In Love with the NL

Chances are you've heard or been told that, all else being equal, you should target a National League pitcher. I've heard it a thousand times, and it makes some intuitive sense. There's no designated hitter in the NL, and that spot is replaced by a pitcher. That itself should make things easier. There's also the opinion- and I have no idea if the data backs this up, or if there even is data- that NL lineups favor defense in general, and have a tendency to trade offense because of it (essentially, no NL team would put David Ortiz at 1st base like the Sox did for half of 2004).

But it's one thing for an assertion to sound good, and another thing for the data to back it up. I've been curious about this for some time, so

I did a little digging. I used my standard pitching data set- all starters who threw at least 100 innings in a season from 2006 to 2010. I assigned each pitcher a designation of "NL" as a binary variable- a "0" if the pitcher spent the season in the AL, and "1" if the pitcher spent the season in the NL. Pitchers who were traded between leagues were ignored (this happened roughly 30 times out of 300 pitcher-seasons); if they were traded but stayed in the same league they were still included.

I took a look at all potential fantasy-relevant stats here: Wins, Losses, ERA, WHIP, Ks, Fantasy points (with and without W-L record) and VUM (my aggregate roto value measure). I did a simple mean comparison (independent t-test, for those who care) for those stats, divided by AL or NL.

This may surprise you, but the only stat that was affected by a pitcher's league was wins. Losses, ERA, WHIP, etc., were all essentially the same. And even in the case of wins, the difference was only about .67 per season. AND the difference favored the AL. It breaks down like this:

Not a whole lot of difference there. Any given year, it doesn't really matter what league a pitcher is in. You should draft a good pitcher, and not necessarily an AL or NL pitcher.

OK, full confession. There is one crucial oversight with this analysis. This treats every season by every pitcher as an independent event. The assumption is that, essentially, who a pitcher is has no bearing on performance. (For example) Each of Roy Halladay's seasons from 2006 to 2010 is treated as a discrete event, and might as well have been 2009 stats for five different pitchers as far as the analysis is concerned. This is, as the Brits might say, twaddle.

There are several ways to handle this. I chose to give each pitcher an ID number and run a MANOVA (multiple analysis of variance). I used ID as a fixed factor and NL/AL as a random factor. Essentially, the analysis controls for the fact that dependent variables might be related, assumes that who a pitcher is (i.e. his ability/skill set) doesn't change, and that the league a pitcher is in is essentially random.

What we find is, unsurprisingly, who a pitcher is matters. A large amount of the variance is accounted for simply by a pitcher's skill set. But once that is controlled for, whether or not a pitcher is in the NL matters. It matters for everything- all relevant 5x5 stats, and any aggregate measure thereof (points or VUM). For any given pitcher, it is preferable from a fantasy standpoint that he plays in the NL.

So what does this all mean?

It means that your #1 concern in targeting a pitcher should be who that pitcher is. A good pitcher is a good pitcher, so target a good pitcher. However, if a pitcher changes leagues, then a move to the NL should raise his value while a move to the AL should lower it. Similarly, if you have two comparably-skilled pitchers, each in a different league, the NL pitcher is to be slightly preferred.

Basically, it's better to have Justin Verlander than Clayton Richard. But if Richard gets traded to the AL, he's worth less. If Verlander gets traded to the NL, he's worth more. And if the choice is between Lincecum and Sabathia, Lincecum is slightly preferable.