Following up Andrew’s post about Uggla and a comment from a reader, I took a look at Uggla and have to admit this is a very tough case to examine.

Talk about a guy that is on fire. Uggla is currently sitting with 14 HR 32 RBI and a .315 BA.  His batting average is propped up by a 37% hit rate so it should regress, right? Well, that’s not so clear.

His xBA is a robust .330, so that at least indicates that his batting average, given what he has done so far, is legit. However, one component of xBA is his Linear Weighted Power score, which is something I mentioned on the Wiretap show from Fantasy Baseball Mafia last week in the discussion of Nate McLouth.

When we look at xBA it is not enough just to take that number as the guide for the future. It is a predictive measure that is very solid, assuming that there is no skills based randomness built in. To put it another way, assuming that the player’s performance is not out of character in comparison with his statistics, xBA will do a good job of precipitating out the luck component.

However, here, and in McLouth’s case, we have an xBA that is propped up by both a lofty hit rate (likely a randomness result) and a gaudy PX score. What do we do in those circumstances?

Job one is to figure out if the power is for real. In Uggla’s case it should be clear that it is not. He has had a tremendous month of May, with a 334 PX score. That means he was more that three times the average in his PX score during the last 30 days.

His career numbers, such as they are, show that he could maintain a PX in the 160 range. This still shows that his season will almost certainly be very solid even on a go-forward basis, but he is due for a relatively large correction. It is possible he could collapse, and in the short term there is a strong possibility of a hard correction. But 30 HR seems like a virtual certainty, which means he should hit approximately 15 more HR for the rest of the year.

As the PX declines so will that xBA, and his actual BA for that matter. Uggla is streaky enough that he could easily hit .150 for June, causing a correction to the mean in one fell swoop. So as the PX and BA decline we will see a hard regression to the mean.  I would still expect a BA around .260 going forward.

It is not easy to ferret out results such as these, and knowledge of the stats and how to use them is not enough. Knowing the actual components of the metrics and how they interact is needed. I am looking for a big correction despite the superficial result that could be obtained merely by looking at his xBA and PX scores. But who knows if this analysis will turn out to be correct? It shouldn’t matter. Process is what matters not results.