Saturday, November 04, 2006

Repeatable Skills

Earlier this week, my Motownsports friend Charles Liston asked whether the ability to hit a sacrifice fly was a repeatable skill. A repeatable skill is one where a player has a high degree of control of the skill and displays it consistently from year to year. One example is walks per plate appearance. A typical player will draw a similar numbers of walks per plate appearance from year to year. A non-repeatable skill is one which is largely out of the control of the player and is at least somewhat dependent on luck. We will see that sacrifice flies per plate appearance is not really a repeatable skill.



One quick way to measure the repeatability of a statistic is to examine the correlation between two years for players that played in both years. If the statistic is repeatable, then players who did well on that statistic in year 1 will also do well on that stat in the second year and players who did poorly on the stat in the first year will tend to do poorly on that stat in year 2. If the statistic is non-repeatable, then what a player does in year 1 will tell us little about what he did in year 2. The strength of the relationship between years 1 and 2 is measured by the correlation coefficient. The correlation coefficient falls between 0 and 1 with 1 indicating a perfect relationship and 0 indicating no relationship.


For this study, I included players who had 400 or more plate appearances (PA) in 2000 and 2001 (145 players), those with 400 or more PA in 2002 and 2003 (134 players) and those with 400 or more PA in 2004 and 2005 (149 players). This gives us a total of 428 observations. Using the 400 PA cut off creates a little bit of bias because players who do much worse in the second year might not get 400 PA in the second year simply because they were not doing well. I included the cutoff in order to avoid small sample sizes. When I tested the data with smaller cutoffs, the correlations were a little smaller but not enough to change the conclusions.


The results are shown in the table below. You will see that sacrifice flies has a very small correlation (.21) and thus is far less repeatable than other stats. This is not to say that no players have this skill but it does say that most probably don’t.



Some of you might be surprised that strikeouts (.80), walks (.83), and power stats - such as homers (.75) and isolated power (.76) - are much more repeatable than batting average (.43). This means that strikeouts, walks and power stats are much more predictive of future performance than batting average. This is because batting average is more dependent on luck (or opponent’s fielders) than the other stats.


Table 1: Repeatability of Hitting Skills

(n=428)

Statistic

Correlation

SO/PA

.83

BB/PA

.80

Isolated Power

.76

HR/PA

.75

Sac bunts/PA

.70

OBP

.67

Slugging Pct

.67

GIDP/PA

.46

Batting Average

.43

Sac flies/PA

.21

6 comments:

  1. Thanks for doing that Lee, that's interesting stuff. It's no surprise from your study that the SF isn't repeatable.
    KS

    ReplyDelete
  2. Charles ListonNovember 05, 2006

    That's really good Lee, the non-repeatability of batting average was an unexpected outcome as far as I was concerned. As for sacrifice flies, my reaction so far is perhaps akin to a terrier who won't put down your shoe.

    I'm thinking that sac fly opportunities are sort of a random event - man on third, less than two out - and won't necessarily be consistent for any particular player from year to year. So perhaps "per plate appearance" isn't ideal as a basis. However, "per sac fly opportunity" might be unattainable without research that is beyond our scope here.

    Even "per opportunity" wouldn't test it properly though, on further reflection. I'm envisioning a guy who can deliver a fly ball "when the situation calls for it", and many of the statistical opportunities would occur in situations that didn't reward a sac fly - a blowout game for example, or losing by 2 runs in the 9th inning, etc.

    OK you've convinced me, there is no statistical basis from which to argue that a sac fly is a repeatable skill.

    ReplyDelete
  3. Great stuff, Lee. Interesting to me that the most repeatable skills appear to be walks and strikeouts. Add reliability to the list of reasons strike zone judgment is an important thing to look at when evaluating a hitter.

    ReplyDelete
  4. Charles, you are right that my method
    probably is not as appropriate for sac flies as the other stats. Looking at sac flies per opportunity would be better. There is a similar issue with sacrifice bunts. Sac bunts were repeatable but that's probably because some players get more opportunities than others.

    ReplyDelete
  5. Good write up.

    I was glad to see BB/PA up there, it's something I always assumed to be consistent, but never did the work to confirm my assumption.

    ReplyDelete
  6. why wasn't Chris Shelton not given any appearance time during series when despite his being sent down to minors, he had been a valuable asset early on in the season? Chris Duncan of Cards went to minors, and upon return was allowed to show his revitalized stuff and be in the Series.

    ReplyDelete

Sabermetrics Book

Sabermetrics Book
One of Baseball America's top ten books of 2010

Blog Archive

Subscribe

501 Baseball Books

501 Baseball Books
Recommended by Tiger Tales

Stat Counter

Site Meter