Single Comment

[Hey, play a game of 8 Camels, half one binding half the other, behind regular Pawns and 8 CG behind regular Pawns! No comparison, probably against G too!] Thanks for input, as piece-values are neglected in favour of designing cvs and defining types. Jeremy's are good if obvious points. What do we have now Grasshopper versus Contragrasshopper? 2.5 and 4.0 were the second try, now estimate in general 2.5-3.0 and 4.0-4.5. They are both measureably over convenient benchmark Camel most embodiments. CG and G using all Queen lines as they do are likely still underestimated those ranges most CVs possible to include them (fortunately CG as yet appears in none). Not only there is nothing wrong with progressive recalibration, it is the only sound approach, because there is no one exclusive piece-value method that can work. There are conflicting and reinforcing methodologies -- until settling on one rules-set and initial position for computer tests. As Jorg Knappen recently says, the only way to get refinements down to +/- 0.1 is to generate play with at least 20 or 50 game scores w/wo computer. Muller somewhat mis-leads that is a 10-minute process, because designers can re-invent rules-set variations and piece mixes at one a minute and the differing CVs can never be overtaken by single established values per p-t. Piece-values have effects from the array and from compatibility with other pieces and from specific rules. Gilman has documented so far about 3000 p-ts here in M&Bxxs and for this subtopic, singled out are two of them, G and CG. Among the other 9,000,000 pairs at least 10% might be interesting to consider one to one. That is still another 900,000 to look at by trial and error and selected methodologies. The only way would seem to be one by one, as pair, within 100s of possible rules-sets each piece-pair to compare by educated guesses, if one will, from growing experience. Just because proficiency at it is difficult or frustrating does not mean CV after CV, as well as M&Bxxs, should always omit serious attempt at piece-values. I did about 30 of these design analyses 8 yrs. ago, Jacks&Witches having all piece values and intend to build on them. The exchange gradient leads to the Move Equation, and the number of moves per game has some empirical testing in Game Courier as well as Brainking.