I've never understood why people referred to Doug Fister's July 2010 as a "sample" of any kind. How could Fister's performances in July 2010 relate in any way to the pitcher he was going to be in 2012? ... yet Fangraphs insists on casting a player's MONTH as a type of 2% REPRESENTATIVE SAMPLE of his career.
If Nick Franklin comes up and takes 100 AB's for the Mariners, that's not going to be a "sample" of ANYTHING, except a very small sample of Franklin's skill level AT THE MOMENT.
It really is like watching a motorcycle zoom around through mountain passes, and you "sample" him going around one curve, and project his overall direction from that. Is it not?
Here's where we run into Gaffney's "moving target" problem. He was careful to specify that he was most dubious about pop sabermetrics as they apply to roster construction issues.
James (and I) couldn't agree more. You won't see Bill James casting much of an opinion on whether Nick Franklin should be called up or not.
Industry trends? Sure, you can conclude that pitchers with 8K per game are better bets than pitchers with 4K per game. Industrywide tendencies, those you can capture. Roster construction? Different animal, scientifically.
This isn't a quibble. Sabermetricians' job seems to be to "sample" that motorcycle going around one corner, and then to forecast where he'll be in 4 minutes and 30 seconds. Can we agree that another 15 seconds' worth of "sample" is useless?
Sure, PECOTA will tell you that the last 10 motorcycles taking this turn ended up "averaging" a NW direction.
You want to know that. The problem comes when Rany Jazayerli insists you "correctly value" the path as NW'erly.
... how were you supposed to use "sampling" to project Kyle Seager, one day before his ML callup? His career arc took an unexpected direction when he collided with MLB pitching.
How would "wider sampling" have helped you project Raul Ibanez when he was 26, or Jason Varitek when he was 24, or Jesus Montero when he's 23? Rauuulllll's career arc took an unpredictable turn in his late 20's. No amount of high-speed photography, of his mountain turns, at age 25, would have helped you.
But we see pop sabermetricians, all the time, see a ballplayer called up for one month, and then start gravely discussing his future based on their "sample."
We don' have a sample of Jesus Montero's career, gennlemen. His 2012-13 is not providing a certain fraction of data as to his 2016 outcomes.
Matt, I'd like your comment on this, if you're so inclined.