/cdn.vox-cdn.com/uploads/chorus_image/image/46337552/usa-today-8558841.0.jpg)
In this edition of the sabermetrics column, we answer another commenter's question, which becomes a jumping off point for related hitting metrics. In honor of Yogi Berra: "When you come to a fork in the road, take it."
Question (from Tallahastros): What determines which BABIPs are sustainable? In a case like Altuve’s, how can we tell whether his Babip has increased for good or if it’s luck driven? And could being consistently lucky be considered a skill?
BABIP means batting average on balls in play. BABIP usually requires a very large sample to normalize. For a typical full time hitter, several years of data may be required for his average BABIP to stabilize. As a result, BABIP results often fluctuate considerably, complicating our evaluation of hitters' seasonal results. The vexing question is whether the hitter's BABIP reflects luck or lack of luck, and therefore is likely to regress positively or negatively in the future.
x-BABIP is a formula for "expected BABIP," meaning that it estimates the normal BABIP for that hitter during the period in question. By comparing x-BABIP to actual BABIP, we gain some insight as to whether the batter has been lucky or unlucky.
A number of different formulas have been devised for x-BABIP. Let's look at this recent fangraphs' formulation of x-BABIP based on Baseball Information Service (BIS) batted ball data. I particularly like the new versions of x-BABIP which include the percentage of batted balls classified as "hard hit" by BIS. If we want to identify factors which cause BABIP differences, the x-BABIP formula is informative. The variables included in this formula are: line drive percentage; true fly ball percentage; true infield fly ball percentage; Hard hit percentage; opposite field percentage; and runner speed score. True flyball percentage excludes infield flies, and true infield fly percentage is reflected as percentage of all batted balls. This x-BABIP explains almost 50% of the variation in BABIP. That's not bad, as far as statistical relationships go, but it also means that there may be significant factors we haven't included or even discovered.
What does this say about the Astros hitters performance so far in the very small 2015 sample? Here is the under or over performance by hitter based on Alex Chamberlain's fangraphs article.
2015 x-BABIP | |||
Name | BABIP | xBABIP | Diff |
Chris Carter | 0.222 | 0.290 | -0.068 |
Colby Rasmus | 0.389 | 0.276 | 0.113 |
Evan Gattis | 0.203 | 0.316 | -0.113 |
George Springer | 0.250 | 0.351 | -0.101 |
Jose Altuve | 0.374 | 0.281 | 0.093 |
Luis Valbuena | 0.194 | 0.273 | -0.079 |
Again, keep in mind that this is a small sample so far, which means that it is only suggestive of future positive or negative regression. Carter, Gattis, Springer, and Valbuena have experienced very unlucky BABIP results. If you buy this x-BABIP formula, both Gattis and Springer are due for BABIP corrections in excess of 100 points. Both Rasmus and Altuve have been lucky, in terms of BABIP. But a substantial amount of regression has already occurred in the few days since this x-BABIP was developed. The BABIP for Rasmus and Altuve has already declined by 60 and 30 points, respectively, compared to the data shown above. This seems to confirm the volatility of small sample BABIP.
As implied by Tallahastros' question, Altuve seemingly tends to over-perform expected BABIP. Altuve's 2014 batting championship season, with a .360 BABIP and a .315 x-BABIP, is the main reason that people question the sustainability of Altuve's BABIP. The 2015 x-BABIP, above, for Altuve is .281, but I have my doubts about the result. Altuve's speed score and infield pop up rate is aberrant (compared to his career results) in the small sample so far this year, and that probably distorts his x-BABIP.
Altuve's career BABIP may be a better measure of his normal BABIP. With over 2,000 balls in play, Altuve's career BABIP exceeds the minimum sample required for his average BABIP to stabilize. His career BABIP is .332, which is relatively close to his current 2015 BABIP of .339. But will his career BABIP line continue to grow, given his tendency in the last two years to over perform his career marks?
Looking at retired players with high career rates for BABIP, Derek Jeter (.350) and Rod Carew (.359) may be the most comparable players to Altuve. Batted ball data is not available for Carew. But batted ball data is available for Jeter. Let's compare career batted ball percentages for Jeter and Altuve.
LD% | GB% | FB% | IFFB% | IFH% | BUH% | |
Altuve | 21.60% | 49.40% | 29.10% | 7.20% | 9.50% | 17.60% |
Jeter | 20.10% | 58.40% | 21.50% | 3.00% | 8.00% | 34.60% |
Pull% | Soft% | Med% | Hard% | |
Altuve | 38.00% | 16.60% | 60.00% | 23.50% |
Jeter | 32.80% | 15.10% | 58.20% | 26.70% |
Objectively, Jeter's line probably is somewhat more conducive to a very high BABIP (particularly given the higher groundball rate and lower infield fly rate). Nevertheless, the batted ball profile for Jeter and Altuve is reasonably similar. They are both line drive, spray hitters to all fields, who can hit the ball on the ground, and have enough speed to get their share of infield hits.
Another way of looking at this question: the confidence interval (95% probability) surrounding Altuve's career BABIP is .309 - .351. That's a large span, large enough that it's not all that helpful. But it does tell us that it's possible Altuve could carry a .350 BABIP with a long enough career. Or perhaps we will see some downward regression in the future. We will have to wait and see.
x-ISO
The question, above, leads me to invent another question: "Should I be worried about the power hitters?" Maybe this is a strange question for a team which has led the AL in home runs. But I think you have heard this question in various forms, given Chris Carter's and Evan Gattis' slumps.
The x-BABIP discussion leads us to a x-ISO metric. ISO is isolated power (or SLG% - BA%), and is among the best measures of a hitter's extra base power. x-ISO is devised to estimate a batter's expected ISO for a period, so that we can evaluate whether a hitter's actual ISO for an interval is lucky or unlucky. Fangraphs' Alex Chamberlain (who also developed the x-BABIP above) developed this version of x-BABIP.
This formula is composed of the following variables: percentage hard hit balls; pull percentage; and fly ball percentage. x-ISO explains 63% of the variation in ISO. Power, as measured by ISO or slugging, stabilizes more quickly than BABIP.
A comparison of Astros hitters' actual and x-ISO is shown below.
fb | pull | hard | ISO | xISO | diff | |
Chris Carter | 0.441 | 0.356 | 0.305 | 0.151 | 0.185 | -0.034 |
Evan Gattis | 0.370 | 0.370 | 0.358 | 0.211 | 0.198 | 0.013 |
Colby Rasmus | 0.547 | 0.623 | 0.396 | 0.272 | 0.305 | -0.033 |
Jose Altuve | 0.358 | 0.402 | 0.254 | 0.150 | 0.148 | 0.002 |
Jake Marisnick | 0.324 | 0.355 | 0.250 | 0.198 | 0.129 | 0.069 |
Luis Valbuena | 0.511 | 0.427 | 0.326 | 0.219 | 0.225 | -0.006 |
George Springer | 0.288 | 0.333 | 0.303 | 0.182 | 0.144 | 0.038 |
The hitters who have under performed their x-ISO in their 2015 early season performance: Chris Carter, Colby Rasmus, and Luis Valbuena. The hitters who have over performed their x-ISO so far: Evan Gattis, Jake Marisnick, and George Springer. Jose Altuve showed just about the same power as his x-ISO. Colby Rasmus is notable for the fact that his high .272 ISO falls below his x-ISO of .305. Rasmus' x-ISO is the highest among all major leaguers this season. If you felt that Rasmus has batted quite a few very hard hit balls into gloves---this substantiates that opinion.
Given the sample size, I can't take this analysis much further. Guys like Carter and Valbuena have experienced some bad luck in the power department, and when they hit into some better luck, maybe it might be enough to more than offset possible future regression by Marisnick. Springer's issues may be more about getting into a groove. Based on his history, we know that he is capable of hitting for more power than his x-ISO for the early part of the season.
Exit Velocity
With the advent of statcast, "exit velocity" (the velocity off the bat) has entered broadcasters' terminology. At this point, we don't know enough about the context of exit velocity to know how it should be applied in sabermetric advanced statistics. Also, the statcast data has not been universally available at all stadiums through the early part of the season. This issue should become a non-issue by the end of the season, though.
However, we do know that (all else equal), high exit velocity is a good thing for batters--because it means the ball was hit hard. The leaderboard at baseballsavant.com provides exit velocity data. Here are a few nuggets for Astros' hitters: George Springer has the 11th highest maximum exit velocity (114 mph) among all major leaguers. For average exit velocity, the highest Astros' hitters are Evan Gattis (91.77 mph), Chris Carter (91.4 mph), and Colby Rasmus (90.67 mph).
Any thoughts or questions?