This past week, MLB announced that they will institute a blood test for hGH in the minor leagues. There's a lot of hype about hGH right now, but the science on the subject is fairly clear: there are minimal if any positive effects of using hGH as an adult. Here, I'm linking a very readable review paper looking at most of the common PED's that I assigned to students in my baseball class. It turns out that perhaps the best demonstration of hGH's effects occurs in a condition called acromegaly, which is caused by elevated growth hormone titers in adults. In this case, affected individuals actually experience a weakening of muscles. MLB is wasting both time and money pursuing a test for what is apparently an expensive placebo.
Good piece by Graham describing he logic behind the positional adjustments used in WAR. One small quibble: the other flawed assumption behind offense-based position adjustments is that position talent (offense+fielding) is equal across positions. It's not. 2B's and 3B's are equivalent as fielders (on average), but 3B's hit better.
The debate rages on DIPS statistics. Essentially, we have three more or less equally accurate (at the population level) approaches at this point. xFIP, Tango's new bbFIP, and SIERA. They all differ substantially in their construction, but there's a sense in which they all do the same thing and come to very similar estimates most of the time. The other player is tRA*, which has yet to be empirically tested, at least publicly. On top of this is a hardcore community vs. BPro social divide, which seems to be driving some hostile feelings. It's an interesting thing to watch...but the payoff is that we're seeing more advances in component-ERA estimators this past week than we've seen in the year+ since tRA was published.
Tom Tango, inspired by SIERA and the conversation around it, derived yet another component ERA estimator: bbFIP. Very simple little construction, but uses batted ball data instead of HR/FB data. And it works as well as SIERA or xFIP. And then, in the same thread, there's also an older stat, kwERA, which is even more simple, looking only at walks and strikeouts...and it performs just as well in terms of prediction! Amazing stuff.
Lee Panas published a terrific sabermetric primer. It's extremely current, with great scope, and will be an awesome resource for those interested in learning more about sabermetrics--especially player valuation statistics. I'm linking to Tango's review of it, but you can find the book on Lulu. If I do my baseball class again next year, I'll probably assign Lee's book.
More testing of PECOTA, this time at its team-level projections. Sky Andrecheck finds that it probably needs to be regressed a fair bit toward the mean, at least over the last two years (2008 & 2009). Whether the same will carry forward in 2010, I don't know--BPro seems to be working hard to get PECOTA back into shape in a variety of ways.
I'd like to throw out a link to a little study that I published at Hardball Times this week. It breaks down the run scoring environment across the minor leagues, and finds (not surprisingly) major differences. What I like about this piece is a) it looks at component statistics in each league and finds, for example, that not all high run environments are constructed equally. And b), that it allowed me to develop a base runs equation that works a bit better, at least for this purpose, than others I've used.
Neat study by Harry showing massive stringer bias in some individual minor league parks, especially Huntsville, where they just stopped using LD!
Jared Cross takes a look at Will Carroll's team health reports. He finds that, at least for players with the "red light," we'd be right to expect players to appear on the DL 10-20 more days than "green light" players. Yellow light results are mixed. To me, this is a pretty favorable result for Carroll's work, as injuries are not easy things to forecast.
John looks at how differences in fastball velocity affect the degree to which CHONE (and, apparently also, PECOTA) misses high or low on pitchers, and finds a real effect--at least for soft tossers. The inclusion of more and more batted ball/velocity/pitchf/x data into projections is the way forward, as John shows here.
After a long period of development, Brian Cartwright's Oliver projection system finally makes its debut. It's a quality system and with a few tweaks should be competitive with the other heavyweights: CHONE, ZiPS, Marcels, etc. (and yes, by including Marcels, I'm saying that the most basic system is one of the best--its solid showing in every year's projection roundup can't be denied).