Ergo, when I pit machine against machine in a serious way, I tend to opt for approximately equal conditions: say 60 seconds per move, ponder off (when possible). But I'm also interested in the effect of programmed opening books on overall performance. I've wondered how much size matters in opening books. I think we've id'd the rankings for largest opening book, though some uncertainty surrounds the Mephisto Montreux:
http://www.hiarcs.net/forums/viewtopic. ... b16347077c
In the thread linked above, Larry said this (emphasis added):
Monstrous opening book sizes will do little to help playing strength if the valuation function once the book runs out can't take advantage of a position deemed better by a professional chess player.
Steve, alluding to the roughly equal strength of Montreux (bigger book) vs RISC 2500 (smaller book), said this (emphasis added):
i guess it means that the book size doesn't matter all that much when playing two computers against each other that have basically the same engine and hardware
basically the Risc and Montreux have the same King engines although released 3 years apart
Selective Search shows the ratings for the two to be very close with the difference due largely to the hash table size...
Montreux(128k)-2210
Risc 2500(128k)-2191
Risc 2500(512k)-2231
Dead right, of course - the debate in the earlier thread shifted to a Montreux/RISC 2500 comparison, which was productive, but left something else unanswered - namely, how much does book size matter?
Larry tackled it above - let me quote him again:
OK. So is it possible to measure the thing he's talking about? It's a facet of the horizon - I understand that. But what's it worth?...if the valuation function once the book runs out can't take advantage...
That led me to another thought: When I turn off the opening book on my various chess phone apps, I tend to win more often.
That got me thinking about something else, a kind of odds-giving scenario: What if you pitted a lower-rated dedicated against a higher-end machine and allowed the weakie to keep its book? When would the relative strengths find the line of balance? Could you (for instance) run Milano (book on) against Atlanta (book off)? What's the widest potential disparity?
The same experiment could apply to lower-rated machines. For instance, what's the strongest book-disabled opponent Super Connie could beat?
Accepting proposals regards,
- R.