No products in the cart.
Seeing as this is a retread of a post I wrote least season, here’s the big block quote from said post:
Spring training stats don’t matter. Except when they matter a little bit. In 2015 The Economist’s Dan Rosenheck presented the evidence that certain stats like walk rate, strikeout rate, isolated slugging, and steal attempt rate could increase the accuracy of projections–especially for rookies. You can view his slides from the MIT Sloan Sports Analytics Conference here. Although Rosenheck describes his math as elementary, I’m going to dumb it down a bit to get to something that’s doable for me and hopefully still useful.
We’re going to compare the performance of White Sox players in the Cactus League relative to the league average, then see how that compares to their ZiPS projection relative to the American League.
This year I used a threshold of 30 PAs for the Cactus League sample (the previous year’s threshold of 40 PAs seemed to exclude some healthy, major-league regulars on other teams this year). The average walk rate (BB/PA) for said players in the sample was 8.67%, the average strikeout rate was 21.18%, the average isolated power was .199 (woah!), and the average home run rate was 3.63%. How does this compare to the 2017 American League season? AL hitters walked at an 8.41% clip last year, struck out at a 21.37% rate, had an ISO of .173, and had a home run rate of 3.42%.
The standard deviations for 2017 AL hitters (min. 60 PAs) were 3.25 BB%, 7.05 K%, 0.064 ISO, and 1.81% HR%. In the small sample size and wide talent variation of the Cactus League, the standard deviations were 5.27 BB%, 8.55 K% , 0.117 ISO, and 3.17 HR%.
This year I added home run% (HR/PA). I also changed the strikeout difference (K Diff) to be positive if the hitter beat expectations vis-à-vis the the ZiPS standard deviations.
|Cactus League Stats||Standard Deviation Comparison to ZiPS|
|Player||BB%||K%||ISO P||HR%||CL BB Z Diff||CL Z K Diff||CL Z ISO Diff||CL Z HR% Diff|
Bolded values in the right four columns mean that the player had a better standard deviation relative to league average in the Cactus League than ZiPS projects him to in the American League. Ryan Cordell is projected by ZiPS to have strikeout percentage 1.22 standard deviations worse than average (30%). Since his very reasonable Cactus League strikeout percentage (12.5%) was 1.01 standard deviations better than the average there, he gets a 2.24 Cactus League to ZiPS K Difference score (1.01 minus negative 1.22 with some rounding involved).
Sometimes this is just a lot of noise. For instance: Omar Narvaez’s 0% home run rate isn’t somehow better than expected just because the standard deviation for HR% in the Cactus League is just about equal to the average HR% in the Cactus League. It’s probably most fruitful to pay attention to the diff scores with absolute values near to or greater than 1. And they’re more relevant for unproven players.
So Cordell’s and Engel’s plate discipline stats and Engel’s power are cause for some measured optimism. Jose Abreu’s plate discipline was impressive, in that he markedly outperformed projections that already saw him as a good hitter. There is a lot of black in the K Diff column, so it will be interesting to see if a majority of White Sox regulars beat their projected strikeout rate (so far, so good in the first game!).
The context from last year’s post is that Rosenheck has said that an extreme spring performance relative to expectations may change a projection by 50 OPS points. Not much.
Cactus League stats were sourced from MLB.com, 2017 American League stats from Baseball Reference and Fangraphs, and Dan Szymborski’s ZiPS projections were found on Fangraphs.