The issues with Expected Outcome wOBA
There is a lot of luck involved in hitting a baseball. When a player violently swings at a pitch, which sometimes travels over 100 mph, even a small change in where the bat meets the ball could be the difference between a homerun or a pop up. A ball could be hit hard in the air but right at an outfielder for an out, or be a weakly hit grounder to 3rd base that is hit so softly the ball can not be fielded in time. Sometimes, wind can push a ball right out of reach of a fielders glove, while on another occasion it pushes it back just enough for a fielder to get under the ball.
We can strip out one layer of luck by looking at statistics on how the ball came off the bat. We now have the ability to see with Statcast, based on how hard a ball was hit and its launch angle, to determine what was expected to happen based on the quality of contact. This statistic has been captured by a metric called xwOBA. xwOBA is the expected wOBA for a player given the exit velocity and launch angle of all his batted balls, which gives us a better idea of the true talent of a player.
However, if our only goal is to determine true talent, that is to say our only goal is to look at past data to predict how well a player will do in the future, then this metric has flaws. If we assume that a player’s goal on every swing is to hit the ball had and into the air, then xwOBA will give too much credit to weakly hit balls or balls hit at a poor launch, and too little credit to balls that were hit well but not well enough.
Hitters don’t necessarily try to hit the ball hard and in the air on every swing. It would certainly make sense on two strike counts that a player’s goal, first and foremost, is to have some contact with the ball. There are also exceptions like bunting where the goal is to hit the ball very softly. However, I think on most swings, hitting the ball hard in the air is the hitters goal, and for the sake of this essay I’m going to make the assumption that this is the case.
Below is a visualization that gives a good idea how expected OBA is calculated. The color of the dots is a metric referred to as scoring value, which is not wOBA, but a more complex estimate of runs created. However, these metrics are highly correlated, and if wOBA was graphed instead of scoring value, the visual would be essentially the same. The darker the orange, the better the expected batted ball outcome, the darker the purple, the worse the expected batted ball outcome.
On the upper right of the graph, there is a cluster of dark orange dots. These have been deemed “barrels” by the guys over at MLBAM. These are balls that are hit very hard at close to ideal launch angles. These are often home runs and extra base hits.
I believe this is a useful statistic in determining hitter skill. When a player hits a barrel and it lands for an out, we can say certainly that the hitter did his job, he hit the ball hard and in the air, but he just got unlucky. Players who hit a lot of barrels, which is to say have high quality of contact often, are going to be the best hitters.
Right next to the barrel area is a light orange one, a line extending from the far right at about 10 degrees of launch angle over to the far left at about 30 degrees of launch angle. Balls hit with these characteristics often fall for singles and more rarely doubles. These are balls that were either hit in the air but were very soft, or balls that were hit hard into the ground. If a player hits the ball 70 mph (very softly) at a 25 degree launch angle, these are balls that go just over the infielders for a hit. These were not hit well persay, but they were hit to a nice voided area. xwOBA would credit this batted ball profile as a positive scoring value.
Now let’s look at the blue area from about 25-50 degrees of launch angle and 75-90 mph of exit velocity. These are almost always outs. Why? Because they are hit in the air just hard enough to land very close to the depth outfielders are standing. These are often referred to “at-em” balls for that precise reason, you hit the ball right at them. Therefore, xwOBA treats the scoring value of these events as very negative.
Coming back to our assumption about the hitters goal, to hit the ball hard and in the air. If the goal of a hitter is just that, shouldn’t we value their outcomes based on how close they were to achieving that goal?
Let’s say I’m playing golf and after a beautiful drive on a par 4. I have a 140 yard approach shot to the green. I take out my wedge, address the ball, and swing. But I skull the ball, badly. The ball gets hit way too hard at a very low angle. However, because the ball is hit onto the ground, the grass slows it just enough to die right as it gets to the center of the green, several feet away from the hole. The result was one of the best I could have expected. I have hit this exact shot many times, I call it the “so bad it’s good” ball. I didn’t hit a good shot, in fact I hit a terrible one, but it had a good outcome.
This is where expected OBA has a similar failure. Balls that are hit badly, but in just the right way, are rewarded but balls hit well but not quite well enough are punished. Let’s take that 70 mph, 25 degree batted ball and compare it to a ball 94 mph at 35 degrees. The latter ball is incredibly close to being a barrel. If the ball was hit a little harder and a little lower, it would have a huge scoring value. Because it is hit right at the area of the fielders, it often lands for an out, and has a pretty low expect OBA.
But the hitter was very close to reaching his goal on this batted ball, while the first batted ball was quite far from his goal. A ball hit 25 degrees in the air is pretty close to optimal (28 degrees is historically where you find the maximum wOBA), but the ball was hit quite softly. Although he succeeded in getting the ball in the air, he failed drastically to hit the ball hard.
Should the the soft batted ball be evaluated as better contact than the “at-em” ball? I think clearly not. In the same vein, does it make sense to treat one batted ball at 96 mph and 28 degrees of launch angle as very positive and a batted ball at 94 mph and a 35 degrees launch angle as very negative? Yes, the 96 mph, 28 degree launch angle ball was hit better, but only barely. Not enough to be evaluated so much better than the slightly softer ball hit a little higher in the air.
I think expected OBA is a big step up from actual wOBA as far as evaluating skills of hitters. It strips out some of the luck of batted ball outcomes and gives us a clearer picture of how well a player hit the ball. But we should also consider the massive amount of luck involved when a player swings quickly at a pitch traveling quickly with a lot of movement. Several milliseconds of difference between when the swing starts can mean the difference between solid contact and a miss. A difference of inches between where the bat contacted the ball can mean a line drive or a ground ball. This may not be luck in the literal sense, the batter estimated where the ball would cross the plate and controlled how the bat moved, but we are talking about humans and not robots. No hitter swings the exact same way every time and no hitter can tell you with precise certainty where the ball is going to cross the plate when he decides to swing. However, the best hitters are able to hit the ball hard and in the air more often than others, even if most of the time they fail.
One possible improvement is to look at each outcomes as a distribution. When a player hit the ball around the actual launch angle and around the actual exit velocity, what was the distribution of outcomes? This would reward balls hit that are close to barrels and discount balls that were hit badly but have good overall outcomes. We could also look at average exit velocity and how close the ball was hit to an ideal launch angle. High average exit velocity, close to ideal launch angle batted balls would be scored as positive outcomes, where low average exit velocity balls with launch angles far from ideal would be scored as negative outcomes. Whatever the method, I think there are still ways to improve the way we look at exit velocity and launch angle to evaluate quality of contact.