The mahalanobis distance? > > > http://stats.stackexchange.com/questions/62092/bottom-to-top-explanation-of-the-mahalanobis-distance > (habanero cookies answer) > > So here's a conjecture. The mahalanobis distance of all golfed programs solving the same problem for all equivalently expressive languages should be the same.
With enough data, maybe we can answer the following questions. 1) Is a particular problem harder/easier to solve for a particular language (e.g. problems requiring matrix manipulation). 2) Is there a best value? Is there no scaling needed? 3) Is there a best value? Is scaling necessary because humans have a bias? 4) Could language A co-opt design from language B based on the above (e.g. generators are cool). I wonder if the AST of the solving program could be useful; count the number of non-terminals. I think one could find a parser for pretty much any language these days. Can anyone thing of any other features? BTW Julia already has code_llvm(), how easy would code_AST() be?
