The Great American GPP: The Home Run Concordance: Exit Velocity and Launch Angle
In tribute to the excellent new Rotogrinder’s video series on Advanced Stats by CheeseisGood and JMtoWin, and the discussions the videos have generated, in this post I decided to jump from applying ideas developed in rhetoric to ideas developed in R). (one of those I know how to use; one I do not—I don’t think you’ll have any trouble figuring out which is which :) ).
Partly, this new discussion focuses using advanced stats, and how important it is to remember when using them that they are all are not created equally, and that two players with the same FB/HR%could be very different hitters, with different outcomes when they step to the plate, and most important for us, that even though the two hitters have the same ratio, they have very different probabilities of hitting a home run each night. Two players may have the same wOBA, but for very different reasons, and only by understanding what went into creating that number for the player, will we get the most effective use out of that stat.
By now we’ve all learned the importance of splits for really getting the most value and information out of the standard baseball stats, and even for some of the more commonly accepted “advanced” stats like wOBA, or ISO splits, or batted ball splits. With the normal stats, and wOBA and ISO, It seems as if the general idea is breaking down the stats to determine if the value of the player, or the odds of the play, are as accurate as possible, and to take advantage of situations when we discover an error in one direction or the other. In this sense, reading the player cards is like reading the Daily Racing Form, hoping to find a situation where our analysis of the past performance of a horse finds the horse to have a certain value, while the public has determined the horse to have a different value, and to profit by exploiting (betting) that difference.
However, it seems that with the majority of the advanced stats, the key lies not in splitting them apart (this does not mean we don’t want to know what went into making them—again see the video) rather the greatest value comes from finding a correlation among the stats. With the advanced stats, it seems as if the idea is to find as many different types of stats—from the different fields measured or methods of measurement, which fit together to craft a 3D model of the player, which simulates the possibility of various outcomes as accurately as possible
Since the K and the HR are the two most valuable DFS MLB events, it may be profitable to find as many stats as possible to correlate together to get the best sense of a batter’s probability of hitting a home run. In that hope, the table below presents some advanced stat data, and ideas on how we can use it to see how it aligns with other data sets for modeling HR probability for hitters.
The following data is from MLB’s BaseballSavant on June 3, 2016. The data updates every day at 3am, so you will find different totals when you visit, and I’ve grouped some of the EV #‘s into buckets.
|Exit Velocity||# BBE||#HR||HR%|
*Exit Velocity clocks the speed of the ball off the bat, of any batted ball event ; it is part of a batter’s skill-set, just as limiting hard contact, or lowering Exit Velocity Against is considered one of a pitcher’s skills. Although a batter can benefit from a low exit velocity on occasion (the donut hole and speed are different discussion), the greater the exit velocity the greater the chance of home run event.
BBE is the number of batted ball events at that exit velocity; *HR is the number of Home Runs hit at that exit velocity bucket.
*HR% is that number in percentile. *wOBA*can also be found at the link, but I’ve left off for now.
Obviously we can see that the Exit Velocities that result in the most home runs. If we just wanted to quickly use this data as part of each slate’s research, we can examine a batter’s overall current average exit velocity average (aEV) to get a sense of which EV buckets they are hitting the most, as an extra bit of information to gauge a player’s probability of going yard or not. With a few more seconds, we could look at the batters most recent games, and see the EV for each at bat…are they trending faster or slower, or do they seem to be locked into a set EV range, hopefully one made up of high Home Run velocity of the bat. This should not replace your current model for estimating Home Run probability; instead we can use this information to supplement it. Perhaps two players seem equally likely to go yard in their match up, but one has been averaging a higher exit velocity for the past week, making that batter the slightly better play. As another quick plug in to current research, add in the current exit velocity allowed by the slate’s pitchers…perhaps you’ll find a spot that other gamers are missing.
So there’s a quick way to plug in another stat which hopefully aligns with your reading of the day’s offense. Remember, the hope is that with advanced stats we find correlation among that stats and methods which seem to indicate the same outcomes. In my next post, we’ll talk about more time intensive applications of this data, and other methods for using it to model hitter, and pitcher performance, as we’ve quite literally only scratched the surface.
Good luck grinding!