The New BvP
If we could somehow get each batter to face each pitcher 500 times, I don’t think there would be many more stats that would be more predictive than BvP. Again, in theory, it rocks. In practice, it sucks because of sample size issues.
If we could somehow simulate what those at-bats would look like, we’d be in good shape. Well we can. Kind of.
One of my favorite ways to analyze players whenever there are issues with sample size is to look at comps—similar players in similar situations in the past. When Nate Silver worked in baseball analytics, he created a projection system called PECOTA that did just this, projecting baseball careers by looking at how similar players performed. Since we can’t know for sure how likely it is that a 20-year old rookie who can hit for power to all fields will crank 500 career home runs, we can look at similar players to him and use those careers as a basis for our prediction.