Hitter xwOBA Underperformers — May 3, 2021, A Review

Today, I continue my in-season metric lists with a review of the xwOBA underperformers through May 1. We know that xwOBA isn’t a perfect metric of what a hitter “deserved”, nor is it meant to be predictive. However, it’s certainly better to use it than wOBA itself when trying to forecast rest of season performance. This is especially true when only a small sample of games are in the books, so expected metrics play a more valuable role. So let’s find out how this group of hitters performed over the rest of the season.

xwOBA Underperformers
Name ISO – Through May 1 BABIP – Through May 1 wOBA – Through May 1 xwOBA – Through May 1 wOBA – RoS wOBA Diff
Dominic Smith 0.111 0.275 0.248 0.385 0.296 0.048
Kyle Tucker 0.184 0.158 0.253 0.367 0.415 0.162
Cesar Hernandez 0.066 0.211 0.248 0.352 0.313 0.065
Albert Pujols 0.200 0.164 0.278 0.379 0.313 0.035
Avisail Garcia 0.128 0.263 0.284 0.374 0.361 0.077
Bryce Harper 0.295 0.358 0.448 0.535 0.428 -0.020
Jorge Polanco 0.077 0.237 0.248 0.334 0.368 0.120
Rafael Devers 0.292 0.318 0.403 0.487 0.368 -0.035
Willy Adames 0.128 0.236 0.225 0.304 0.374 0.149
Freddie Freeman 0.245 0.200 0.353 0.431 0.385 0.032
Francisco Lindor 0.052 0.197 0.247 0.325 0.333 0.086

Out of the 11 hitters on this list, nine of them improved their wOBA marks over the rest of the season. The two that failed to do so were already sitting pretty with wOBA marks above .400, so expecting them to enjoy even better results over the rest of the season, regardless of their xwOBA, would have been unrealistic. Four hitters actually posted a rest of season wOBA mark above their early season xwOBA marks. All the wOBA improvements were driven by a combination of both increased ISO and BABIP marks.

After his 2019 and 2020 performances, albeit over small samples, Dominic Smith looked like an obvious buy low candidate through the first month of the season. Unfortunately, if you bought low, you still ended up with a poor performer, though less poor than he had been that first month. Through May 1, Smith mostly suffered through a power outage, but his ISO barely increased over the rest of the season. His ISO finished at an embarrassingly low .119, driven by a sub-10% HR/FB rate, both of which were career worsts and well below even the most pessimistic of projections. The Mets recent spending spree guarantees he’s out of a job, and only the adoption of the DH will give him a potential chance at significant playing time.

Kyle Tucker was one of the few who went bonkers after the first month, even outperforming his impressive wOBA over the rest of the season. He was a very clear buy low candidate, or at least attempt at a discounted purchase given his absurdly low BABIP. Over the rest of the season, both his ISO and BABIP surged and he finished right where his owners hoped he would when they drafted him. The improved strikeout rate is exciting, so we’ll have to see if that sticks.

After a poor first month Cesar Hernandez transformed himself into a complete different hitter. Suddenly, he was a power hitter with a below average BABIP, as he set career bests in both ISO and HR/FB rate, while posting a BABIP below .313 for the first time, as that mark slid all the way down to just .266. The steals are gone, so he’ll need to keep hitting for power to deliver any fantasy value as he joins his new team.

Albert Pujols has underperformed his xwOBA every season of the Statcast era, and he highlights one of the missing ingredients of the metric. He pulls grounders into the shift at a significantly higher rate than the average right-hander, which kills his BABIP, but isn’t accounted for in the xwOBA equation. Still, he did improve over the rest of the season, thanks to a BABIP rebound from well below .200, but still fell short of his full season xwOBA.

Just once in seven seasons of the Statcast era has Avisail Garcia met or exceeded his xwOBA, though I’d have to dive into some deeper metrics to try to explain why. Still, he came pretty close to matching his first month xwOBA over the rest of the season, as his ISO nearly doubled, and his BABIP increased. Given the difference in home park, I’m not particularly enthused about him as a member of the Marlins.

LOL at Bryce Harper’s .535 xwOBA through May 1! He still managed to post an elite .448 wOBA over that first month, so no one in their right mind would expect better over the rest of the season. It’s a testament to his year that he didn’t regress all that much, posting a still-elite .428 wOBA the rest of the way. What’s amazing is that his ISO increased and BABIP remained stable, so I am guessing some combination of a slightly lower walk rate and higher strikeout rate is what caused his actual wOBA to decline.

Jorge Polanco made for one of the best buy lows on the list as one of just three who increased their wOBA marks by over .100 points over the rest of the season. He more than tripled his first month ISO over the rest of the season as his HR/FB rate jumped into double digits for the first time and he knocked a career high 33 homers. He’s just 28, so we can’t automatically call this a fluke.

Like Harper, Rafael Devers was already enjoying a fantastic season, so you couldn’t expect him to be even better the rest of the way. Most of his wOBA slide was driven by a drop in ISO, but he still managed a career best mark for the season and posted his first HR/FB rate above 20%.

Willy Adames was likely dropped in many leagues after his weak first month, but he nearly doubled his ISO over the rest of the season and his BABIP went through the roof. He ended up finishing with an ISO barely below his career best last year, while his BABIP ended up in a normal range. The good news is his 2020 short season strikeout rate spike proved to be a small sample fluke. A full season in a much more hitter friendly park in Milwaukee should help him maintain this level of performance, but his power will be dependent on where his FB% goes, as that spiked to just over 40% after being stable around 31% in his first three seasons.

Freddie Freeman with a .200 BABIP?! That could never last. His BABIP rebounded back to normal elite levels over the rest of the season, but it was somewhat offset by a drop in ISO to just under .200. Some of that was due to a decline in FB%. At age 32 now, his projection is up in the air until he signs.

Not only was Francisco Lindor a massive disappointment over the first month at the plate, but he also failed to swipe a single base. He did end up as one of the few on this list that posted a rest of season wOBA higher than his first month xwOBA, which is nice, but his counting stats and batting average still fell far short of expectations, ensuring he remained a bust all year, even if he was less busty over the remainder of the season than he was during that first month. The Mets have a new lineup that could be pretty strong, but you have to then wonder if it means Lindor’s steals won’t rebound. That actually might be the biggest question mark.

Mike Podhorzer is the 2015 Fantasy Sports Writers Association Baseball Writer of the Year. He produces player projections using his own forecasting system and is the author of the eBook Projecting X 2.0: How to Forecast Baseball Player Performance, which teaches you how to project players yourself. His projections helped him win the inaugural 2013 Tout Wars mixed draft league. Follow Mike on Twitter @MikePodhorzer and contact him via email.

Newest Most Voted
Inline Feedbacks
View all comments
Joe Wilkeymember
2 years ago

Can we please stop blaming the xStats on Baseball Savant for not accounting for grounding into the shift? Or at least if we’re going to do it, can we actually look to see if it’s true? On Baseball Savant, Pujols had an actual batting average of .196 on 107 GB, with an expected batting average of .207. Do you know what that difference is? Approximately 1.2 hits, which would take his BABIP from .226 to .231, which is basically no difference, and certainly within the range of randomness. He loses twice as many xHits from the difference in his line drives, in less than half as many events. So can we please please please stop this “StatCast doesn’t account for the shift” nonsense when it’s not warranted?

Joe Wilkeymember
2 years ago
Reply to  Mike Podhorzer

I’m arguing that it globally doesn’t make that much of a difference all the time, especially in this case. You are saying that Albert’s BABIP should be lower than his xBABIP because he grounds into the shift. If that’s the case, his xBA on ground balls should be much higher than his actual BA, which it’s not, meaning there must be a different explanation for why his BABIP is lower than his xBABIP.

Albert’s pulled ground balls have an xBA of .223 compared to an actual BA of .111, so it obviously does not account for the shift. However, this is only 63 of his 107 GB. His 44 GB that are straightaway or opposite outperform his xBA by .318 to .183, so the shift both giveth and taketh away. In the end, based on his batted ball distribution, it evens out.

I will be the first to tell you that StatCast’s major shortcoming in the xStats is not accounting for horizontal batted ball direction. But simply stating that the difference is because of the shift without that being the case is inaccurate. There are certain times where that is the case! But this is not one of them.

2 years ago
Reply to  Joe Wilkey

You can line out into the shift as well…

Joe Wilkeymember
2 years ago
Reply to  jemccla

Agreed, but based on how StatCast defines line drives, it also doesn’t have that much of a global effect, or really a local effect. For 2021, the 1173 pulled line drives for RHB into the shift, the xBA is .675. The actual BA is .672, all of three points lower.

Part of the reason for this is less than 28% of the line drives in 2021 as defined by StatCast are hit where a shifted infielder can make a difference. If you look at pulled line drives for all hitters, xBA starts to equal BA at about 13 degrees of launch angle. I included only pulled line drives, because the middle fielder tends to play deeper, making for a higher launch angle to get it over their head. So, theoretically, nearly 3/4 of all pulled line drives should not be affected by the shift.

Furthermore, only about 37% of all line drives were pulled in 2021, which is really not that much more than the theoretical 33/33/33 distribution if you put all line drives in random buckets for “pulled/straight/oppo”. So if you take 28% times 37%, you get about 10%. Obviously this number will vary season to season, but I’m guessing not by much. It also certainly varies player to player. Albert did hit 26 pulled line drives out of 45 total this season (58%), and 10 of those were below 13 degrees (22% of the total).

Which brings me to my last point: a good portion of a line drive’s xValue is generated by exit velocity. I don’t care how well positioned you are, if you have less than one second from contact to when the ball gets to you, if you’re not within a 15 feet of that line drive, you’re not getting it. Even for pulled line drives of less than 13 degrees in 2021, the observed BA of .632 is about 100 points less than the xBA of .741. So if we take that 10/45 of Albert’s LD in the “LD shift zone”, and multiply it by the .109 difference, we get a .024 difference in line drive BA, which accounts for less than half of his .054 difference.

I’m not saying Pujols is going to be a world-beater going forward. All I’m saying is before we cast aspersions about the legitimate shortcomings of the StatCast system, we need to look at whether they apply to the case we are evaluating.

Jonathan Sher
2 years ago
Reply to  jemccla

But that’s not what Mike claimed:

“He pulls GROUNDERS into the shift at a significantly higher rate than the average right-hander, which kills his BABIP, but isn’t accounted for in the xwOBA equation.”

And as Joe Wilkey demonstrates, it make much more sense to analyze the actual data rather than assume hypothesis that should instead be testes with the data. The fact one can line out into the shift doesn’t tell us how often that happens compared to line outs against a standard infield.

Coming up with a hypothesis is only the first step; we must also test that hypothesis against the data.