Prospectus Toolbox: WHIP-ping Through the Mailbag

Yes, I know, another mailbag article. All I can say in my defense is that as long as folks send me thought-provoking e-mails, I’m going to get distracted and run a little off-course. So it’s all your fault, really.

Anyway, we’ll start off with a purely technical detail about the hitters’ side of the Minor League Stats & Translations report before going off on a bit of a tangent. Reader E.M. asks:

One other question on the minor league translations. After their positions, are listed numbers-some negative, some positive. Is that their FRAA?

Yes, that is FRAA, listed in a format you might be familiar with from the PECOTA cards or the last several iterations of the Baseball Prospectus annual: Games-Primary Position, Fielding Runs Above Average (FRAA). So if you look up Brandon Wood‘s Triple-A numbers and see 70-SS -10 at the end of his stat line, that means he’s played 70 games at his primary position of shortstop, and he’s been 10 runs below average as a defender at that position.

Now that that’s out of the way, reader Julien Headley had a few questions about the “relatively unheard of” stat, POW, which I erroneously called Power on Contact:

I was interested to see you use the POW stat. As far as I know, I invented it when I started my blog in 2003. The form you are using was my original form; I settled into an ‘extra-base hits per contact’ form over the years. I was wondering how you came across it, and what experiences you’ve had with it.

I first came across POW while looking at Clay Davenport‘s minor league translations. The figures in the POW column looked like Isolated Power (and if I recall correctly, once upon a time Bill James used the abbreviation POW to stand for ISO), but a quick check verified that the figures didn’t quite match. So I looked in our glossary, and then in several other glossaries, both print and online, but to no avail. Finally, I asked Clay what it was, and he replied with the equation. Not being familiar with the formula, and due to the lack of data about the statistic, I figured that it was something new or obscure. That’s my mistake, for which I apologize.

Now that I’ve had the chance to check out Headley’s blog, I’ve been able to add the statistic’s proper name (“Power Percentage”) and attribution to the Glossary. If anyone has objections to Mr. Headley’s claim to originating POW, I trust that you’ll settle it in the traditional method of dispute resolution for sabermetricians: gladiatorial combat in a hexagonal arena.

Now that I know what POW is, I momentarily find the pitchers’ version of the statistic more compelling than the hitters’ version. The list of lowest POWs in the majors last year gave us some of the game’s strongest sinkerballers:

Name               RA     RA+    GB%   POW
Tim Hudson        3.49   1.37   62.7  .108
Chien-Ming Wang   3.79   1.26   58.4  .119
Brad Penny        3.25   1.49   50.6  .121
Brandon Webb      3.47   1.42   63.6  .125
Fausto Carmona    3.27   1.45   64.8  .125
Roy Halladay      4.03   1.21   54.8  .125
Joe Blanton       4.15   1.15   48.6  .130
Jake Westbrook    4.62   1.02   55.7  .132
Kelvim Escobar    3.63   1.30   44.4  .141
Adam Wainwright   4.14   1.17   49.6  .144

Meanwhile, at the opposite end of the spectrum, you had some less-accomplished names, along with one high-profile fellow who stands out in a “one of these things is not like the others” way:

Name               RA     RA+    GB%   POW
Ervin Santana     6.18   0.77   36.9  .264
Chuck James       4.30   1.12   32.8  .260
Adam Eaton        6.51   0.75   40.5  .257
Johan Santana     3.62   1.32   39.5  .254
Shaun Marcum      4.30   1.14   42.4  .245
Woody Williams    5.46   0.89   41.8  .240
Jamie Moyer       5.33   0.92   41.6  .240
Josh Fogg         5.38   0.93   41.1  .236
Scott Olsen       6.83   0.68   40.2  .232
Javier Vazquez    3.95   1.24   41.2  .228

Unsurprisingly, the list of high-POW pitchers included some of the most extreme fly-ball pitchers in the majors. While the discovery that fly-ball pitchers give up more extra bases per contact than their ground-ball brethren is hardly profound, it did take me back to the DIPS conversation that we had last year. I still had DIPS on my mind when our customer service liaison, Kathy forwarded me the following e-mail from blogger Nate Rose:

I’m pretty sure I just created a new statistical method of evaluating pitchers. I wrote a post about it on my blog, and you can find the link to it here. Of course, there’s always a chance that this has already been thought up, but I’ve never seen it in any statistical glossary that I’ve ever read, and I checked before publishing the post just to make sure.

Sometimes, we’re criticized at Baseball Prospectus for not responding to outside critiques. It’s a conscious decision, made at the management level-we’d rather talk about baseball than about ourselves or about our colleagues in the world of baseball analysis. While we don’t respond to every broadside in every blog, we also by and large don’t spend much time critiquing others’ work-with the exception of a situation like this, where someone has basically written in asking for feedback.

So I checked all of the usual sources, and I came to the same conclusion that Nate did-no one seemed to have picked up the particular formulation for “mistakes” per inning that he did: M/IP=((HR*4)+BB+HBP-IBB)/IP. However, that’s hardly definitive-the DIPS family of statistics has sprouted more variants than the Power Rangers, and it’s pretty hard to keep track of them all.

Still, I’m not sure that just having a unique formula is enough to lead to a triumphant declaration of “we have a new stat!” After all, some metrics have enough variables that they’re almost infinitely modifiable, but not every modification is worth the trouble. Successful metrics should ideally bring something new to the table, something that’s an improvement over the status quo. For example, that something can be simplicity-someone takes a stat that required fifteen steps and proprietary data to calculate, and does it in four steps with the kind of information one could find on the back of a bubblegum card, and without an extreme loss of accuracy. Or it can be an element of context-league, ballpark, era adjustments-or a closer relationship to game events like runs scored or games won.

When I see Rose’s metric, the main thing I think of is Fielding Independent Pitching (FIP), which is the brainchild of Tom Tango (a similar formula was created by video-game designer Clay Dreslough, and called DICE, or Defense Independent Component ERA). That formula is FIP=((13*HR)+(3*BB)-(2*K))/IP. So basically, in FIP the proportional values of homers and walks are close to the same as they are in M/IP, so the big distinction between FIP and M/IP is in losing the positive value of the third of the Three True Outcomes, or strikeouts. To see if this is an improvement over the previously existing metric, I ran correlations between the two stats and various run-prevention standards, to see how well the metrics matched up to the results: if “mistakes,” as Rose terms them, are important to a pitcher’s job, we would expect a strong relationship where a lower rate of mistake-making would mean a lower rate of allowing runs to score. I also did the same correlations for a couple of other metrics-POW, which we discussed earlier, and Walks plus Hits per Inning Pitched (WHIP), which Rose named as an inspiration for M/IP:

          RA       FRA       RA+
WHIP    0.826    0.837    -0.707
FIP     0.718    0.711    -0.638
M/IP    0.670    0.662    -0.503
POW     0.581    0.575    -0.410

All of these are rather strong correlations. POW is touted mainly for its predictive value, so it’s not too surprising that M/IP fares better across the board. Still, over the last 50 or so years, FIP shows a substantially stronger correlation to run scoring, and WHIP-which is basically pitcher’s OBP allowed-laps the field. So is “FIP-now with fewer strikeouts!” a good enough slogan to base an entire stat upon? If the argument in M/IP’s favor is that it’s more specific than FIP, does it have any benefit that you wouldn’t get by looking at stats like BB/9 or HR/9 separately?

To date, I haven’t seen any evidence that M/IP improves on FIP in terms of its predictive value, either. So while I’m generally in favor of having more statistical toys to play with, I don’t quite see where M/IP fits into the larger scheme of things. To clear my head, I ran M/IP on our database, to take a look at the top performances of all time:

Pitcher          Year    G    IP     HR/9  UBB/9  HBP/9   SO/9    RA     RA+   M/IP
Greg Maddux      1997   33   232.2   0.35   0.54   0.23   6.85   2.24   2.12   0.24
Greg Maddux      1994   25   202     0.18   1.25   0.27   6.95   1.96   2.55   0.25
Greg Maddux      1996   35   245     0.40   0.62   0.11   6.32   3.12   1.66   0.26
Greg Maddux      1995   28   209.2   0.34   0.86   0.17   7.77   1.67   2.93   0.27
Randy Jones      1978   37   253     0.21   1.57   0.00   2.52   3.70   1.08   0.27
Ron Reed         1975   34   250.1   0.18   1.65   0.14   5.00   4.24   1.02   0.28
Bob Friend       1964   33   229     0.28   1.30   0.16   4.68   3.46   1.17   0.28
Bill Gullickson  1981   22   157.1   0.17   1.72   0.23   6.58   3.09   1.30   0.29
Bob Ojeda        1988   29   190.1   0.28   1.47   0.19   6.29   3.50   1.14   0.31
Gaylord Perry    1968   39   290.2   0.31   1.46   0.12   5.3    2.88   1.18   0.31

Looking at the leaderboard, aside from Greg Maddux’s dominating 1994-1997 performances we see a couple of near league-average seasons, and then a few nice ones in the 1.14-1.30 RA+ range. Those are some respectable seasons from ground-ball pitchers with good control; but it’s a bit of a letdown when you consider that these are the top performances in M/IP over the last 50 years. Given Maddux’s peculiar monopoly on the top spots, maybe a better name for M/IP would be something like Maddux Pitching Index. It’d take a slight rescrambling of the abbreviation, and it still might not be an improvement over FIP, but at least the name would be more descriptive.