BTW, the Dale-Chall formula uses two factors to score readability: (1) Average sentence length (longer is worse), and (2) percent of "difficult" words. They've done scientific validation of this small and easy set of metrics. (See The New Dale-Chall Formula.) To determine whether a word is "difficult", they created a list of 3,000 "easy" words that 80% of 4th graders know. (It was developed to help design textbooks for children.) It's an interesting algorithm which is an easy-to-implement proxy for deeper linguistic analysis.
I've found that this very simple algorithm does a decent job of putting the easiest-to-grok definitions at the top. Meanwhile, sentences with parentheticals, lots of prepositions and passive voice are rated less readable and pushed towards the bottom of the page.