In my last post, I introduced my new metric: xbFIP. Expected-ballpark, extra-base, fielding independent pitching is my take on FIP. Fielding independent pitching was created, as I outlined in my last article, in an attempt to evaluate a pitcher on what he can truly control: walks, strikeouts, innings pitched and home runs. Given that, as Voros McCracken discovered, the amount of balls in play which fall for hits do not correlate well on a year-to-year basis, currently FIP is the best way to quantify a pitcher’s success. I previously described that I felt doubles were on the pitcher and that home-runs yielded weren’t entirely on the pitcher. I accept that in my rather vague introduction, I failed to provide any kind of evidence that would prove my ‘theory.’ So, here I am, expanding upon my introduction of xbFIP.
Before I get into a pitcher’s control of doubles and triples, I’m going to focus on the home-run element of my metric. The general premise was that pitchers don’t truly control home runs. While, yes, xbFIP tried to combat this, personally, I don’t think it has. Assuming that a pitcher will allow a certain percentage of fly-balls to leave the ballpark, as opposed to using their absolute home-run totals was a step in the right direction. However, assuming that 10.5% of their flyballs will leave the ballpark is a little of base, in my opinion. For example: Madison Bumgarner pitches in the pitcher friendly confines of AT&T Park, where, on average, 6.77% of fly-balls leave the park. Surely, it’s wrong to assume that Bumgarner will yield a home-run on 10.5% of flyballs. On the contrary, Masahiro Tanaka pitches in the home-run friendly confines of Yankee stadium, where, on average, 14.68% of fly-balls leave the park, far from the 10.5% league average xFIP uses.
Given that there’s a vast difference between certain ballparks concerning the percentage of fly-balls which leave the park, I felt it was only right to change that element of FIP. So, what I did, as I outlined in my previous article, was: I took each pitcher’s ballpark-factor, added it onto the league average HR/FB% – to account for the starts made on the road – and rather simply divided by two. Typically, each pitcher’s new ballpark-factor adjusted HR/FB% won’t differ too much from the 10.5% league average; but with some ballparks being extremely home-run/pitcher friendly, I still felt it was important to use. In the following chart, the correlation between the ballpark-factor adjusted HR/FB% I calculated for each team and the team’s actual HR/FB% in 2014 can be seen.
As can be seen in the chart above, there is a very strong correlation between my ballpark-adjusted HR/FB%, and each team’s actual HR/FB% last season. Hence, it’s pretty fair to assume that my ballpark-adjusted HR/FB% that I’ve used in my xbFIP deserves to stay. Despite there being two anomalies; the correlation is still a very strong .906. With the adjusted HR/FB% now proved, on to the harder part: proving that pitchers have some control over doubles and triples. Well, enough control to accept xbFIP as a useable metric. In order to do so, I’ve compared a handful of pitchers’ – randomly selected, same as last time – FIP, xbFIP and next season ERA. Once more, I’m going to use the next season ERA not to test the predictiveness of it, but rather; given that xbFIP is an attempt to illuminate how a pitcher truly pitched, his ERA should shift towards his xbFIP, if accurate.
At first look, xbFIP seems to perform pretty poorly. With a mere correlation of .267, there seems to be no relationship between xbFIP and the ERA the pitcher posted the year after. However, when we look at the relationship between FIP and ERA; xbFIP looks a hell of a lot better. Remember, after all, that the purpose of these two graphs isn’t to show how predictive each metric is. Rather, to compare the results of the two metrics to see if doubles (and triples counted as doubles) have a positive/negative impact. Now, on to FIP and ERA.
If you thought the relationship between xbFIP and ERA was bad, then just have a look at FIP. The aforementioned premise of not testing the predictiveness can’t be overstated. I mean, okay, the correlation between xbFIP and the ERA the pitcher posted the year after is pretty minimal; however, including doubles (and, again, triples counted as doubles) – or, to put it better: using xbFIP as opposed to FIP – gives more accurate results. Does that mean pitchers can control doubles? Nope. However, it is another step in the right direction. And, thus far, xbFIP is performing rather well.
To continue comparing xbFIP to FIP, I’m going to look at how well each metric predicted the trend in the starter’s ERA. Did xbFIP/FIP predict that the starter was due a regression in his ERA? Or, did xbFIP/FIP predict the starter was set to lower his ERA? Well, xbFIP successfully predicted 16 of the 19 starters’ trends correctly – it’s still a tiny sample size. On the other hand, FIP successfully predicted 15 of the 19 starters’ trends correctly. Once more, xbFIP edges out FIP – in this tiny sample size, that is.
So, in conclusion: in the second installment of introducing xbFIP, I hope I’ve managed to convince you all a little more. My ballpark-factor-adjusted HR/FB% is almost perfectly accurate and hence should almost certainly be used in my new FIP. Doubles – and triples counted as doubles – are far from set it stone. But, it is going in the right direction. Thus far, it has outperformed FIP and hence, it isn’t too far off base to consider that a pitcher might actually be able to control doubles.