Author: Eduard Nemeth
Date: 09:10:16 01/01/03
My original "german message here" on my homepage: Eduards World: http://www.beepworld.de/members37/computerschach/kommentar.htm Translation with Internet Translator: 2003-01-01 - If tests have retired recently gave it in a German computer chess forum on the same day of two totally different results of two programs those to have played against each other to see. A tester brought a result of 10-10, while the other one presented 9.5-20.5. The reason for this extreme deviation lies in used opening books. While a tester a Remis Einheitsbuch used, the other tester used a short book. The short book contains only few courses e.g. in such a way: 1.d4 d5 2.c4 e6 3.g3 Sf6 4.e3 and now had to show the programs themselves which them drauf to have! The portion with this zugfolge nevertheless lasted about 70 courses, and it won the program also 20.5 points had brought in! After 70 courses to win here, thus no coincidence can be, or? I have the program that was subject clearly here with strange opening courses also tested, because I wanted to see which it about chess strategically understand, thus no festgefahrenen paths (longer variants) am given. In such a test portion the program played also black once a7-a5, pulled its Springer thereupon deplatziert after a6 over after exchange runner against Springer with the tower from a8 to to strike. That did not see any longer well out however it followed next still the tower course Ta6-Ta8. Black wasted here thus several courses around strategically a backward step instead of progress to erspielen. No miracle thus that another program, which knows to use this strategic weakness such positions then wins! Over the result 9.5 - 20.5 (and it was not the none like that) I was not surprised therefore. With the Remiseinheitsbuch the thing ran completely differently. The variants are much longer, a complete opening with determined plans are already on the board, in addition (like the name already it says) the tendency is inclined to the fact that the portions will end with high probability Remis. Here such a variant from the mentioned Remisbuch: 1.e4 e5 2.Sf3 Sc6 3.Lb5 a6 4.La4 Sf6 5,0-0 Le7 6.Te1 b5 7.Lb3 d6 8.c3 0-0 9.h3 Lb7 10.d4 Te8 11.Sg5 Tf8 12.Sf3 Te8 13.Sbd2 Lf8 14.a4 h6 15.Lc2 exd4 16.cxd4 Sb4 17.Lb1 c5 18.d5 Sd7 19.Ta3 f5 20.exf5 Sf6 ends here the main variant with the highest Ausspielwahrscheinlichkeit, whereby the repetition of course is highest actually with the course 12 according to Ausspielwahrscheinlichkeit, I selected with 13. Sbd2 however a course with that, with tournament option provided book only with 31% probability was out-played, while the repetition of course with 67% stands to beech! Probably thus an imminent Remis is after course 11. There a result of 10-10 does not surprise me thus also, and even if the portion had not kept running after course 20 like that is it for both sides simply to win. Nevertheless, I agree with both test methods and find her good, if one interprets the results only correct thereafter understands! Thus us very beautifully the short book shows which programs without book assistance is capable to achieve. Profits becomes here that program of the game of chess the more understands! With the Remiseinheitsbuch refinement are in demand. The program will usually win also there more chess knowledge possesses and the refinement therefore better converts is able. Since however the Remiszone is large it close results will usually too much come, and it should not surprise also if 10-10 stands there! Before five years about, the first large Testerei began with such short books on broad level, whereby the variants were longer as the short book and shorter than the Remisbuch. The idea still comes from the magazine CSS. one wanted on the basis short variants, in addition with exchanged colors, finds out which program without tournament opening books the strongest is. One assigned the GM Dr. John Nunn to prepare 10 positions which are best suitable for it. Supplied then a very interesting test to the GM with, with 10 short variants. It was designated Nunn 1-Test. The variants were as said very interesting, but I was not astonished nevertheless badly as I the variants under the magnifying glass took somewhat more exactly. Here too far if I each individual variant would go out of my view exactly explains. However that the variants demanded altogether sharp continuations, likewise a higher measure of tactical being able stands firmly instead of strategic! As far as I can remember, the GM Dr. Nunn is a exzellenter technician, however altogether rather an attack player, at least in former times. For me it was therefore also no miracle that from the 10 variants, three Sicilians are thereby, and all three have it correctly fetzig in the sense, it come in all three variants to sharp play, or doubt someone if one it the name "Schweschnikov variant" and "farmer robbery variant" whispers in the ear? Also some other continuations led to immediate exchange in the center, and in the king Indian were the courses f7-f5 and f2-f4 quasi pre-programmed Springer for the Vordre farmers already place made. Altogether in this test very little strategic knowledge of programs demanded, required was rather tactical being able! Now, here at that time 1998 Fritz 5 were actually also garnet, although otherwise in the final game a still very stupid program! At that time it was therefore also no miracle for me that Fritz 5 against all its competitors in this discipline won! Genius, Rebel and Mchess or also Hiarcs were clearly defeated. A test for itself the requirement demanded at that time to be able to test programs purely objectively without tournament books after their schachlichem being able would have therefore also today the same requirement for itself to raise, if it were regarded at that time as absolutely objective, or? Today we have the 01,01,2003 and I a while ago on the Downloadseite of the CSS umgeschaut ourselves. There is not a Download of this Nunn1-Tests to find there any longer! In order to make it short: the test has retired, because otherwise would be it still there, or? It ask itself thus only does why he has retired? This question can be answered easily: The today's programs (particularly Fritz) are not any longer further tactically trimmed. Yes, particularly with Fritz one is now particularly anxious to implement it as much as possible knowledge! Fritz is to be able to play chess also beautifully and supply apart from tactics now! A test like the Nunn1 (with action purely) is now no longer in demand now thus. The same which to Fritz applies, applies nowadays also to the other programs. But one invented now different tests, how the Nunn2 test (less tactics) or the WM test (consisting of test positions divided in categories "final game", "position play" and "king attack" - both tests are on the CSS homepage to the Download, the Nunn1-test there are however no longer there, at least today! The Nunn1-Test had served its purpose and has today retired. The tactics monster at that time "Fritz" was umjubelt at that time, while other very beautifully playing programs, thereby the tragic role to have played. Much harm! Which happened today to the Nunn1-Test, can be issued the today's tests tomorrow likewise. Result: there are not objective tests, even if some testers or their promoters liked to put it in such a way. There is however also tester (fortunately) no Absolutheitsanspruch of its tests raises - in majority however unfortunately such testers are those exactly to know want which program at which rank lists place to be led are, unfortunately! Eduard
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.