Unaccounted For Changes In Exit Velocity

Predicting bat speed using the publicly available Statcast data is easier said than done. For much of the past few years there was a section on Baseball Savant which displayed a bat speed number of each player, but without much explanation for how it may be calculated. I haven’t inquired for an explanation, but I feel rather comfortable saying it was probably a derived stat using a formula published by Alan Nathan.

This formula takes the pitch speed and batted ball speed, and manipulates them using laboratory tested values for the various relevant coefficients–basically the bounciness of the ball and the bounciness of the bat. If you assume values for those coefficients, you can get a rough estimate for bat speed by plugging in the pitch speed and batted ball speed.

I don’t have proof that this is how bat speed was being estimated by Baseball Savant, but I feel it is the most likely explanation for the numbers.

Two weeks ago I proposed a formula for estimating future exit velocity using past exit velocity and launch angles. This method is far from perfect, and there is a whole lot more research that can be conducted into this area.

Over the past week I have been thinking about what performance changes may or may not be predictable from one season to another. Part of the variance that we see from season to season are large dips or climbs in offensive production, which often in retrospect we might be able to explain. Maybe there were signs that pointed towards decline, but we overlooked them for one reason or another. Maybe we didn’t know what the signs meant until further research had been conducted.

No doubt, these mistakes are often due to a lack of information. In some cases it may be bat speed. We don’t really know how much of a role bat speed plays between seasons or during the course of a career. We don’t know how injury plays a role with bat speed, nor do we understand the aging curve.

Part of my method for predicting future exit velocity relies on a fudge factor. I am not entirely sure what this fudge factor is, but I do believe a large part of it is bat speed. Another aspect is almost certainly plate discipline. While I do have other ways of measuring plate discipline, which I’ve written about recently, decoupling discipline from bat speed is too large a burden for me today. For that reason, I will only speak of this fudge factor itself, and from this point forward I will refer to it as bat speed for the purposes of simplicity.

I have run predicted versus actual exit velocities for the past three years for each batter with at least 100 plate appearances in two of those years. One aspect jumped off the page right away: exit velocities in 2016 are higher than they are in 2015 and 2017. In fact, 2017 is the lowest, and 2015 is just about halfway between 2016 and 2017. I only have data from these three seasons, but this information leads me to believe 2015 had a roughly average year, with the following two years representing normal variance.

Of course there are other ways to interpret this situation, if you wish to attribute the velocity changes to the physical characteristics of the ball. Available evidence suggests the ball may have changed towards the middle of 2015 and into 2016, leading to more home runs. But higher exit velocity? I’m not sure.  And did the ball change for 2017 to remove this? I’m not sure. And what is the normal year to year variance between baseballs? I’m not sure.  We don’t have any answers to these questions, so let’s call it normal variance.

Aging Curves

First, I tried to find the aging curve for exit velocity data. I compared year to year changes in velocity, and the data is a bit messy. I split it into high exit velocity batters (>88mph) and low velocity (<88mph). This helped a bit, but not much.

The low exit velocity group seems to gain velocity up until around age 29, then steadily lose it as they age. After around age 33 or 34, they tend to gain velocity again, but this is almost certainly caused by survivor bias, most likely due to injury patterns. Players around this age are more likely to get injured, but still young enough to return to health and put up a solid season afterwards.

The high exit velocity group has a different pattern entirely. They tend to steadily lose velocity at all age ranges, although the rate of loss seems to follow a pattern. There might be some peak around age 25 where their rate of loss is minimized. They seem to lose more velocity as they age, until around age 31. At this point the trend reverses, likely caused by survivor bias.

This chart makes it appear that the high velocity group has their velocity rapidly tank from high EV to low EV as they age, but that isn’t usually what happens. Rather, expect a career long change to be more gradual decline when compared to the low EV group.

This data is from 680 batter-years worth of data. Meaning If a batter played in 2015, 2016, and 2017, it would count as 2 data points. This obviously isn’t a large sample, so the trend might become more pronounced with a larger dataset.

Three Year Exit Velocity Changes In Batters

The chart above depicts the average change in exit velocity between years 1 and 3. I have also calculated year 1 exit velocity and predicted year 3 exit velocity, plus year 1 exit velocity versus actual year 3 exit velocity. So let’s see how certain players trend.

First, Justin Upton. He is a member of the ‘high EV’ group, and he has had a pretty consistent average launch angle making him an easy player to compare. In 2015 he was 28 years old, and in 2017 he is 31. From the chart above you can see the average dip in exit velocity between 28 and 31 is just under 1 mph, .93 to be exact. Using 2015, Upton’s predicted exit velocity in 2017 is 89.7 mph when factoring in the small change in launch angle. A 0.8 mph difference, right about what you’d expect given his age. In reality, Justin Upton has fallen from 90.5 to 88.5, twice the loss you’d expect given his average launch angle and aging curve.  Hmm.

However, a 2 mph drop in exit velocity is within a standard deviation, so it isn’t a particularly interesting example. I picked Upton to show you the general sort of process I’m going through here. On average, there is a .33 mph difference between the actual 3 year change in exit velocity and that predicted by the age chart I created above, and the standard deviation is 1.7. For the purposes today, I’m going to throw out anyone within 1.5 standard deviations.  And, hey, I’ll limit them to a minimum of 200 PA, too. That should make the list small enough to fit in a table.

 

Okay, so this chart has all sorts of players worth talking about.

First off, Stanton and Trout are both in the high exit velocity camp, so you should expect a steady velocity decline over time. In addition to this, Trout has raised his launch angle considerably, adding power at the sacrifice of exit velocity. On the flip side, Stanton appears to be making a conscious effort to cut down on his bat speed in order to increase bat control. Both of them have intentionally sacrificed exit velocity, which may explain why high velocity batters tend to lose velocity over time. It could be part of their maturation.

Next you have Marte, Jaso, and Piscotty who each appear to have left the ‘high exit velocity’ group of batters now that their average batted ball is well below 88 mph. Jaso has dramatically increased his launch angle, which could explain his losses.

Then you have exit velocity gainers, most notably Jose Ramirez.  Ramirez has not only increased his exit velocity but taken his whole game to another level this year, becoming one of the most valuable fantasy players. Khris Davis, Addison Russell, and Avisail Garcia are all in the same camp, with Khris Davis representing the most extreme case. Davis has gained at least one mile per hour each of these three seasons, while simultaneously increasing his launch angle. This points towards an increase in that ‘fudge factor’ I referenced earlier. Plate discipline? Bat speed? Maybe a little of each? Maybe something else? Whatever flavor the fudge may be, Khris Davis has gone back for seconds.

Of all of these names, though, none of them are particularly interesting ‘bust’ cases going into next season. Stanton and Trout are the most interesting on the down side, but their losses can be explained.  Let’s dig a little deeper to see if we can find a real question mark.

 

These batters are between 1 and 1.5 standard deviations below the mean. Neil Walker is a solid question mark, but he’s been recovering from back injury and he isn’t exactly the most fantasy relevant guy out there. Inciarte is a glove first outfielder. Lucroy’s problems are already well known. On down the line, not many of these are particularly interesting.

Except, maybe, Corey Dickerson. Dickerson has fallen from the high EV to low EV group, without obvious injury concerns. However, Dickerson is suffering from a pronounced plate discipline problem. He has the second most out of zone swings to date, behind Salvador Perez, and he doesn’t have the out of zone contact rates to make up for it. Dickerson is chasing bad pitches and missing them. Even when you do make contact with an out of zone pitch, your exit velocity suffers.

Dickerson hasn’t always had this out of zone approach. In 2015 he swung at far fewer out of zone pitches and his exit velocity was better off for it. His batting average and slugging were higher, too, but he also played in Coor’s Field so we’ll toss that aside for the moment. Dickerson is having a good season, but this is a very worrying trend.

Do you see any other players on these lists that you find interesting? Are you willing to shrug off some of the down years Miguel Cabrera, Lucroy, and Carlos Gonzalez are having?  Or are you going to be more cautious? Personally, outside of the notable exceptions, I will avoid each of these players.