Firstly let’s start off with the fact that I am not a plant, stooge or shill for MLB, Statcast or Baseball Savant (I would like to be, but I am not). But secondly, I need to tell you all that if you want to learn more about MLB, especially the statistical side, you should be using Baseball Savant regularly.
Off the back of their January big drop, infield Outs Above Average (OAA), which I have covered in a few articles already, they dropped a new update on 3 February, which allows us to see how infielders have performed for each hitter and pitcher individually, and for the hitting teams combined.
You can click through to the same breakdown pages as before, but you can now filter by pitcher and batter, in addition to the previous fielder option. Here is Marcus Semien’s page as a batter.
This means you can now see which pitchers benefited most from their infield defence and which hitters got unlucky. But that isn’t all we can do. With some of this new info, we can see how it impacts some long-standing sabermetric questions about soft contact pitchers, BABIP & shifting.
The lucky ones – Pitcher
Above you have the pitchers who were helped most and least by the defence behind them. What this is saying is that based on the positioning of the defence, the quality of contact and the speed of the batter, Dakota Hudson would have given up 16 more hits if he had an average defence behind him. Michael Pineda would have allowed eight fewer hits.
A rough estimate for outs-to-runs is 0.75-0.80 runs per out. That means that Hudson conceded roughly 13 runs fewer than he would have with an average defence. Him being top seems to match well with other metrics. He had the largest difference between ERA and FIP of any qualified pitcher last season (-1.58 difference between 3.35 ERA and 4.93 FIP).
You could say that Hudson is lucky and Pineda is unlucky, and you would be partially correct, but it isn’t the whole truth as the OAA isn’t the whole picture. There is another part which could be described as luck, and this is the estimated success rate (ESR) of the infield defenders on his plays.
Pitching BABIP, wOBACON, ESR – Luck or Skill
Based on the research of Voros McCracken and others, the current understanding is that a pitcher’s BABIP is mostly a function of a pitcher’s defence and luck, rather than persistent skill. Thus, pitchers with abnormally high or low BABIPs are good bets to see their performances regress to the mean.
Above is the year-on-year BABIP for the last five seasons. This shows the work of McCracken still holds true today. But today we also have more data about pitchers, and we hear discussion about players who can pitch to soft contact. There is some data which shows that a pitcher with a lower average exit velocity would have a lower wOBACON (wOBA on contact) and average exit velocity is slightly more predictive. But once again, the year-on-year trends show no predictability. So luck not skill.
So both BABIP and wOBACON are luck and not pitching skill, so to me, that means that we should expect a pitcher’s estimated success rate (ESR) on infield plays to be the same, that is, luck and not skill. With three seasons of available data, this looks to be the case.
Less data is available, and the data is rounded to a full percentage point, but it does seem to show that a pitcher’s seasonal ESR is once again luck and not skill. With this, we can determine each pitcher’s “luck” on infield hits by comparing the success rate of their infield when they are pitching, to the league’s infield actual success rate (88%). Any difference is the combined luck of defensive skill, positioning or randomness. I will call this Outs Above League Average (OALA).
Hudson still tops the list for good luck, but Jon Lester leaps into the lead for worst luck. One thing to add here is that these negative OAA for pitchers may or may not be reflected in their stats like ERA, as we don’t know if these missed outs were charged as errors or not. But we do know that it would have impacted their R/9.
The lucky ones – Batters
The OAA figures are always shown with regards to the defence so a hitter having a negative OAA means they have got on base more often than expected. So this would suggest that Shohei Ohtani had the most luck getting on base and Brett Gardner had the worst. But, as I talked about at the end of the pitcher analysis, Ohtani’s -10 OAA doesn’t mean that he got 10 more hits than expected because we need to account for errors, and the data isn’t broken down for that (in these tables).
If you go onto his player page, you can see that Ohtani had five errors on these plays and isn’t getting credited with hits. Most of the top 10 luckiest players had numerous errors on these plays: Arenado (5), Choo (6), Castellanos (9), Alberto (7), Wong (5), Ramirez (3), McNeil (2), Galvis (4) & Bryant (2). So, these players aren’t getting the full amount of OAA as hits, but there is still some potential for regression.
On the opposite side though, you have hitters not getting hits because of infield defenders performing better than average against them. So you could say, Gardner (who had no errors) missed out on nine singles which would have improved his wOBA to .358 (from .344) and given him another 0.6 WAR. That is enough to move him from the 47th best hitter to the 34th.
Batting BABIP, ESR – Luck or Skill
Unlike pitchers’, a batter’s BABIP has an element skill involved. The predictive nature of BABIP gets even stronger if you look at qualified hitters (550+ PA), the year-on-year r-squared is 0.28. This being a skill for batters makes some sense as some hitters can hit the ball harder than others (creating more difficult grounders) and some can run to first faster (decreasing the time for defensive plays to be completed).
So, is the estimated success rate (ESR) on a hitter’s ground-balls predictive?
The answer looks like yes. Over the small amount of data we have for the last three seasons, we see some year-on-year trend which suggests that some skill is involved. That is to be expected as one of the key elements in the calculation of OAA is a player’s average sprint speed, but it doesn’t include all grounder plays, as when the ESR is 0% the play is excluded. If we were to compare a player’s ESR to their average sprint speed, we would expect to see a trend.
Which we do, it shows a reasonable correlation between the two. But there is still some random variation down to the location of the defenders, the push/pull angle of the bat and other factors. But something really interesting happens when you split this for handedness.
The correlation is much lower for left-handed hitters (LHH) than it is for right-handed hitters (RHH). I have ignored switch hitters here as I cannot get the info for their ESR split between their handedness. And what do we know is different between left-handers and right-handers when it comes to infield defence? The shift!
Shifting Expectations
In 2019, 41.9% of LHH plate appearances were shifted while just 14.3% of RHH were. Sadly the OAA data isn’t currently split into shifts and non-shift at-bats, so we cannot look to see if a player’s estimated success rate is different in these two scenarios. But if we account for the overall shift percentage of a hitter as well as their sprint speed, can we predict their ESR better than using just sprint speed?
There is as least one other thing which may impact the ESR – the exit velocity of the grounders. I will build some models that will account for that as well, by including each player’s average EV on grounders. To do this, I built some very simple linear regression models, for each handedness and include or exclude these three inputs. This is for hitters with more than 100 infield OAA plays per season for each of the last three seasons.
Before I go into the results, I wanted to say the correlation between exit velocity and ESR is positive (very minimal but still positive), which suggests that it is either easier to get the out when someone hits the ball faster or has no impact. That seems counter-intuitive but starts to make sense when you remember the OAA metrics and calculations don’t include plays where the chance of making the play was 0%.
As we are only talking about plays which have a chance of being made, what this is saying is that the difficulty of fielding a harder-hit ball is countered by having longer to get the ball to first. And maybe countered so much that it is easier to get the out on a harder hit ball.
These models show clear differences between for LHH and RHH. Average exit velocity and overall shift percentage give little added value to the sprint speed RHH model. But for LHH, accounting for both definitely improves on the sprint speed LHH model, but still doesn’t bring it close to the sprint speed RHH model.
This is a very simplistic view of this data, and shifting as a whole, but it does seem to suggest that infielders may have had easier chances of getting left-handed hitters out when they were shifted more often. Which suggests from a fielding and positioning perspective, the shift might be working.
There is a large number of caveats for all of this work as we are just looking at high-level data and not individual plays, but this is something worth monitoring, especially if Baseball Savant keeps adding to what is available.
Photo courtesy of Dilip Vishwanat/Getty Images
Russell will shortly be presenting research at the 2020 SABR Analytics Conference in Arizona. Make sure you’re following Russell on Twitter @REassom
Have you ever wanted to write about baseball? Do you want to tell the story about your baseball fandom journey? How about a favourite player? Or, just get silly and write something to make people laugh? Slide into our DMs, contribute now. We’ve never rejected a post.
Make sure you subscribe to the Bat Flips and Nerds podcasts and follow us on Twitter @BatFlips_Nerds. News, views and interviews, all with a British twist.