Number correct matches – 10 * number of false positives – number of false negatives
The top entrants and scores were:
- Alex R, 18207 (Java, 1.7 s)
- Steve H, 17887 (Python, 60 s)
- Perry T, 14442 (Python, 50 s)
Congratulations to Alex R, who won $500 for his efforts!
The maximum score possible was 19633. Alex’s solution dominated all categories: most correct matches, fewest false postivies, and fewest false negatives. There were 13 entries with an average score of 12330 and a standard deviation of 4058.
Unsurprisingly, given the scoring function, those who avoided false positives did the best. In general, and because of the focus on precision, ugly heuristics did much better than more elegant probabilistic models. Some people employed a mix of the two with good success.