Explore the inner workings of any major-league front office or baseball think tank, and you'll find legions of sharp minds, hard at work trying to unearth the next big breakthrough in analytics.
From Branch Rickey and Allan Roth harnessing the power of platoons to Voros McCracken informing us that much of what we believe to be good pitching is actually luck or good defense to the brains at MLB Advanced Media telling us that results on balls in play often aren't actually that random, new rocks get overturned all the time.
The challenge for the rest of us is keeping up. Maybe you're caught up on Wins Above Replacement, Weighted Runs Created Plus, and Defensive Runs Saved. But the wheel of progress never stops spinning, with new concepts emerging so fast, it's tough to keep up.
With that in mind, we asked the smart folks at three leading analytical outlets -- MLBAM, Baseball Prospectus, and Fangraphs -- to help guide us through some of the newest and most exciting fresh work in stats land. Here's what they had to say.
One of the biggest advances in publicly available data over the past few years is the ability to quantify how balls are hit. Thanks to MLBAM's Statcast platform, we can now measure how fast the ball travels off a player's bat (exit velocity), and the angle at which a ball is hit (launch angle).
"The harder you hit the ball, the more success," said Tom Tango, senior data architect for MLBAM. "But you need to reach a threshold for that to happen, close to 95 mph, after which each mph adds a great deal of production. It's better to hit a ball at 70 and 110 than to hit two at 90. Launch angle is interesting because it needs to be paired with exit velocity for it to matter. Launching at 30 degrees is great if you can hit the ball at 100-plus mph, as it has a good chance of being a home run, but terrible at 85-90 mph, as this will be caught in the outfield, but good at 70-75 mph as it will drop between the infielders and outfielders."
At the intersection of a fast exit velocity and optimal launch angle is a new stat called barrels. Tango developed barrels as a way to note batted balls that produce a minimum batting average of .500 and a minimum slugging average of 1.500. As Tango notes, hitters only "barrel" the ball 7 percent of the time. But when that 7 percent outcome hits, big things happen for a team's offense.
Fangraphs senior editor Jeff Sullivan said he's most interested in stats that offer predictive value, or tell you something you might not already know. For instance, Expected Weighted On-Base Average (xwOBA) takes batted-ball data like exit velocity and launch angle, then boils down all those batted balls into a number that runs along the same scale as plain, old, on-base percentage. The difference between xwOBA and wOBA is that the expected number strips out what the defense did and just considers the quality of contact by the hitter. So if a batter smoked a bunch of screamers that happened to land in fielders' gloves, xwOBA will reflect that.
As an example, Red Sox outfielder Jackie Bradley Jr. posted one of the biggest gaps between his actual production and his expected production in 2018 (.311 wOBA vs. .341 xwOBA). Why? Because despite ranking among the league leaders in exit velocity and frequency of hard contact, bad luck resulted in a lot of hot smashes off his bat going for outs.
On the counter-intuitive front, Sullivan likes to sift through metrics like Statcast's Sprint Speed. Designed to measure how fast a player runs once he's reached full acceleration, Sprint Speed recently told Sullivan something he never would have expected.
"Avisail Garcia is apparently fast!" exclaimed Sullivan. "I never would have guessed. He's a big boy but he runs at about 90th-percentile speed. I would have thought he had, like, Wilmer Flores speed."
In other words, if the Rays didn't already get a colossal bargain by nabbing a player just a year removed from a massive .330/.380/.506 season for just a one-year, $3.5 million contract, here's another little data point which backs up Garcia being a potential heist for Tampa Bay. (Though we're still bummed that Garcia's signing came at the expense of the affable nomad, Oliver Drake.)
One of the most prolific purveyors of cutting-edge stats is Baseball Prospectus. Top researchers like Jonathan Judge and Harry Pavlidis have developed a passel of metrics to help fans better understand what's happening on the field. To help navigate its extensive menu of sortable stats, Prospectus hired Bryan Grosnick to be its director of reference and training.
The biggest item on Grosnick's plate at the moment is helping readers understand a new Judge creation, Deserved Runs Created Plus (DRC+). The goal of DRC+ is to tell you how well a hitter performed relative to his peers. So a hitter's score of 100 is league average, 150 is terrific (50 percent better than league average) and 50 is terrible. In 2018, your typical average hitter by DRC+ would be superstar-turned-oft-injured third baseman Josh Donaldson (100 DRC+). A great hitter would be out-of-nowhere Dodgers slugger Max Muncy (150 DRC+). A terrible hitter would be Cleveland catcher Roberto Perez (53 DRC+).
If you were following along at the beginning, or if you've familiarized yourself with some older-generation advanced stats, you know about Weighted Runs Created Plus (wRC+). Baseball Prospectus touts DRC+ as a better indicator of what happened on the field, for a bunch of reasons.
To lend it more predictive value, DRC+ uses a "mixed-model approach," said Grosnick. "It takes everything into account -- park effects, temperature, handedness of pitchers, ballparks changing, the balls themselves changing."
By adjusting to changes in the game in real time, DRC+ can then make apples-to-apples comparisons to past players and past seasons easier.
"As numbers change, doing everything as an index makes it easier to understand," Grosnick said. "Take home runs. Hitting 50 in a season used to be a huge deal, then lots of guys were able to do it, now maybe one guy might do it. By indexing offense [as a whole], we can compare across years" without getting stuck on changing league norms."
You'll find that mixed-model approach in other Prospectus stats too. Deserved Run Average (DRA) runs along a similar scare to the number of runs a pitcher allowed per nine innings. But like DRC+, it adjusts for contextual factors that ERA and even advanced stats like Fielding Independent Pitching (FIP) do not. With DRC+, if a hitter ran into the top of Houston's stacked rotation more than some of his peers, the final number would reflect that.
And with DRA, "if James Paxton got stuck facing the Yankees and Red Sox on the road two weeks in a row, he doesn't get dinged as much" for giving up a bunch of runs, said Grosnick.
On the catching front, BP's mixed-model approach loops pitch-blocking, framing, and throwing into one catch-all stat, Fielding Runs Above Average. So if you're evaluating, say, the one-year, $18.25 million deal that Yasmani Grandal got from the Brewers, you wouldn't focus on one terrible playoff game, or even on Grandal's sometimes iffy ability to prevent balls from scooting to the backstop. Instead, you'd note all of his defensive attributes, which marked Grandal as the best overall receiver in the game last season.
But wait, there's more. Want to measure anything from the number of 22-year-olds who've ever topped Bryce Harper's 2015 season (one, Ted Williams) to more arcane information dating back to the 19th century? Baseball-Reference's Play Index becomes indispensable. Want to dig into baseball's endless array of splits, then rank players against each other using all of those splits? Fangraphs' splits leaderboard goes back 17 years, and you could spend days flipping through it, answering hundreds of splits-related questions.
And still more. For a granular look at one play's result, there's Catch Probability, which can instantly tell you if a seemingly spectacular grab was as great as it looked, or if the fielder put a little extra mustard on it. Coming soon to Statcast will be Infield Defense, slated for release before the start of the 2019 season.
More broadly, Tango said the pursuit of new sources of knowledge -- both in the public eye and behind front offices' closed doors -- boils down to two key goals.
"The first is to break things down into components," he said. "While a Statcast WAR will be an eventual byproduct, and it may serve as setting a target, the true objectives are all the components that we can measure and create metrics for that would go into such a metric. WAR is the culmination, but Barrels and Catch Probability and Sprint Speed are the true advances.
"The second is to answer the question: 'How?' How did that play result in a hit? How did that lead to an out? It's not about trying to put things into some sort of final probability, though that's a necessary step. It's how to identify and value all of the components that would lead to the thing we actually saw. We saw the hit. We don't want to know in some context some number like 72 percent. We want to know the 72 percent and the 12 percent and the 16 percent that together gives us 100 percent. A hit happened. How? That's what we'll answer."
