For those of you that visit Fangraphs, wOBA has become a favorite stat of not only this website, but many other sabermetrically-inclined sites as well. The theory behind it is simple: assign value to individual offensive events via linear weights and then apply those numbers to a player’s performance. We go ahead and adjust the formula so that it looks like OBP, meaning that league average usually falls ~.333 and is properly distributed along the curve.
But there has recently been more research done into linear weights so that we can see the value of not only a single or a double, but of certain pitches. While this has been mostly popular in the domain of Pitchf/x, we can also use this information to further our understanding about hitters. For example, let’s say Miguel Cabrera and Magglio Ordonez come up to the plate consecutively. Miguel Cabrera works a 3-1 count and then lines a fastball into the right field gap for a stand up triple. Next, Magglio Ordonez hits a ball in the exact same spot on an 0-2 count and slides in safely at third. Back-to-back triples. For our purposes of wOBA, we would assign both Cabrera and Ordonez equally with ~1.56 runs, the linear weight value of a triple relative to an out.
But was that all we could gauge from those at-bats? With count-based linear weights, we can actually do more. Miguel Cabrera worked a 3-1 count before his triple; there’s value in knowing that information. Since we can say that a 3-1 count is worth somewhere around .14 runs, why not credit Cabrera in some capacity for getting into that count? Likewise, we know (based on the same run value charts linked last sentence) that an 0-2 count is worth roughly -.104 runs. Why not also take that into account? The moral here is that those triples were not made equally.
But we do have to be somewhat careful we don’t double count. A player is more likely to hit a triple (or for our purposes, get a higher run value) if he gets to a 3-1 count, and the oppposite is true for an 0-2 count. Those count-based linear weights are based on how many runs are likely to proceed from that count, so we would probably have to regress the run values somewhat so we don’t double credit a hitter. Maybe the most interesting experiment would be to just take a batter’s count-based linear weights for an entire season and compare players, or even apply their batted ball linear weights for if/when they put the ball in play to their count-based run total.
This is a thought experiment, so I’d like to see what people think. The next step may be crunching the numbers.
Print This Post