23 September 2013

BABIP, or Why Some People Might Make the Wrong Statistical Conclusion

Note: Most of this article was written in early September, so the information on Morse may not be especially important.  However, my desire to finally complete this post is to address something broader about statistical analysis and Morse is a good example to use in this instance.

Wander around the internets or, perhaps, spend time (a lot of time) at the ball park and you may hear about how certain hitters have been lucky or unlucky.  The comments could be as specific as how line drives are being hit straight at the third baseman, less specific by mentioning BABIP, or even less specific by indicating how performance marks in general are not up to peak career marks.  What I want to comment on is the conversation right in the middle: BABIP.

The general idea of a player being unlucky with with respect to Batting Average of Ball In Play (BABIP) is based on the population trend that the value is consistent.  That is if a component of the population is hitting 50 points over their average, they tend to revert back to their average.  Mike Morse's career BABIP is .333 while his current season BABIP is .267.  If Morse has suffered a simple fluke in the distribution of his batted bats (like most players most of the time) then we should expect that he will revert back to a .333 BABIP in the future.

That expectation dismisses his current performance of 226/283/410 and suggests he should be more in line with 280/340/480.  Having hits drop in, as you can see, can greatly increase a player's worth.  However, we need to remember that this assumption is based on the measure of a population as compared to the measure of individuals.  In other words, a forest might look incredibly healthy from afar, but up close you see that a couple trees are dying.  It is a matter of resolution.

In general, it is fine to say that given such a broad resolution that it is plausible to expect a player like Mike Morse to bounce back from such an uncharacteristic BABIP.  If this is as far as you go, then you might forget what actually impacts changes in BABIP.

The three major components of BABIP are:
1) Defense - Although the unbalanced schedule makes for the possibility of BABIP to be team specific due to differences in stadium design (e.g., Fenway) or general differences in defensive aptitude, usually it all washes out.  Defenses are very similar to each other and there is enough overlap between schedules that a batter essentially faces a league average defense over the course of a season.

2) Random Variation - For no good reason that we have been able to discern, sometimes a player tends to hit more balls in defensible territory than normal.  There seems to be little ability for a hitter to deftly place a ball to an empty spot on the field.  Think of the crazy shifts that are employed today and how no MLB hitter can exploit them with any regularity.

3) Talent - Talent is a combination of foot speed, batted ball type, as well as just how those batted balls travel.

Below is a good video that briefly reexplains what I just noted:

I do want to drill a little deeper on the talent portion of BABIP.  Why?  Because that is what changes in practice.  These chances in BABIP occur because players get stronger, more experienced, and refine their approaches.  Players can also age, adversely affecting their production, and get injured.  I compact them down to three things:
1) Foot speed - as a population, players tend to see a decrease in BABIP over the course of their careers as they lose speed.  This effects grounders almost exclusively and this trend tends to be very small due to how good defense is in the Majors unless a physiologically catastrophic event occurs to the player.

2) Batted Ball Type - again, as a population, balls that are grounders become hits 23% of the time, line drives fall in 69% of the time, and fly balls are wonderful for the hitter 13% of the time.  A player who tend to have more line drives than fly balls will tend to have a higher BABIP.  If something about the player changes the types of batted balls he hits, then that BABIP will change as well.

3) Types of Batted Ball Type - As we all know, a fly ball is not a fly ball is not a fly ball.  We tend to think of a Chris Davis moon shot flyball differently as opposed to a Nick Markakis fly ball.  Within grounders, line drives, and fly balls we get a lot of variety within those classifications.  Those differences can mean a great deal with respect to how well a player will perform.  Soft liners are easily to turn into outs than hard liners.  Shallow flies tend to be caught more often than deep flies (and we are not even considering how deep flies can turn into home runs).
Now, let us use Mike Morse as a vehicle for this discussion.  Below is a table indicating his previous performance.  His 2013 time with the Orioles is not included because I find performance after a trade to be not important with respect to making a trade because that portion is unknown.  Later we will revisit what he did for the Orioles.

2005 23 SEA 258 10 1 3 18 50 .278 .349 .370 97 .341
2006 24 SEA 48 5 0 0 3 7 .372 .396 .488 131 .421
2007 25 SEA 20 2 0 0 1 4 .444 .500 .556 184 .571
2008 26 SEA 11 1 0 0 1 4 .222 .364 .333 91 .400
2009 27 WSN 55 3 0 3 3 16 .250 .291 .481 101 .303
2010 28 WSN 293 12 2 15 22 64 .289 .352 .519 133 .330
2011 29 WSN 575 36 0 31 36 126 .303 .360 .550 147 .344
2012 30 WSN 430 17 1 18 16 97 .291 .321 .470 112 .339
2013 31 SEA 307 13 0 13 20 80 .226 .283 .410 98 .267
Provided by Baseball-Reference.com: View Original Table
Generated 9/1/2013.

A good exercise is now to run through the different variables because upon looking at his .267 BABIP so far in 2013, we would think that he has been incredibly unlucky.  Why?  Because the greater population of baseball players tend to perform at their established career BABIP.  In other words, we should expect Morse to have a BABIP of about .330 to .340 if what has transgressed so far this season is due to random variation (aka balls finding gloves).

However, has his talent level changed?  If it has, then we would expect the BABIP to not be random and instead be a characteristic of that player.

Foot speed

Although Morse began his MLB career as a shortstop, foot speed was never a quality he enjoyed.  He may be slower, particularly with his quad injury that forced him on the disabled list in June through almost all of July.  That said, running out ground balls was simply not a major part of his game.  True, a loss of speed will negatively affect his performance somewhat as a loss in footspeed would result in a slight alteration in singles and doubles, but probably nothing profound.  His game is not a singles game and we do not see a great increase in hitting singles from would-be doubles.  I severely doubt foot speed has caused his BABIP to drop.

Change in Batted Ball Type

Another possibility would be that a leg injury could impact his mechanics and result in him creating a different proportion of fly balls.  The table below shows his batted ball types over the past nine seasons.  Again, I am not considering his time in Baltimore.

Apparently, the table failed to load properly when posting.  I will have to wait until tonight to correct it.  Until then, feel free to go over to Fangraphs here and see how similar his distribution is to his prime years to date.

From that table, nothing appears to be out of the ordinary.  It also appears that from this there is no great reason why we should suspect his poor hitting performance with respect to his BABIP is not a product of a change in batted ball percentages.

Types of Batted Ball Type

One thing that concerned me about Morse was that leg injuries can greatly impact power transfer from a player's lower half into the bat.  If that transfer is off, then swings become all arms, which greatly reduces the distance of those balls.  Even when you see a guy like Chris Davis flip a bat into the stands with only his arms, you can still see a portion of his power is coming up through his legs.  Below is a table showing Morse's average distance for his line drives and home runs.  These numbers are somewhat difficult to find on your own, so I decided to include his few Baltimore data points here.  To be clear here are the three designations for 2013: (a) Seattle before leg injury, (b) Seattle after leg injury, and (c) Baltimore.

Line Drive (n) Fly Ball (n)
2011 266 ft (72) 316 ft (141)
2012 249 ft (59) 309 ft (81)
2013a 276 ft (31) 302 ft (54)
2013b 230 ft (12) 298 ft (16)
2013c 189 ft (3) 269 ft (4)
When I look at the data above, I see potential red flags.  Looking at 2011, 2012, and 2013a, we see some variation in Line Drive distance, which I believe is normally highly variable.  That does not concern me.  What does concern me is the decrease in Fly Ball distance, which I think is a more stable number (again, though, I have not studied this, so this is a bit anecdotal).  I think his early season numbers this year were inflated due to solid line drive contact.  The decrease in his fly ball percentage makes this difficult.  For every 10 feet you lose in your average fly ball distance, the player experiences about a 20% decrease in home runs.  Now, I would not write him off completely long term, but I would be concerned that he might be a player who will flame out early even if he is healthy.

That said, the long term prognosis is not what we have interest in with Morse as a trade candidate.  As a trade candidate, we have more of a concern with respect to how well he performed after coming off the disabled list.  Now, 12 ground balls and 16 fly balls are a very, very small sample size (I assume, I do not know how many batted ball events are needed to stabilize one of these numbers, but I assume this is an accurate statement), but a drop of 46 ft in line drives and his fly balls falling below 300 ft were big concerns of mine.  That suggests to me that even though he was healthy enough to play baseball, he likely was not healthy enough to be Mike Morse and that is a significant issue.  That drop was a major reason why I panned the Avery for Morse deal.  The distance is an illustration that even though his BABIP is low in comparison to his career line, his current BABIP might be completely founded in a change in ability due to his leg injury.

Population vs Individual

This is a major issue for some who are new to baseball statistics and projections.  A good rule of thumb is to expect players to behave like the population of players behave.  This really is true about any population model.  It makes sense for drug trials and it makes sense for ecosystem assessments.  You will often be right more often if you make this assumption.  However, if you are able to assess a single individual, you may have some ability to tease apart the causes for a change in behavior.  For instance, your patient not doing well in the trial may have a secondary issue that causes him to respond differently to the drug.  Or perhaps, your pristine habitat does not act like other pristine habitats because someone is dumping all of their excess paint into the stream.  In Mike Morse's case, it should not have been expected of him to bounce back perhaps because there was an underlying cause (leg injury) and evidence (lack of batted ball distance) that suggested there were real issue there.

In other words, one should not be fundamentalist when it comes to assigning population responses to an individual.  It has been shown consistently that players suffer from variability in their performance.  Much of that wobbles back and forth each year as a result of chance, but there are also real issues that can cause a player to deviate from the expected path.  It is important to recognize these things.  Luck is not all.

How Has Mike Morse performed for the Orioles?

In 27 plate appearances, Morse has three singles and a walk.  He has a slash line of 115/148/115 and a OPS+ of -27.  His offense has been worth -4.1 runs, roughly meaning that in those 27 appearances he was responsible for about half of a loss in comparison to a replacement level player.  He saw 49 innings in the outfield and was credited for -7.5 runs when extrapolated out to a full season.  In other words, he played poor defense.  In a general sense, he has done nothing to improve the team and he cost the team a million dollars.  As you know, money is not a thing that is spent wildly in Baltimore, so losing a million hurts a little bit.  Add that to the loss of a genuine 4th/5th outfielder with options in Xavier Avery and the deal looks slightly worse.  Twenty three year old toolsy, poor skilled outfielders are not exactly a dime a dozen, but they are probably close.  Avery certainly has value and a scant chance at become something more like a second division starter.  With the paucity of talent in the Orioles upper minors, it does potentially hurt a little to give up on Avery's slight promise and his certain versatility.

What Happened to Xavier Avery? 

Avery logged 14 plate appearances over three games in AAA Tacoma.  He walked once, had a sacrifice hit, and had six hits, including a home run and a double.  He also managed a stolen base.  He started once at each outfield position.  For those three games, Avery was a wunderkind.  He was not added to the expanded roster in Seattle.  He is expected to be given a legitimate shot at the fourth outfield position with the club next year.  He is fully expected to log time at some point in Seattle next season.

No comments: