clock menu more-arrow no yes mobile

Filed under:

Talking Sabermetrics: Sampling The Size

This is one of a season-long series of articles on different sabermetric topics and how they relate to the Houston Astros.

Whoa, Nellie. It's just the second week and I'm already throwing you guys a changeup. Or was that a slider? I can never tell...In last week's article, I mentioned how I'd touch upon using statistics in tandem this week, but I think it's more appropriate to touch up on a singular topic near and dear to Spring Training.

That's right, we're going to talk about Small Sample Size (SSS).

Do you want to know what was at the heart of the conflict in Moneyball? Forget that "stay close to my daughter" or the "I'll prove them wrong" tropes. It's all about SSS.

The cool kids around sabermetric analysis throw this around all the time. Sometimes, they even do it carefully, instead of just throwing it out there before plunging ahead with analysis. As in, "Carlos Lee is 0-for-30 this April. I know, I know, small sample size, but..." Sometimes, they will appreviate it to just SSS, but it's always there as a hedge. They're still going to draw their conclusions, with the caveat that the sample size may not have normalized.

But, I digress. How was SSS at the heart of the matter in Brad Pitt's world? There's not many statistical concepts that completely throw what happens between the lines away for the cold, hard power of numbers. Oh, you're guy made adjustments at the plate? Show me a month's worth of data before I'll believe it. Oh, he's in the best shape of his life? Let's see how he's throwing in July.

It's one of those statistical components that's vitally important to understanding what's going on with the numbers of a team or player, but it's pretty useless in helping that player or team win more games.

Yep, SSS is the proverbial wet blanket.

I bring this up in Spring Training because these few games are the worst sort of small sample size nightmare. You've got varying talent levels of competition that can change even in the later innings of the same game. You've got varying degrees of the kind of competition, with some players choosing to refine a single aspect of their game (hitting opposite field, working on the feel of a curve) that changes the game situations. Heck, you've even removed the elements of scouting or game planning for opposing players.

If all that's going on, why do we care about stats at all in spring ball? Well, because we're human. We like big numbers, because we think they can translate to the regular season. We think Chris Johnson has finally turned the corner and we hate Jack Cust's lazy bones for not showing up at all this season.

Fans have that luxury, but teams do not. General managers and managers have to make decisions this week on who they want to keep on their squads. Can they really say after a handful of at-bats how a player will perform this season?

The answer, courtesy of SSS, is most assuredly no.

Here's a handy little chart showing just how long certain stats have of normalizing during a seasons (or when the sample size stops being too small), thanks to FanGraphs:

Offense Statistics:

  • 50 PA: Swing%
  • 100 PA: Contact Rate
  • 150 PA: Strikeout Rate, Line Drive Rate, Pitches/PA
  • 200 PA: Walk Rate, Ground Ball Rate, GB/FB
  • 250 PA: Fly Ball Rate
  • 300 PA: Home Run Rate, HR/FB
  • 500 PA: OBP, SLG, OPS, 1B Rate, Popup Rate
  • 550 PA: ISO

Pitching Statistics:

  • 150 BF – K/PA, grounder rate, line drive rate
  • 200 BF – flyball rate, GB/FB
  • 500 BF – K/BB, pop up rate
  • 550 BF – BB/PA

Yeah, about the only thing we can learn from a hitter in spring ball? How much he's going to swing the bat. We can't even get to a point of definitively knowing how pitchers are going to do with key stats like line drive rate or K/PA.

So, how do you evaluate a guy like Cust, who doesn't have the plate appearances to show what he could do?

That's where the GMs and scouts of the world earn their money. They have to see how he's gotten there, regardless of the sample size. Adjustments to his swing, his work ethic, if his defense or positional flexibility can impact the club are all secondary ways a club can analyze a player.

Cust doesn't do well there, either, since he barely plays first base and would be stretched pretty badly in the outfield. Also, injuries haven't helped him one bit. That's another factor in the club's evaluation, I'm sure.

For all those reasons and probably some other ones, the Astros decide to cut ties with Jack Cust. They did it mainly on those secondary factors, not his spring stats. After all, 0 for 24 is waaay too small a smaple size to judge anyone one fairly.

All SSS really does for fans is to temper expectations. In that way, it is most definitely a wet blanket. If Chris Johnson hits .350 in the first two weeks, feel free to go crazy just like you did in 2010. Going back to last week, though, don't expect that early batting average to stay high until you've seen more of a sample size.