Monday, December 06, 2004Prospectus and planning their roto teams than they do watching actual games. Football fanatics now have Football Outsiders, and fantasy football is a multi-million dollar business. But hoops fans, especially college hoops fans, tend to spend their time watching games; media coverage tends to be either reporting or punditry, with very little quantitative analysis beyond the usual box score stats. (Semi-notable exception: John Hollinger's Pro Basketball Prospectus. More on this later.)
So should we care? After all, if there's nothing you'd rather do than watch games, why waste your time with stat-heads? Well, I'd say that at their best, statistics enhance my experience as a fan (Disclosure: When I'm not guest-blogging, which is most of the time, I work as an economic consultant and spend my days running regressions and financial models, so maybe my interest for stats is a little higher than most). To paraphrase Moneyball, a great stat acquires "the power of language" -- it can confirm what we think we know, teach us things we didn't know, and provide context for historical debate. Ideally, composite stats should let us test propositions like "Duhon is overrated" or "Okafor makes the biggest defensive impact" -- let us confirm or reject ideas that commentators spout every day without justification. A great stat can allow us to identify undervalued skills and predict players' and teams' futures. Baseball has Bill James' Win Shares and Prospectus' VORP, which estimate an individual player's impact on his team's wins and runs, respectively; Prospectus' PECOTA forecasts are a highlight of the baseball offseason. Pro basketball has only Hollinger's Player Efficiency Rating (though I wonder what kind of models Mark Cuban is using for his new sports gambling hedge fund), and the college game, to my knowledge, has nothing.
Tomorrow I'll write about some obvious and not-so-obvious problems I perceive with the current use of stats in college basketball and the development of better stats, and after that I'll write about some preliminary ideas for solutions. A lot of what I write will be as analogy to baseball, with its better-developed stats. I don't pretend to be totally original in looking at this topic, and I'm sure some efforts have been made along these lines that I don't know about. Your input and ideas will be extremely welcome.