Spring training is on the horizon and we baseball addicts are once again ready to invest our hearts and souls into a seven-month long trek that hopefully leads to the promised land. Some of you may be lucky enough to enjoy March exhibition baseball in the warm Florida weather. As such, you will most likely wear the team paraphernalia, collect autographs, and keep score. Hey, fans need to warm up too, right?
However, that’s all spring training is: good fun; a meeting with a long-time friend whom you haven’t seen in a while; a warm-up before running the marathon of the upcoming 162 games (and the playoffs). Spring training is not a harbinger for the regular season.
You may have seen me use the phrase “small sample size” here one or five thousand times. Spring training is a breeding ground for SSS’s. Consider that Phillies pitchers completed 308 innings in spring training last year but they were compiled by 31 different pitchers. That’s an average of about 10 innings per pitcher. Additionally, 87.1 of those innings (28%) were compiled by 14 hurlers who did not appear for the Phillies during the ’09 regular season, leaving 220 and two-thirds innings for 17 pitchers, or about 13 innings apiece on average.
Certainly you wouldn’t make inferences about a player’s ability based on a measly 13 innings. As an example, in his first ten innings last year with the Cleveland Indians, Cliff Lee had an ERA of 9.90. Ten or thirteen or twenty innings don’t tell us anything about a player.
To show us how useless spring training data is, I made a scatterplot with Phillies hitter and pitcher performances in both spring training and the regular season. For hitters, I used a cut-off of 35 AB in ST and 100 in the RS; for pitchers, I used a cut-off of 7 IP in ST and 25 in the RS.
What the r-square tells us is how much of X can be explained by Y. In other words, how much of the players’ regular season performances can be explained by their spring training performances. The hitter data had an r-square of .016 which means 1.6% of their regular season production is explained by how they fared in Clearwater, Florida. Meanwhile, pitchers had a lower r-square of .0037, or 0.4%.
Not all spring training data is entirely useless, but you have to be very aware of the SSS. For instance, the r-square for Phillies pitchers’ K/9 rate in spring training and K/9 rate in the regular season was 0.6.
Weaker relationships are also found: the r-square for BB/9 rate was .15. ST and RS OBP was correlated negatively with an r-square of 0.17; SLG was correlated positively with an r-square of .11. For more on correlation coefficients and the r-square, this primer by Phil Birnbaum is a must-read.
When you take in spring training, do so simply for the enjoyment of the great game of baseball. Don’t use spring training performances to perfect your fantasy baseball roster or to settle an argument with a friend about Brad Lidge’s future. Spring training tells us remarkably little about a player’s skill level.
Note: The numbers used to compile r-square data also constitutes a small sample size, so don’t take them as gospel. Including data from all 30 teams with the same — if not more stringent — qualifiers would give us a much better idea as to which stats are truly meaningless, which are mostly meaningless, etc. However, the Phillies’ 2009 ST and RS data would have to be an extreme outlier for the findings here to be completely irrelevant.