Skip to main content

Winning March Means Nothing In October: Comparison of Spring Training and Regular Season MLB Performance

It's March, which means the weather will even more Indiana-y (1), the variety of seafood specials at restaurants will increase, and baseball will be in Spring Training. MLB's preseason, in which the 30 teams play around 30 games each (2), is longer than the 4 game NFL preseason or 8 game NBA and NHL preseasons. Which begs the question: since the sample size is larger, might it be more predictive than the NFL's preseason? Sparked partially by the following Facebook post by my brother-in-law, I decided to determine if that 7-0 record is a sign of optimism for the upcoming season (if you are an Angels fan).


Using data from ESPN of the 2003 to 2016 Spring Training (3), and Lahman’s Baseball Database (4) for the 2003-2016 regular seasons, I compared the Pythagorean Expectations of each team's Spring Training and Regular Season performance. By using the Pythagorean Expectation, a better sense of a team's quality is determined than the win-loss record, with less variability due to chance.

The results confirmed the conventional wisdom of Spring Training not being predictive of the Regular Season. The correlation coefficient of the Pythagorean Expectation between the regular season and Spring Training is .2272, indicating the correlation is weak. For example, the 2016 Chicago Cubs were under .500 for Pythagorean Expectation (48.53%) in Spring Training and had a Regular Season Pythagorean Expectation of 67.87%; this would rank 7th since integration (the 1947 Season), compared to a quite middling Spring Training. While there are teams that were great in both or terrible at both, examples of mixed records abound as well. Some of this is due to the fact primarily players tend to be limited in Spring Training to prevent injury, giving the lineups a difference look than the regular season, and some is due to luck. So it would be unwise to put much stock in Spring Training performance; what happens in March stays in March, and everyone is 0-0 come April.


Similar to the NFL analysis, there was a strong correlation (in this case, -.7267) between a Spring Training Pythagorean Expectation and the difference between Spring Training and Regular Season Pythagorean Expectation. This is not too surprising; outliers (high or low) in Pythagorean Expectation are unusual in a league where most teams are relatively close to 50%, and as sample size increases there should be some progression to the mean. 



What was surprising in the data was the comparison between the results from the NFL preseason and MLB preseason. Despite more games in Spring Training, the NFL's preseason was more predictive, with a correlation coefficient of .2618 compared to MLB's .2272; the coefficients of the preseason and the difference between the preseason and regular season was -.6843 for the NFL and -.7267 for the MLB. There are two factors that likely contribute to this. The first is despite the fact there are more games in Spring Training than the NFL preseason, they represent a small amount of games relative to the length of the regular season; 25% for the NFL compared to 18.41% for MLB. Given the nature of greater rotation among key baseball players than in football (pitchers rotate; QBs do not), that factor has an even greater impact in producing small samples. Another factor is in the NFL, the separation of team performance has been greater; there was a standard deviation of Pythagorean Expectations of 16.23% in the NFL compared to 6.882% for the MLB. Mostly driven by the longer season causing grouping of performance around the mean, teams tend to be less exceptional; the 2016 Chicago Cub's Pythagorean Expectation was 7th in the MLB since 1947, but would only rank as 47th between 2006 and 2015 in the NFL. This grouping around the center means that the difference between great and mediocre teams tends to be smaller in the MLB than the NFL.

While Spring Training (and other preseasons) may not be predictive of the Regular Season, that doesn't mean they aren't important. It gives teams opportunities to evaluate talent, get players into shape, and gives managers a chance to experiment with lineups. Most importantly for baseball fans, it is baseball, and a sign Opening Day is coming soon.

Footnotes and Citations
(1) Indiana-y weather is characterized as containing extreme swings of temperature, wind, cloud cover, and/or precipitation over the course of a short period of time, including within a single day. There is also the potential for extreme events including violent storms, tornadoes, massive quantities of snow, and very low and very high temperatures.
(2) The number of Spring Training games per team calculated from data available from ESPN for 2003-2016 is 29.86
(3) http://www.espn.com/mlb/standings/_/group/league/season/2016/seasontype/pre
(4) http://www.seanlahman.com/baseball-archive/statistics/

Comments

Post a Comment

Popular posts from this blog

Brohm and Calhoun: Purdue's New Top Two Choices Analyzed

Earlier in the silly season  coaching search, the top two coaching candidates floated by Purdue's fan base were Western Michigan's P.J. Fleck and former LSU head coach Les Miles. In recent days, it has appeared neither may end up in West Lafayette. Yesterday, news-ish broke-ish that a deal was done-ish with Purdue and current Western Kentucky head coach Jeff Brohm.  Western Kentucky was revealed to be beginning its own coach search, while coach without an agent Jeff Brohm stated no deal existed and he would not think about future plans until after the C-USA championship game today. Another name floated was current Air Force Academy head coach Troy Calhoun. Which are two odd choices when considered together; at Air Force Calhoun ran a run-heavy option offense (although he has experience coaching quarterbacks in the NFL under Gary Kubiak) and Brohm's offense at WKU was a pass-oriented spread offense. Using the same methods I used to look at Purdue's last few coaches , I...
2 comments
Read more

Disney Princesses Are Not All, In Fact, Princesses

This past weekend, I went to see Disney's new (and very good) animated film Moana . There was a (genre aware) exchange between the title character and the demigod Dwayne "The Rock" Johnson  Maui regarding whether or not the title character was a princess. Maui's evidence is as follows: she  a) is "daughter of a chief," b) "wears a dress," and c) has  "an animal sidekick." Of course, the definition of princess is typically understood as a woman who fit one of the two descriptions: is daughter of a monarch, or the wife or widow of a prince (in turn defined as the son of a monarch, a monarch in his own right, or the wife of a princess). Disney markets 11 individuals as "Disney Princesses": Snow White, Cinderella, Aurora, Ariel, Belle, Jasmine, Pocahontas, Mulan, Tiana, Rapunzel and Merida. They also market the two female protogonists (and daughters of monarchs) of Frozen, Anna and Elsa (who is in fact a monarch), in a similar man...
1 comment
Read more

NASL Power Rankings, Games Through 5/5/2017

Definitions: Pythagorean Expectation, Preseason = The previous year's Real Pythagorean Expectation, with a factor of regression to the mean based on the year-to-year correlation of  Real Pythagorean Expectation. Pythagorean Expectation, Real =  Pythagorean expectation of points , based upon goals scored and goals allowed so far this season. An exponent of 1.27, derived from analysis of previous NASL seasons, is used Pythagorean Expectation, Bayesian = The main power ranking. Using Bayes' Theorem, updates its value accounting for new information. Begins with the preseason ranking, then updates week to week after that.  Expected Final Points = Using the Bayesian Pythagorean Expectation, the number of points a team has already, and the possible points remaining, is found Why am I doing a power ranking for a Division II soccer league? Because it is the league Indy has a team in.
Post a Comment
Read more