Roger Craig had never been on Jeopardy! before, but by the end of his first day of taping, he'd won five games in a row, the most lucrative day for any contestant in the show's history, including the most lucrative game in the show's history. His secret? A web app that modeled the show's all too predictable question sequences.
Craig, to be sure, is a very smart cookie. In addition to game show winnings totaling more than $195,000, he earned a computer science Ph.D. from the University of Delaware. In the talk above, delivered to the New York "Quantified Self Show & Tell," he explains how he developed a web tool and various supporting programs to analyze and effectively train himself on a database of past questions. Among his findings: high value questions are culled from academic fields like art, science, biology, and architecture, while low value questions tend to come from more popular topics like food, "firsts," and inventions. Based on these types of weightings and his performance on random questions selected by his app, Craig was able to deduce topics he needed to study, like fashion.
Three other contestants used the web app to good effect on the show, at least according to the anonymous quotes collected by Craig. His next venture: an iPhone app that will let anyone study in a similar fashion. Jeopardy! producers should hope it's not half as good as Craig's original software. No one wants to watch three bored nerds plowing through their questions.