Or maybe, “The One Mil­lion Dol­lar Algorithm”.

A com­pe­ti­tion to improve the rec­om­men­da­tion engine of the online DVD rental com­pany, Net­flix, has been run­ning in to problems.

As the con­tes­tants edge toward an improve­ment rate of 10% (the point at which the $1,000,000 prize will be awarded), their progress grinds to a halt thanks to a small selec­tion of films that are noto­ri­ously divi­sive and dif­fi­cult to pre­dict. The New York Times reports that this prob­lem is being called the Napoleon Dyna­mite Prob­lem:

Math­e­mat­i­cally speak­ing, “Napoleon Dyna­mite” is a very sig­nif­i­cant prob­lem for the Net­flix Prize. Amaz­ingly, Bertoni has deduced that this sin­gle movie is caus­ing 15 per­cent of his remain­ing error rate. […] And while “Napoleon Dyna­mite” is the worst cul­prit, it isn’t the only trou­ble­maker. A small sub­set of other titles have caused almost as much bedev­il­ment among the Net­flix Prize com­peti­tors. When Bertoni showed me a list of his 25 most-difficult-to-predict movies, I noticed they were all sim­i­lar in some way to “Napoleon Dyna­mite” — cul­tur­ally or polit­i­cally polar­iz­ing and hard to clas­sify, includ­ing “I Heart Huck­abees,” “Lost in Trans­la­tion,” “Fahren­heit 9/11,” “The Life Aquatic With Steve Zis­sou,” “Kill Bill: Vol­ume 1″ and “Sideways.”