So what is PAVE, where did it come from and why should you use this new stat? It all starts with the quality of a hit. Sure, BABIP tells us how many batted balls fall for a hit, but why do star players carry a higher mark? Why do Miguel Cabrera and Robinson Cano have BABIPs of .346 and .324 while Conor Gillaspie and Dustin Ackley check in at .294 and .288? Besides the obvious talent disparity, Cabrera and Cano hit “better balls.” What constitutes a “better hit ball?” It’s actually quite simple. Line drives are the best quality, for obvious reasons, and fall for a hit at a significantly high percentage. Ground balls and fly balls are much poorer and rarely equal a hit. Going back a decade, we find that 71 percent of line drives fall for a hit (league average). Meanwhile, ground balls become hits just 25 percent of the time, and fly balls only 21 percent. You may see some differing percentages on fly balls if you search on your own, but that is because some sites separate fly balls, pop flys and infield fly balls (IFFB). We’re actually going to cover the IFFB issue in a bit.
Knowing the league averages, we can say if a player hits 100 line drives, 100 ground balls and 100 fly balls that his average should be .390 (71 LD hits, 25 GB, 21 FB in 300 ABs). Therefore, we have an expected level of production. However, we need to circle back to the types of fly balls while also accounting for speedsters. For IFFB, those have a zero percent chance of falling for a hit. Even when they hit the ground, it’s because of an error (see: miscommunication as three guys stare at each other like it was the other’s fault). With ground balls, we also have IFH (infield hits), which have a 100 percent success rate thanks to a hitter’s speed. Therefore, we need to account for IFFB never being hits and IFH always resulting in a hit, and the same goes for home runs, as those never result in an out… obviously.
We started with LD*.71 + GB*.25 + FB*.21, and now we have the following:
LD*.71 + (GB – IFH)*.25 + (FB – HR)*.21 + HRs + IFH – IFFB = PAVE
Make sense? Basically, we’re accounting for the success rate of all GB and FB that aren’t automatic outs or hits, then adding or subtracting the auto-results back in to get our PAVE.
So how does this play out when we look back over the past decade? Taking ever hitter with at least 400 PAs gives us 2,112 points of data, or PAVE results. The resulting r-squared is .505, which is just above moderate and into strong relation territory. Strong runs from -0.5 to -1.0 and 0.5 to 1.0. Moderate is +/-0.3 to 0.5. For those not familiar with r-squared, it basically tells you how much correlation there is between two sets of data. In this case, it’s PAVE and players’ true batting averages. To have a r2 over 0.5 for this many players with such a variable statistic is quite good, and it tells us that we can definitely use PAVE to predict future success or failure. It’s not perfect, no predictive stat is just as with BABIP, but it is definitely an added metric that can help you in evaluating players for your Fantasy Baseball team, even in your DFS lineups.
Of course, you want to see some PAVE numbers for yourself, so here you go (from 2014).
Looking at the biggest outliers, Giancarlo Stanton and J.D. Martinez, we have one player that might make you question PAVE (Stanton) and another that makes you question the player (Martinez). With Stanton, he is a career .271 hitter, only five points (.005) over his 2014 PAVE. In addition, his .266 PAVE is seen in his .274 second half average. Stanton was hitting above his norm in the first half (.295) and hit .226, .289 and .231 over the last three months. Stanton actually had a better home run rate in the second half (11.6 ABs per HR versus 16.8 per in the first half), but his average dropped thanks to his normalized luck and LD%. For Martinez, he too saw an average dip in the second half, going from .346 to .292. Martinez was carrying a 22.8 HR/FB percentage in the first half, which fell to 16.4 and is more reasonable for Martinez’s ability. After all, only two players had a higher HR/FB mark than Martinez’s first half 22.8: Jose Abreu (26.9) and Stanton (25.5).
Let’s check out one more group before slapping a bowtie on this.
Carlos Gomez and Danny Santana immediately jump out at you. Gomez has more years of hitting for a mediocre average than good. His PAVE would indicate his last two years are a bit above his hit quality, but Gomez could be a late bloomer in that aspect as well. Nevertheless, like Stanton and Martinez before him, Gomez saw a dip in the second half. His average dropped from .304 to .253. Gomez’s big problem wasn’t his home run rate, but that he dropped nearly 10 percentage points for line drives and jumped nearly eight in ground balls – both of which moved close to his career averages. With Santana, we all know about his insane (and unrepeatable) BABIP of .405. Unlike everyone else, Santana didn’t see a significant second half decline (.314 AVG versus .328), but when your BABIP is .405 and Jose Altuve’s is .360, you know you’re as lucky as they come! For reference, Altuve hit .341, Santana carried that .319 AVG even with a far superior BABIP. Again, instead of figuring out what that BABIP means though, you can look at Santana’s PAVE and know that the luck is going to end… it just is going to happen this year instead of during last year.
And that’s the final point with this. Like wOBA, BABIP, etc., PAVE isn’t a perfect stat or infallible. If it was, the r2 would have been 1.00. However, PAVE is a great new metric to help you predict future success by making you dig deeper for reasons behind a PAVE-to-AVG difference. Is it a crazy HR/FB rate? Has the player started hitting more line drives? Does Santana have a rabbit’s foot around his ankle? All of these factors matter in evaluating a hitter’s success, and it starts by PAVE-ing a new outlook on performance.
Oh, and don’t worry. You won’t have to calculate this for yourself. Until I can get Fangraphs to add it *wink, wink* I’ll give you updates weekly and highlight the significant outliers worthy of your attention.