Sample Size and Spring
by, 03-18-2012 at 05:27 PM (769 Views)
I love this site. Thanks guys! I enjoy reading the thoughts and projections for the opening roster. However, I cringe every time I see anyone use spring data in support of a player making a roster.
All winter long, I listened to Gleeman and the Geek remind me that Parmelee's 86 plate appeaeances last fall were not predictive of the future. They are essentially correct. There have been some significant studies about sample size and at that size we might start learning something about contact rate. It is also true that every sample means something. I am very certain that Drew Butera will never have an 86 PA stretch that matches Parmelee's OPS over that stretch. That it means he is almost certainly a better hitter than Drew Butera is not really useful. We knew that from the AA data.
The spring data is essentially useless and shouldn't be used in support of a player making a roster or not. If the Twins assess that Michael Hollimon belongs on the opening day roster, they must not base it on his data from a few spring games. We shouldn't either.
Last spring, Danny Rams hit .500. Steve Holm .400. Delmon Young had an OPS of 1066 (he is doing quite well this spring also). Nishioka hit .345. Dusty Hughes pitched 12 scoreless innings. Matt Capps' ERA was also 0.00. This data was just as useless last year as any of the incomplete spring data we are using this year.
So if it's is useless, how do we join in projecting a roster for the Twins? How do the Twins make decisions? I hope they throw away the statistics this spring as they make decision about the roster. If not we may end up with a Dusty Hughes in the bullpen.