Unusually High XERAs in Projections

[Stat Boy]

I've noticed that the XERAs of NL Pitchers this year seem, almost without fail, to be not only higher than the pitcher's ERA, but a lot higher. This has not been the case in year's past.

Furthermore, players who seem to have promising XERAs in the Forecaster have had a large reversal of fortune in the current projections.

For instance, Sean Marshall, whose Forecaster XERA of 3.77 is a bit of an improvement on his ERA prediction of 3.88, gets hammered in the Projections XERA of 4.19 -- for reasons I'm unclear on. Nothing that has happened this spring should have affected this prediction. If anything, his spring was impressive.

Carlos Zambrano -- never a favorite of HQ's XERA goes from 4.04 to 4.45.

I don't mean to limit this to just Cubs.

Tim Lincecum's xERA went from 2.88 in the Forecaster to 3.21. Ubaldo Jimenez from 3.33 to 3.72.

Virtually every pitcher's XERA is like this. Traditionally, there are a fair amount of pitchers whose XERA is markedly below their ERA and that's often something I look for.

Is something amiss? Why have these XERAs changed so dramatically from the Forecaster and why are they all so much higher?

Please let me know ASAP as I'm prepping for my draft and would like to have informed values (or range of values).

[RAY@HQ]

One component of the xERA formula is a normalizing factor, that basically levels the xERA league-wide to the projected ERA league-wide. I'm not 100% sure about this (my head's not in Forecaster mode right now), but that normalizing factor probably got adjusted, or segregated by league, since the book.

In general, the values on the site are always more current. And no, nothing's amiss.

[Stat Boy]

Ray, it's still very weird to have values so out of whack with the book.

Derek Lowe, for instance, is at 3.89 ERA in the Forecaster and 3.90 on the website. Fine.
His XERA, however, is 3.57 in the Forecaster and 4.06 on the website.

The Forecaster seems to suggest that Lowe's ERA is due for a reasonable positive correction. The website seems to suggest the opposite.

I appreciate that the website is more current. But this seems to be the case for virtually EVERY pitcher.

Something is either off in the Forecaster or off on the website with XERA. There is no reason why Lowe should have a half run swing in that number -- as other pitchers are consistently showing as well.

Why would a normalizing factor up XERA figures by .50? That's a huge swing. I can't believe these numbers are accurate on the website. If they are, it suggests that all the pitchers (with the exception of a few), but certainly the vast majority of starters ALL have XERAs significantly worse than their ERAs.

Could you please look into this? Something is clearly off here.

[RAY@HQ]

Upon further research....

I was correct that the discrepancy is as a result of the normalizing component of the xERA formula. Our tech guys are going to change that component in the site's formula to (nearly) match the book.

Thanks for pushing the issue.

[Stat Boy]

Thanks for looking into it, Ray.

My question is -- what are the ramifications? I've always looked to XERA to give me an insight into the hidden potential (positive or negative) that a pitcher has going into the season (and me into my draft).

The Forecaster had me looking at a lot of pitchers positively. The website erased that.

Not being a mathematician, I'm unclear what a normalizing component is, but I'm a little unsettled that it could so completely affect what has always been one of my go-to leading indicators for pitchers.

Should I be looking at the book more closely before my draft for a more accurate number?

[RAY@HQ]

Just to confirm, this will be fixed in tomorrow's stat update.

The ramifications, really, are minimal. The xERA will basically just end up being re-scaled across the entire pitching pool. I'm oversimplifying, but for illustrative purposes, let's take this example: let's say the total projected ERA of all pitchers in our data set is 4.50, and the total projected xERA is 4.20. The purpose of the normalizing component is to level-set the two, so in this case it would globally adjust all xERAs so that they tie out to 4.50, because that's where ERA is. But that adjustment is different from year to year, which is why the factor changes every year... it's retrofitted, basically.

That normalizing factor was higher in the web site formula than the book formula. So the net effect tomorrow will be that all of the xERAs will slide downward a bit. No one pitcher will improve more than another, percentage-wise.

The other, larger point, which I should have made in last night's response, is that you're not really using projected xERA the right way. You said:

I've always looked to XERA to give me an insight into the hidden potential (positive or negative) that a pitcher has going into the season (and me into my draft).

What you describe here is really the right way to use in-season xERA vs. ERA comparisons. But with the projected ERA/xERA, what you're really looking at is basically two different ways of coming to an ERA projection: the projected xERA is totally skills-based, the projected ERA includes non-skills elements like park factors, consistently low/high hit/strand rate history, etc. It's great that you're looking at both, and large gaps should be questioned (and usually explained by the elements above), but xERA/ERA gaps in projected data don't necessarily say "opportunity" the way they do in-season.

The other thing to keep in mind is that what you're getting in projected xERA is very similar to what you're getting in BPV, but without the normalizing factor. So you can always use BPV as your skills-focused metric in conjunction with projected ERA.