Baseball is being played again, and it feels great. But do the Grapefruit and Cactus varieties of America’s Pastime really matter? More specifically, do team records and player statistics from Spring Training actually correlate to regular season success?
We all want to go crazy over Dominic Smith’s batting average, or be upset at Zack Wheeler’s ERA, but often these outliers get shrugged off with the phrase, “it’s only spring.” However, these are real baseball games being played, and one may ask how can they not correlate? Let’s just see what the numbers have to say.
You can see that there was a slight positive correlation between a team’s Spring Training record and Regular Season record last year. But we all know correlation does not mean causation, and with the linear regression model only having a slope of 0.35 this is a weak correlation at best. A strong relationship would have a slope closer to a 1:1 ratio. Also, the r-squared value shows that only 9% of the data set is explained by the linear relationship, making this scatter plot more scatter and less of a real trend. The orange dot by the way was the Mets in 2018.
So, team records in spring are not crystal balls to regular season winning percentages, but teams aren’t really playing to win in Spring so this makes sense. Teams are playing to a) practice for the upcoming season and b) to see who makes the team for opening day. Spring is more about the success of individual players than the success of the team.
This proves to be false as well, with another weak relationship between spring success and regular season success. Unfortunately, there were only 39 hitters who qualified for both the Spring and Regular season (the Mets lone representative is the orange dot Brandon Nimmo).
Brandon Nimmo being the only Mets qualified hitter raises a good point about who is playing in these Spring Training games. Paul Goldschmidt did not have enough plate appearances to qualify in last year’s spring season; however, his former teammate Kristopher Negron did. Not heard of Negron? It’s probably because he had a grand total of three plate appearances for the Diamondbacks while playing most of the year in Reno before being shipped to Seattle. These are the type of players playing the bulk of spring games. Remember Nimmo was a 4th outfielder at best this time last year.
Only 40 pitchers reached the number of innings to qualify for 2018 Spring Training, and many of them failed to qualify in the regular season making the sample size for pitching too small. This just further shows that Spring Training is for the 25th men on the rosters, not the stars.
So why do we obsess over these “meaningless” stats? The answer is that while they do not have much of a correlation to regular season success, they matter for the players trying to make the team making them matter to us.
We love rooting for players to succeed because frankly baseball is all too often an overly difficult, unfair, and just plain cruel sport. Players can hit a ball 100 mph directly into the third basemen’s glove, or a pitcher can paint the outside corner but the ball still leaves yard. Freak injuries can happen right as a hot streak starts. But when an unproven player finally overcomes the odds makes it to the big leagues it excites the fans. That might be what we are seeing with Smith, so his 1.2615 OPS is very meaningful in a sense, a sense of hope only found when “it’s only spring.”
Thanks for actually using stats to show things here. Often times baseball confuses stats and metrics, with very little statistical analysis of data.
I think it was Collins that said overall numbers in March and September mean almost nothing.
It is also worth noting the level of data and relevance to the desired characteristic being assessed. Duda was the classic performer when things were out of reach. I created a weighted scoring/tracking system a few years back where value was measured against the leverage of the situation. A solo home run in the 9th with a 5 run lead means less than a single with runners in scoring position in a tie game in the 6th.
By the way, the worst thing about the second plot is that the R^2 is being driven by a small number of spots on the fringes. If you got rid of the 2 points with reg season ops >1 and the ST ope at .4 then refit the data I bet it would be worse. Not great for outliers to control the fit.
Good to know for next time thank you Chris!
A more interesting question to me is how a ST sample compares to one of the same (or larger) size from the regular season.
We know that it takes well over 400 PA for OBP numbers to stabilize, so there’s not any point judging ST OBP numbers. Most numbers need well over 100 PA. If I remember right, K rate is one of the few that takes under 100 PA to stabilize. So, we might be able to infer changes in K rate from ST numbers, if ST numbers compare to ones from the regular season.
Last Spring, Nimmo had a 25.7 K% in 70 PA in ST, slightly down from his 27.9 rate in the majors in 2017. During the regular season, he had a 26.2 K%. In this one isolated case, ST numbers were reliable. But we would need to try multiple player on multiple teams from multiple seasons.
For those of you old enough to remember. In late 80s early 90s. Darren Reed played like Ted Williams in spring training but was a bust when got the call.
I’ll tell ya one spring training metric I dont like: the defense is horrific, absolutely putrid. The errors today sickening.