April 21, 2010

Fantasy Beat

When Sample Size Matters

by Marc Normandin

Printer-friendly

I face a dilemma almost daily as a fantasy analyst, given my background in sabermetrics. I know that sample size is important--without a proper sample size, it's hard to take a player's recent performance (for better or worse) seriously. But fantasy baseball is a game that requires quick decisions--you aren't the only person wondering if a player is for real or not, and there is always someone more desperate than you are for any help they can get. We know when statistics matter for baseball analysis of the real thing--you can thank a certain Baseball Prospectus author for that information--but oftentimes (okay, all of the time) you don't have enough time to wait for the information you need for an informed decision.

For example, I wrote about how Jeff Francoeur was worth keeping an eye on due to a potentially newfound understanding of the strike zone. He's taking more walks even if he isn't taking more pitches, and it may just have to do with recognizing which pitches he should and can drive and which ones he should sit on. As far as regular baseball analysis goes, I'm more cautious towards Frenchy because we don't have the sample size to know if this is random variation or not, but as a fantasy analyst (and owner) I scooped up Francoeur off of waivers in one league in the hopes that his performance is for real. When you can't afford to wait for the necessary information, you have to learn to read what is available to you to the best of your ability. You're going to swing and miss like our friend Francoeur, but when you do connect, it's going to go a long way.

This makes the early season a bit tough, as you see players struggling or succeeding at levels that they normally do not, and your desperation can often get the better of you. I tend to have a roster spot or two that I rotate--churn and burn can work if you make enough moves over the course of a year, even if your success rate looks like Rob Deer's batting average--so I can scratch that itch while maintaining the core of the team I had enough faith in to draft just weeks or months earlier. It's also tough from an analysis standpoint, as you're looking for something, anything with meaning at a time when there isn't any.

Here's a real-world example: Chase Headley doesn't hit for a lot of power, so he was largely ignored in many fantasy leagues when he played in the outfield. Now he's at third base (his natural position) and although he still isn't going to set the world ablaze with his pop, he's much closer to useful than he was. He's also a player who appeared to have a bright future, one that struggled to learn the outfield to boot, so there was reason to be optimistic about 2010. He's off to a hot start, hitting .358/.414/.491 with a .400 BABIP. It's clear the average will drop, and his Isolated Power is around the same lines as in the past--the question is whether he will retain enough of his batting average to make him a solid option at third.

Headley's issues in the past had to do with strikeouts and contact rate--he whiffed in over 31 percent of his at-bats in 2008 and nearly 25 percent of the time in 2009. This year, he's at 13.2 percent, a vast improvement at first glance, but to refer back to the earlier linked piece, you need 150 plate appearances to have confidence in an improved strikeout rate for a hitter--Headley is 91 short of that figure as of today. Headley is making more contact in the zone, more contact outside of the zone, and has cut down on the number of pitches he swings at inside of the zone (potentially sitting on pitches he can't do anything with in order to wait for one he can). Problem is, you need 100 plate appearances to know if things like contact rate are for real, and these are splits of contact rate, which makes things even more uncertain.

What's a fantasy owner to do about Headley? Let's forget for a moment that I've seen most of Headley's plate appearances, and can tell you that his swing looks better and more effective than last year, and that he appears to be making better decisions. Let's say you haven't watched a single Padres game or highlight all year, and have no idea about any of this--all you have to go on is the information on his statistics page. You have to make an educated guess without the necessary data, which is the kind of thing that drives analysts insane--it's a necessary evil for fantasy baseball though. The unexpected happens, and there are holes to fill, which often means you need to take a gamble on someone like Headley before someone else in your league makes the decision first.

You almost have to pick Headley up if you've got the hole--you can always cut him later to take a chance on someone else, and if your gamble pays off, and Headley is all of a sudden capable of sustaining, say, a .295/.360/.440 line or so (which, remember, would be better outside of Petco) then you look brilliant. In deep leagues, where players like Headley are stashed on the bench before free agency is even an option, you may have to roll the dice with players of a far lower caliber because it's necessary. Fantasy and real baseball have very many similarities, some of which can help us learn more about the other one, but when it comes to sample size, understanding how the rules work in reality--and then twisting them into the shape you need for fantasy--is the way to succeed. It sounds like knowing the rules so you can break them, but you would rather make an educated guess than a regular one, wouldn't you?

Thanks to DRays Bay for reminding us all about the work on this subject.