The analysis has some of the same flavor that might be applied to quant model evaluation in Net Interest Margin optimization error analysis.
Raymond Smullyan is probably the world’s greatest expert on the logic of lying and the logic of chess. He is still writing books well into his 10th decade. Last year he published a new textbook, A Beginner’s Guide to Mathematical Logic, and in 2013 a new puzzle book named for Kurt Gdel. His 1979 bookThe Chess Mysteries of Sherlock Holmes introduced retrograde analysis—taking back moves from positions to tell how they could possibly have arisen—to a wide public.
Today Ken and I wish to talk about whether we can ever play perfect chess—or at least better chess than any one chess program—by combining output from multiple programs that sometimes might “lie.”
We will start with Smullyan-style puzzles today, but they are prompted by an amazing and serious fact. Even though human players have been outclassed by computers for over a decade, humans judging between multiple programs…
View original post 2,669 more words