Author: F. Huber
Date: 01:45:12 05/29/05
Go up one level in this thread
On May 28, 2005 at 20:08:29, John J. J. Smith wrote: >On May 27, 2005 at 18:19:42, F. Huber wrote: > >>On May 27, 2005 at 17:56:34, Uri Blass wrote: >> >>>On May 27, 2005 at 16:57:35, F. Huber wrote: >>> >>>>On May 27, 2005 at 16:53:49, Rob Basham wrote: >>>> >>>>>Well, at least on this move...:-) >>>>> >>>>>Mate in 10.... >>>>> >>>>>[D]4r1k1/rp2Bppp/p1b5/1q2Q3/R7/1P5P/P4PP1/3R2K1 w - - 0 1 >>>>> >>>>> >>>>>Both the Risc (dedicated unit) and Fritz 8 on PC find Qxg7+ >>>>>in just 9 seconds! >>>>> >>>>>P3-850 >>>>> >>>>>Rob >>>> >>>>Well, and ChestUCI finds it in less than 1 sec on my slow Celeron/400! :-) >>>> >>>>Regards, >>>>Franz. >>> >>>But ChestUCI *default setting* needs 50 minutes to find it on A3000 >>> >>>New game, 10'/40+10'/40+10'/40 >>>4r1k1/rp2Bppp/p1b5/1q2Q3/R7/1P5P/P4PP1/3R2K1 w - - 0 1 >>> >>>Analysis by ChestUCI Ver.3.7: >>> >>>1.Qxg7+ >>> +- (#10) Depth: 10 00:50:05 1081113kN >>> >>>(, 28.05.2005) >>> >>>It proves that there is no mate in 9 in the process. >>> >>>Uri >> >>There are 3 important points to mention: >> >>1) Do you really think, that it is fair to compare ChestUCI´s default settings >>(where it _proves_ the shortest mate with a brute-force search) with a normal >>chess engine, which does all kinds of search extensions, cutoffs, ... ? >>Or is it really so hard to switch _one_ option to Automatic-mode? >> >>2) And as I can see from your output, you´re still using a _very_ old version >>of ChestUCI, which doesn´t even have the new ´ThreatDepth´ feature - >>and exactly _this_ function speeds up its search dramatcally! >> >>3) When comparing the solution times with normal chess engines (like this >>dedicated RISC unit or Fritz 8) it´s not necessary at all to _prove_ that >>there is no mate in 9, simply because NO engine actually _proves_ this, when >>looking for a mate - so why requiring such a prove from ChestUCI? >> >>Franz. > > >Franz > >Can you please explain exactly which options you set to get such a fast >solution? Do you always use autoamtic mode? What about Autoturbo mode? > >Thanks. >John Hi John, well, for the above mate ChestUCI´s solution is: 4r1k1/rp2Bppp/p1b5/1q2Q3/R7/1P5P/P4PP1/3R2K1 w - - bm #10; 00:00 @ C2/R1/K3/P8/X34; But of course you don´t have to set all those parameters manually - simply use ´AutoTurbo´-mode and you´ll get them within a few seconds (except C2, which has to be tried and set by yourself, if wanted). And your other question is not so easy to answer - there isn´t really a general rule which mode to use or which options to set, otherwise I would have used exactly those parameters as fixed settings in ChestUCI. ;-) But let me describe shortly, how I usually try to solve mate problems with ChestUCI: First of all I always try ´Normal´-mode (i.e. a brute-force search), because this is the main purpose of Chest - finding (and proving!) the shortest mate. Only if I can see (from the progress in depth), that this will be hopeless, I´m switching to a mode using special parameters, normally to ´AutoTurbo´-mode. In this mode I´m using ChestUCI´s default values (except maybe setting the ´ThreatRange´ to 6 instead of 3), and again I´m looking how it progresses. If for higher ´ThreatDepth´-values (R4,R5,...) the search becomes too slow, I often abort the calculation, switch off threats completely (ThreatRange=0), and let ChestUCI look only for the other parameters (K/P/X) - in this case it´s sometimes also quite useful to change the ´SearchOrder´ to ´XPK´ (instead of ´KPX´), because many mate problems are solved much faster in this way. And finally if even those ´tricks´ are not successful, _sometimes_ I´m setting one or two parameters manually (mostly ´K´ or ´P´), but that depends on the current position: looking at it sometimes you can guess, that e.g. the king (of the defender) could not move at all (K1) or might have only one move (K2) in the complete solution, or that only a few defender pieces could move (Px). Of course _this_ method requires some experience with ChestUCI and mate solving generally, because there don´t exist any universally valid rules - but mostly the ´AutoTurbo´-mode finds the solution by itself, and even quite fast! :-) Best regards, Franz.
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.