Bill [reading history book]: "The only true wisdom consists in knowing that you know nothing."
Ted: That's us, dude!
It's early in the season, and there's an all-too-human temptation to latch on to the earliest indicators from a week's worth of games and extrapolate to the rest of the season. Morse is in over his head! Marquis is the new staff ace! Bunting actually works! Then the killjoy stat-heads come along and solemnly intone, "small sample size" (or SSS in its abbreviated talismanic form). "Okay, fine," you think as a casual fan, "but how long until the samples aren't small? When can I point to someone's ERA or batting average and know whether they're a H3R0 or a bum?" As with all-too-many things in life, the answer is, "it depends."
After the jump, a summary of when sample sizes aren't "small" anymore. As your reward for reading (or for skipping past it to the good stuff), I've pulled a selection of early-season statistical wackiness, too.Why do we worry about "sample size," anyway?
Statistics, when applied properly, can bring patterns and even insights out of things that look random or disorganized at first glance. The complicating factor is that stats are more reliable if you more observations. A batter who starts the season with 3 hits in his first 5 AB looks like a .600 hitter. Wait, but then if he has an 0-10 slump right afterward, he's a .150-hitting sub-Mendoza bum. A subsequent 6-15 stretch would take him to an even .300--at what point can we stop and say, "Okay, we've seen enough--this player is good/bad/so-so in this stat." Bear in mind that this is a tricky enough question if we assume the underlying ability behind the stat isn't changing. Trying to use a limited stretch of stats to "prove" that a player's talent has changed because of injury/"figuring it out"/whatever only makes it harder. When you can say the player has gotten better/worse than career average instead of having a fluky clump of hits or whiffs?
Slice and Dice
How do you tell when stats "stabilize" (at what point do they become worth having arguments about on the internet)? A hardy internet researcher by the name of Pizza Cutter did analyses of both hitting and pitching data in the last few years to see whether and how quickly stats stabilized to a steady value over a season. This is from the invaluable fangraphs sabremetics library section on sample size, which has links to the original research (math-heavy, but quite interesting for those who find that sort of thing interesting).
In short, here's how many plate appearances (PA) it takes for batting stats to become reliable (a full season with regular playing time is 650ish PA):
50 PA: Swing %
100 PA: Contact Rate
150 PA: Strikeout Rate, Line Drive Rate, Pitches/PA
200 PA: Walk Rate, Groundball Rate, GB/FB
250 PA: Flyball Rate
300 PA: Home Run Rate, HR/FB
500 PA: OBP, SLG, OPS, 1B Rate, Popup Rate
550 PA: ISO
For me, what stands out here is that a lot of "diagnostic" hitting statistics stabilized pretty quickly: contact rate, line drive rate, strikeout and walk rate, etc. Those are good to have serious discussions after about a couple of months of regular playing time. However, the stuff that really matters for fan discussions: SLG, OBP, etc, takes almost a full season of PA (well past the AS break) to become reliable. And the great cornerstone of baseball stats, the batting average, doesn't stabilize even after a full season's PA! I'll say that again, because it was quite surprising to me: you can't consider a year's worth of batting average data to be a reliable measure of player talent. (The same is true of BABIP--I can't find the link right now, but I believe it takes something like 3-7 years for AVG/BABIP to stabilize, a long enough time period that you have to worry about the player's underlying talent changing over the course of "measurement.")
Here's how many batters faced (BF) a pitcher has to have for his pitching stats to become reliable (a full season of starting pitching is about 800 BF):
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
What's interesting here is that the batted-ball profile and strikeout rate stabilize pretty quickly. Walk rate takes a little longer to stabilize, but HR rate does not stabilize over a full season of pitching. Defense-independent pitching statistics assume the pitcher controls strikeouts, walks and HRs--it seems here that the Ks and BBs become debate-worthy pretty quick, but that random variation plays a big role in HRs (possibly because of low numbers overall? Most pitchers have far fewer HRs than BBs or Ks). Overall, the lesson is that you can't tell that much about a pitcher from one season's worth of data, other than walks, strikeouts, and what kind of hits people get off of him.
I don't care about reliability, I WANT STATS!
Okay, but keep in mind that no National has even had 50 PA or BF yet (okay, Marquis has 53).
- Ryan Zimmerman has racked up 0.5 WAR in 7 games--he's on pace for a 12 WAR season on the strength of his 357/486/536 hitting line (177 wRC+). And the first batch of defensive stats haven't even come out yet. MVP!
- Danny Espinosa has quietly put up a 280/406/480 line (129 wRC+) as a middle infielder.
- Jordan Zimmermann leads the starters with a 3.18 ERA, although his 4.8 K/9 is nowhere near team-leader Tom Gorzelanny's 13.5 K/9 (wha-huh?).
- Tyler Clippard and Sean Burnett are both rocking 0.00 ERAs, while Drew Storen isn't far behind at 1.35
- Mike Morse is struggling through a 148/250/148 slump (22 wRC+). That's the same as Pudge's 176/222/235 (22 wRC+).
- Brian Broderick's 23.14 ERA could be better, as could Chad Gaudin's 8.31 and Tom Gorzelanny's 8.44.
- Livan Hernandez is striking out less than one batter for every one he walks (versus Marquis', Clipp's and Burnett's rates of 5 K/BB or better!).
Versus the league
- Overall, the Nationals' offense is 14th in the NL, with an average team team line of 217/321/328 (78 wRC+) that's ahead of only the Braves and the Cards.
- The Nationals lead the league in walks (44) and walk% (12.3) thanks to their patience and to R.A. Dickey and the Mets bullpen.
- The Nats' team ERA of 4.09 is 10th in the NL (105 ERA-, 5% worse than average). Walks, strikeouts, and HR/9 are all around league average.
- The Nats have the second-most errors in the league (7), although the advanced stats are currently around league average or a bit below.
What have we learned?
Good rule of thumb: don't try to "prove" anything with stats until you have at least a month or two of playing time to talk about, and it really takes about a season to evaluate a player's talent statistically. I'll leave it to you readers to pick out which of the stats I pulled above will turn out to be flukes, and which will turn into something that lasts over an entire season.