How Statcast Will Impact How We Value and Describe Players

1 of 2

Oct 22, 2014; Kansas City, MO, USA; MLB newly elected commissioner Rob Manfred speaks at a press conference before game two of the 2014 World Series between the Kansas City Royals and the San Francisco Giants at Kauffman Stadium. Mandatory Credit: Christopher Hanewinckel-USA TODAY Sports

Last Friday at the Sloan Sports Analytics Conference, MLB Network’s Brian Kenny interviewed new MLB Commissioner Rob Manfred. The two discussed a wide variety of topics, but perhaps the most exciting one was StatCast, Major League Baseball Advanced Media’s flashy and impressive player tracking system that debuted last year.

StatCast measures things like the angle and velocity with which the ball leaves a player’s bat, how quickly fielders react and what route they take to the ball, and how big of a lead a runner takes off a base. MLBAM has made some impressive videos that display the technology’s ability, but we haven’t heard much about how much data will be available for public analysis.

On Friday, Manfred indicated that StatCast data will be available on MLB.com and the MLB At Bat mobile apps, with some raw data becoming available to analysts at some point. While it’s still unclear how granular that data will be, we have another chance to wonder how this new data will impact how we look at the game.

I like to break down stats into two categories: valuation, and description. Valuation refers to any stat that tries to put an actual win or run value on a player’s performance. Stats like Wins Above Replacement, Batting Runs Above Average, Defensive Runs Saved, and Win Probability Added all fall under this umbrella. They are important because they are designed to estimate how many runs or wins a player added to his team. A larger positive total is always better, since more runs (or fewer runs when looking at something like Defensive Runs Saved) gives the player’s team a chance to win more games. Most valuation stats combine multiple categories: WAR combines offense, defense, base running, and playing time, while Defensive Runs Saved combines fielding plays and throwing plays.

Descriptive stats on the other hand, simply help us understand how a player executes their game plan. Swing rate, contact rate, even on base percentage are all descriptive stats. While these stats often correlate strongly with value, only tell a part of the story. A player with a .300 OBP may be a more valuable offensive player than a player with a .330 OBP if the first guy hits more double and home runs. Strikeout rate is a classic example of this, with many valuable hitters nowadays striking out often while still being valuable. While these stats have limited use cases, they are still very important for being able to fully understand how a player creates runs and wins, and they help us see what parts of his game are doing well or poorly when his value is above or below what we expect.

With StatCast data, we’ll be able to create better valuation statistics AND better descriptive statistics. The actual mix depends on the type and quality of the raw data released to the public, but one can begin to imagine what the impact will be as long as there is some data made available.

The most common topic that arises regarding new stats is always defensive metrics. Stats like Ultimate Zone Rating (UZR) and the aforementioned Defensive Runs Saved (DRS) use video scouts to estimate each batted ball’s trajectory and location in order to come up with an estimate of the likelihood that a batted ball will turn into an out. This method works rather well for putting players into tiers after gathering decent samples (usually three or more years is considered relatively reliable), but always causes controversy when an outlier arises and is rated as 20-30 runs better than a league average player at that position. While some of this is simply people not liking things that go against their instincts, there is a significant margin of error on these numbers due to the inexact method with which the play probabilities are created.

Next: Will StatCast Fix These Issues?

Sep 23, 2014; Cleveland, OH, USA; Cleveland Indians left fielder Michael Brantley (23) reaches for the ball on a two-RBI double by Kansas City Royals catcher Salvador Perez (not pictured) in the fifth inning at Progressive Field. Mandatory Credit: David Richard-USA TODAY Sports

Will StatCast fix all of these issues? No, it won’t. First of all: there will always be people who refuse to accept numbers that disagree with what they see. The next issue is that when dealing with fielding, we will always be dealing with a small sample of players that are actually contested. Many plays are either easy outs or hits that are impossible to catch. Let’s use Michael Brantley as an example. Inside Edge has video scouts assign an out probability for each batted ball, in buckets from 0-1%, 1-10%, 10-40%, 40-60%, 60-90%, and 90-100%. Of the 338 batted balls hit into Brantley’s vicinity in 2014 (at both center field and left field), only 28 of those balls (8.3%) had an out probability between 1% and 90%.

Some perspective on how small that sample is: that’s the same as the number of plate appearances Brantley had with a runner on first with second and third open and nobody out. You would certainly not want to judge Brantley’s batting on such a small sample, so it’s not surprising that there may be some room for error in his fielding numbers simply due to the fact that there are not that many players that are genuinely contestable.

More from Cleveland Guardians News

The difficulty of distribution of out probabilities may not change when we have a better estimation of them, but it will lead to more precise valuations due to helping us understand the difference between 90% plays and 99% plays. So while our overall valuations will be better, they will still be heavily limited by the sample size. This means that a fielder who is 10 runs above average by StatCast will still have a very decent chance of having a lower true talent level than a fielder who is 5 runs above average. This is always a possibility, but it’s important to note because valuing defense is such a vital part of WAR.

With that said, while our valuation stats may only improve by a bit, our descriptive stats will become almost unimaginably better. With decent samples, we’ll basically be able to simulate a player’s chances of getting to any ball. All you need to know is: a) the hang time of the ball, b) where the ball lands, c) how quickly the fielder reacts to such a hit, d) how fast they can cover the necessary distance (taking into account initial positioning and route efficiency), and e) their chance of actually catching the ball once they get to its location. When you have all these inputs, you can then model how a fielder would theoretically perform when given a certain distribution of chances. The key to the accuracy of that model would depend on how granular the descriptive stats created and distributed publicly by StatCast will be.

The cool thing about this is that it skips the attempt to value past performance and goes directly to true talent. You’ll be able to simulate how valuable a fielder would be when playing behind different pitchers, when playing in different positions, and playing with different fielders. For example, it would make it much easier for the Kansas City Royals to determine the value of putting Jarrod Dyson into the game as a defensive replacement in the late innings, while adjusting for the pitcher, ballpark, other outfielders, and the hitters that will be batting.

This disconnect between true talent and the value of performance is going to be the most noticeable effect of StatCast’s release. I only discussed defensive metrics in this post, but hitting stats will be impacted almost as greatly. With the knowledge of how hard a player hits the ball and the angle of contact, we’ll get a much better sense of when a player has gotten lucky hits. While those lucky hits certainly led to runs and wins, they’ll look the same as outs when packaged into a model meant to estimate the player’s true talent at the plate. Again, this is going to bother some people and excite other people, but more than anything it will lead to more questions and more descriptive answers.

We may never be able to put a precise and accurate valuation on a fielder’s performance over one season. But with StatCast, we should be able to analyze their performance and understand their underlying talent in a much more interesting and evidence-based manner. All we can do now is wait for the season to start and hope that MLBAM releases granular data to the public.

Home / Cleveland Guardians News

How Will StatCast Impact The Way We Value and Describe Players?

Schedule