These are two alternate pitching metrics, developed by Tom Tango, both scaled so that values are similar to earned run average, but which only focus on certain parts of pitching.
FIP is “Fielding Independent Pitching”, a metric that ignores results that are dependent on fielding. So it takes a pitcher’s walks, hit batters, strikeouts, and home runs allowed as inputs, and outputs a number on the scale of ERA, so one can quickly sense what good or bad performance is in the number.
The original FIP formula, and what I’m using on the RotoValue site, is:
FIP = 3.203.10 + (13 * HR + 3 * (BB + HBP) - 2 * K) / IP
While the coefficients seem arbitrary, they weigh events in relative importance. Home runs are much more damaging than walks, and a walk hurts more than a strikeout helps. The 3.2 constant added to the ratio is added so that the statistic’s range roughly matches that of ERA.
Fangraphs shows FIP in its leader boards, but it uses a different constant for each year, computed so that the cumulative FIP of all MLB pitchers matches their cumulative ERA for each season, and indeed they’ve computed FIP constants going back to 1871!
I’m sticking with 3.10, however, for a few reasons. First, in most of the discussions of FIP, I’ve seen the formula presented using 3.10, rather than a year-specific constant. And while it would be possible to add a constant for each year, taken to the logical extreme, if I’m showing FIP for a part of a season (RotoValue easily shows you statistics for any date range within a year), then to be consistent I should compute a separate constant for that specific date range, which makes things messier. It’s just simpler to use 3.10 all the time.
So the FIP data I show won’t exactly match Fangraphs, but it should differ no more than by a constant.
Tango has a more detailed discussion of the event weightings for FIP here, noting among other things that the inclusion of IP in the denominator implicitly does take fielding into account somewhat, and also discussing whether there should be different coefficients for events based on different run environments. This helps explain why he’s chosen the coefficients the basic formula uses.
kwERA is an even simpler metric, relying solely on strikeouts, walks, hit batters, and plate appearances:
kwERA = 5.40 + 12 * (K - (BB + HBP - IBB)) / (BF - IBB)
The basic structure is similar – coefficients times raw statistics, plus a constant, intended to make the number have a similar cumulative value as ERA. Here this removes HR (which are affected by ballpark, and I suppose stellar defense from the likes of Mike Trout, or, back in the day, Gary Pettis), and it also switches the denominator from innings pitched (which depend on how often the defense turns balls into outs) to plate appearances. One other minor difference is that this metric excludes intentional walks from both the numerator and denominator, which makes sense for a metric used to assess a player’s skill. You should not be penalized in the metric because your manager tells you to walk a player intentionally.
So RotoValue uses the above formula for kwERA (sometimes called “Strike Zone ERA”).
From a fantasy baseball perspective, these can be interesting either as scoring categories themselves, or as better predictors of ERA than ERA itself. While most fantasy players are well aware of the very high degree of luck in wins, many may not realize that ERA itself is subject to quite a bit of variation based on luck, too, and both FIP and kwERA fluctuate much less. So they’re a better indication of a pitcher’s true talent than ERA typically is.
I’ve added FIP and kwERA as display categories for baseball demo leagues, so you can see lists of pitchers for AL and NL leagues, sort by different statistics, and even change date ranges for the data:
These demo leagues also compute RotoValue prices assuming a 10 team league with a $260 salary cap per team, using 10 active pitchers, 3 corner infielders, 3 middle infielders, 5 outfielders, and 2 catchers per team. The AL teams also use a DH.
RotoValue prices are customized to your particular scoring and settings, taking input some set of statistics, and returning a theoretical auction value for a player given those statistics.
Update Tue Dec 25 13:31
Craig Kimbrel was simply amazing last year in kwERA – 0.18! 116 strikeouts against just 14 walks will do that!
Update #2 Tue Dec 25 13:44
I’ve changed the constant in FIP from 3.20 to 3.10, which matches the formula shown on the Wikipedia page for Defense Independent Pitching Statistics and happens to be much closer to the 3.095 value Fangraphs uses for 2012.