Not criticizing it or anything, but I just wanted to note for both of our sakes that NYRoyals’ stuff about rotation slots and my compilations of projections and WAR calculations based on them are two separate things.

Second, most people are aware of the limitations of FIP. In figuring Team WAR, when we do defensive projections (if you buy those), we figure that using ERA would be redundant.

If you look back at my older Driveline Post on Pitcher Value, I discuss some of these things and ways of accounting for them. ONe way is to average FIP and ERA, although for relievers, I thikn FIP alone is still the best because of sample size.

I discuss my favorite pitching stat, tRA, from Stat Corner, which is like FIP, but with more park adjustments, and accounting for the run expectancies of LDs, GBs, FBs, etc.

Good thoughts.

I don’t have much time, and I’ll probably respond more fully later, but I did want to throw something out there. I don’t think FIP rewards finesse pitchers. But it does reward groundball pitchers, and rightly so. Inducing a high percentage of GB’s can be almost as valuable as getting a lot of K’s (because GB’s very rarely become extra base hits and never become HR’s). So, if a pitcher has a high GB% and a low FB, he’s going to have fewer HR/9 (which is one of the three elements of FIP, as you know). And one of the reasons that HoRam and Ponson have decent FIP projections is because they have high GB and low FB. In recent years, Ponson has been in the 53-55 range for GB’s while HoRam has usually been in the low 50’s.

Also, I want to second d_f’s endorsement of tRA. It is basically a bigger, better version of FIP which includes more batted ball data. If I had tRA projections for Royals and AL Central pitchers, I would have used that. I just had to use the best stat available.

I asked Tom Tango a while back about this subject and he devised a Simple tRa when I was looking into FA pitching:

Link to Tango

Link to my Blog

I was wanting a stat that is projectable from one season to the next and this one is not too bad.

I know the Tango crowd just dropped a good dime on measuring the time of various balls til they hit the ground to get a good data set. We will have to see if it become public or gets sold/put in a book.

Then I read more about FIP on fangraphs, and though it had a problem because it didn’t take into account things like GB/FB/LD% (like xFIP does). I thought it would be giving an inherent advantage to pitchers who had a high LD% but an abnormally low HR/FB% relative to their LD%.

A flukey HR/FB percentage (anything significantly more or less than about 11) is going to skew a pitcher’s FIP. You can see this last with with HoRam (a 2.9 HR/FB really helped keep his FIP very low) and Peralta (his 18.8% killed him). But I don’t think this hurts FIP’s predictive ability much because projections use anywhere from 3-5 seasons of data, which should level out the deviations from the HR/FB mean caused by good or bad luck.

Then I read more about FIP on fangraphs, and though it had a problem because it didn’t take into account things like GB/FB/LD% (like xFIP does).

Does xFIP do this? It was my understanding that xFIP just normalizes HR/FB% in an attempt account for the good/bad luck discussed above. I know THT’s stat glossary describes xFIP in this way. But there may be multiple versions of xFIP. Is someone doing a xFIP with more batted ball data.

But I do agree that FIP is limited to the extent that it doesn’t take into account more batted ball data. That is why I think tRA is superior. But FIP remains clearly superior to stats like ERA in isolating and accurately describing pitching performance.

Sure [Chris George’s] 2003 was awful, but in 2004 he started 7 games for the Royals. He struck out only 3.21/9 and walked an awful 5.36/9. Yet due to only giving up 1 HR, he had the FIP of a #3 pitcher: 4.43. Despite the good FIP and sabrmetrics taking a firmer footing in front offices, he has not pitched in the majors since. So, am I off on this point? Am I missing something?

First, there’s a significant sample size issue. His 2004 MLB stats cover 42 innings. That’s not enough to evaluate how good a player really is both because it isn’t much data and because small samples can skew results. As you pointed out, him giving up only 1 HR really lowered his FIP. You really have to take sample samples of data with several grains of salt. Second, gopherballs was right on when he said that when you look at a pitcher’s FIP, you also have to look at the batted ball data to see what else was going on and if it looks like he the pitcher suffered some good or bad HR/FB luck. Augmenting FIP with LD, GB, FB% and HR/FB gives you a much more complete picture of a pitcher’s performance. While George’s FIP was ok, his LD% was very high, his GB% was low and his HR/FB% showed that he was the beneficiary of a lot of luck.

Not to rehash the argument, but this post, along with the subsequent explanations of tRA by devil_fingers and NYRoyal, has helped to explain my confusion with the apparent suckage of Ponson and his seemingly generous FIPs.

When looking at Ponson’s career tRA’s, I am now convinced more than ever that Ponson is terrible — however, I’m no longer as convinced that he isn’t an improvement over the rest of the AL Central’s 8th starters. (I’m looking at you Dontrelle…)

Now it would be interesting to see a comparison of the AL Central’s rotations with regard to tRA. It’s too bad there aren’t the same sort of forecasted figures available for the pitchers. Guess we’ll have to wait until year’s end to get a sense of what happened.

Guess we’ll have to wait until year’s end to get a sense of what happened.