is mate. I'm less worried about that (that is fixable in fact, just don't append # if the score does not lie within the checkmate boundaries) than I am about actual bogus moves being in the PV, but not in the actual path that leads to the score shown at the root.
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.