One thing I look at when mining for Draftstreet bargains is the “last 30 days” filter on Fangraphs leaderboards. While there has been much work done on the predictiveness (or lack thereof) of “streaks” (for one example, see The Book), you aren’t necessarily falling prey to the Gambler’s Fallacy when looking for guys who have good underlying peripherals. Four weeks is not a long time, but for many players, it may be enough to recognize fundamental shifts in skill (batting eye, contact rate, etc.). More importantly, since valuation of players inherently is skewed towards counting stats (in Draftstreet’s case, fantasy points per game), you can get a good sense of whether a “hot” or “cold” player’s value has been artificially inflated/suppressed and exploit arbitrages that way.
Some examples of the stats I look at?
Read the rest of this entry »