FanGraphs Baseball


RSS feed for comments on this post.

  1. Begin applause now.

    Comment by Ruki Motomiya — August 29, 2012 @ 9:16 am

  2. Very glad to see this. Thanks to you and the Dark Overlord.

    Comment by chuckb — August 29, 2012 @ 9:19 am

  3. Dave,

    Nicely done. I like having these kind of metrics to help explain the outliers comparing FIP to actual run prevention.

    The only thing I’m not clear about is whether the impact of defense falls into the BIP wins or bleeds into LOB wins. If the BIP wins are based on linear weights of singles vs expected and (2B+3B) vs expected, it look like the impact of errors vs. expected flow into the LOB wins.

    Comment by tz — August 29, 2012 @ 9:22 am

  4. When do you think Steve will have the glossary updated?

    Comment by Tyler — August 29, 2012 @ 9:27 am

  5. Thank you Dave. This is a great compromise, and I applaud you and the FanGraphs team for being innovative and flexible. I’m looking forward to seeing how these changes are implemented and how they affect pitching evaluation.

    Comment by Matt Hunter — August 29, 2012 @ 9:39 am

  6. nerdgasum

    Comment by Tom — August 29, 2012 @ 9:49 am

  7. Excellent job. It would be helpful for the site to have a “user’s guide” for pitcher statistics with links, as I now see the statistics here quoted regularly in the mainstream media without a full understanding of their meaning.

    Comment by Mike Green — August 29, 2012 @ 9:50 am

  8. Fantastic stuff! One question — Is a pitcher’s propensity to allow batters to reach base on error included in the hit prevention metric?

    Comment by craigtyle — August 29, 2012 @ 9:51 am

  9. This is excellent! I have been hoping something like this would be put together although I would have expected it to be based on batted ball data or swing/discipline data rather than hit/XBH and strand rates. However since FIP is about results (K/BB/HR) more than components of those results (GB%,swSTR%,HR/FB etc.) this makes sense.

    I’d like to see a SIERA implementation that factors in the batted ball profiles along with FDP input that could possibly be the ERA estimator/pitcher skill evaluator to beat them all.

    I’d also like to see if there is a correlation between O-swing%/O-contact% and FDP success, especially on bip.

    Comment by Scott Clarkson — August 29, 2012 @ 9:53 am

  10. Looks like Matt Cain finally won.

    Comment by Jack — August 29, 2012 @ 9:55 am

  11. My company hasn’t been hesitant about promoting dips, either.

    Comment by Well-Beered Englishman — August 29, 2012 @ 10:02 am

  12. Awesome, just awesome. Finally a crack in the wall surrounding this subject.
    The gap between FIP and explaining exactly what goes on just got smaller.

    Comment by Joeskil — August 29, 2012 @ 10:08 am

  13. This is why I love your website.

    That said, however, I fear that this still oversimplifies the ability of pitchers to mitigate good contact with the ball through good sequences (some of the credit on that belongs to catchers), ‘late’ movement, high velocity, and generally great ‘stuff’. We’ve all watched a dominant pitcher throw a game where the contact he induces tend to be weak ground balls or lazy fly balls, which, unless they are placed in exactly the right spots on the field, are likely to lead to outs. Similarly, poorly thrown pitches can get hit hard on the ground and be more likely to find holes, lined into the outfield, driven to the gaps, or pounded over the walls. While I understand that it’s difficult to isolate for this and that statistics can only do so much of a job to achieve this, I’d love to see some of the more sophisticated technology used for PitchFX used to read the speed of the ball off the bat, consider how far from a fielder’s starting position it was hit, and factor the time it takes for a ball to travel from the plate out to a fielder into the outcomes of a play to be able to identify the extent to which the outcome was caused by the pitcher, the batter, and plain old luck.

    This seems like a great step in the right direction to identify which pitchers have successfully mitigated runs through making it easier on their defenses, but taking it to the next level seems like the only way to truly identify a pitcher’s abilities.

    Comment by Kevin — August 29, 2012 @ 10:14 am

  14. Check out the Fangraphs Library in the Glossary tab near the top of the page.

    Comment by williams .482 — August 29, 2012 @ 10:16 am

  15. Defense certainly impacts both areas, as I note in the Jordan Zimmerman example in the follow-up post about FDP and pitcher WAR.

    To answer your question, though, LOB-wins is essentially a catch-all that quantifies the remainder of run prevention that isn’t captured by FIP or context-neutral hit prevention, so yes, reaching base on errors would be part of LOB-wins.

    Comment by Dave Cameron — August 29, 2012 @ 10:25 am

  16. These metrics are more about allocating past runs into specific bins than about projecting future runs allowed, but there is definitely a lot of room for adding various pitch-based components into a projection to determine whether we can get a better estimate of how much control of FDP a pitcher actually has.

    Comment by Dave Cameron — August 29, 2012 @ 10:27 am

  17. There’s no question that having access to the kind of information you’re talking about would be very helpful. However, right now, there is no publicly available source of speed-of-bat data for most contacted balls, and the tools that are measuring such things (HITF/x, Trackman, etc…) are made exclusively available to MLB clubs.

    Eventually, some of that data might become public, and we would jump at the chance to do the kinds of analysis you are talking about. For now, though, we hope that we can continue to do the best we can with the data we do have.

    Comment by Dave Cameron — August 29, 2012 @ 10:29 am

  18. I’m excited

    Comment by Carlos Collazo — August 29, 2012 @ 10:50 am

  19. Thanks for the reply Dave. I was curious as to the predictive vs. descriptive properties of the new metric. I’d love to see some studies about whether FDP can be used in any sort of predictive way or if we could come up w “xFDP”.

    Comment by Scott Clarkson — August 29, 2012 @ 10:51 am

  20. Interesting article. I wish we had some pitching in Cleveland, heck I wish we had a baseball team that could last the season in Cleveland….

    Comment by Cleveland Sports 360 — August 29, 2012 @ 11:02 am

  21. So, what is the different in FDP and ERA, in layman’s terms (i.e. something that might ever have a snowball’s chance in hell of making it to a game broadcast or mainstream publication)? The way FIP has been talked up and used over the years, it seemed like ERA pretty much *is* “fielding dependent pitching”.

    Comment by Snowblind — August 29, 2012 @ 11:18 am

  22. I’m slightly confused since hit prevention sounds a lot like what SIERA already accomplishes while still keeping the defense out of it. In a sense, it predicts BABIP. Why would we look at the results of balls in play when SIERA is already pretty good at this? The value of stranding runners may be one that is not encapsulated by SIERA, so I understand that, just not this BIP data.

    Or were you wary of using the somewhat unreliable batted ball data to calculate WAR?

    Comment by Tony Fernandez — August 29, 2012 @ 11:19 am

  23. “introducing fielding dependant pitching”

    this is a new and interesting way of saying

    “watching Jason Vargas pitch against anyone the twins roll out”

    Comment by illtakebothdakotas — August 29, 2012 @ 11:27 am

  24. Basically, you split ERA (or more accurately RA9, runs per 9 innings, that includes earned and unearned runs) into FIP and FDP.

    Comment by Tangotiger — August 29, 2012 @ 11:35 am

  25. Sidney Ponson also scores very high in other areas as well. His 9.27 GDP (gravy dependent pitching) figure is the second highest in the last decade, nestled comfortably between CC Sabathia (8.97) and Rich Garces (an amazing 10.18). Hideki Irabu scores very highly as well (7.74) despite his abbreviated career.

    Comment by Jason B — August 29, 2012 @ 12:26 pm

  26. @craigtyle

    I think Dave answered this above when he said:

    “…LOB-wins is essentially a catch-all that quantifies the remainder of run prevention that isn’t captured by FIP or context-neutral hit prevention, so yes, reaching base on errors would be part of LOB-wins.”

    Comment by Matthias — August 29, 2012 @ 12:56 pm

  27. no surprise at all about palmer, with that defense behind him

    Comment by jim — August 29, 2012 @ 1:44 pm

  28. Did fangraphs drop some sort of weird subconscious hint that this was going to happen? Because I had this idea last week and was going to ask Dave if it was any good in today’s chat. Maybe I can read Dave Cameron’s mind…

    Comment by Thomas — August 29, 2012 @ 1:50 pm

  29. Answer: “now introducing fielding dependant pitching”
    Question: “what should the twins public address announcer say if he forgets a pitchers name”

    Comment by illtakebothdakotas — August 29, 2012 @ 2:15 pm

  30. Answer: “now introducing fielding dependant pitching”
    Question: “what should the twins public address announcer say if he forgets a pitchers name”

    Comment by illtakebothdakotas — August 29, 2012 @ 2:15 pm

  31. Palmer is the biggest FIP outlier in the part of baseball history that at least resembles the game that is played today.

    Exactly what you’d expect when you consider who was behind him, notably Brooks Robinson, Mark Belanger, Bobby Grich, Cal Ripken, Jr., Davey Johnson and Luis Aparicio.

    Using Fangraph’s Fielding Runs data and approximating a win as 10 runs it looks like those players alone had about 53.3 fielding wins. The other players in that time period that I looked at, Boog Powell and Frank Robinson, were close to zero wins fielding. I’ll make a rough correction and assert that in that run environment, a win cost about 9 runs, so I’ll credit them for about 59.2 fielding wins.

    Palmer pitched about 14% of the time across that time period. From those players alone, pretending Palmer pitched the same number of innings every season and smearing the fielders’ performances across those innings, those players alone should account for about 8.3 fielding wins behind Palmer. That’s about a third of Palmer’s 28 Balls In Play Wins.

    Of course, the contributions of Robinson, Belanger and Grich in particular are being underestimated by smearing Palmer’s innings across his career instead of measuring out innings and contributions on a year by year basis.

    Taking 1975 as an example, Palmer pitched 323 innings, which we’ll call 22% of the season. Together, just Robinson, Grich and Belanger were worth about 7.6 fielding wins and so just those three players ought to account for 1.7 of Palmer’s 3.3 Balls In Play Wins.

    It’s nice that at, in this case at least, the, well, ballpark numbers add up like they should.

    Comment by Nathan Nathan — August 29, 2012 @ 3:31 pm

  32. From the chat:

    “…Johnny Cueto doesn’t give up HR with men on base, which leads to LOB-Wins…But not giving up HR with men on base *could* be a result of luck as well. ”

    “We’re not trying to factor out luck, we’re trying to factor out defense. Luck is inevitable, and we don’t take a hit away from a hitter just because he squibbed a ball in between three defenders and got a bloop single. ”

    Wouldn’t that answer suggest that RE24 and not wRAA should be used in WAR?

    Comment by Tonyc — August 29, 2012 @ 4:28 pm

  33. Dave,

    Have you guys looked at the year-to-year correlation of an individual player’s FDP stats?

    Phrased another way, do these stats contribute to separating luck from skill?

    Comment by Scott — August 29, 2012 @ 4:45 pm

  34. year-to-year correlation among pitchers that switched teams is going to be fairly low.

    It’s going to be a little higher for pitchers who don’t switch teams, because it’ll have the bias of the same-fielders and/or same-park.

    Comment by Tangotiger — August 29, 2012 @ 5:10 pm

  35. How are park adjustments done?

    Is there any issue in parks which have divergent run and HR factors (which I think is an issue with FIP).

    For example a park like Fenway which has a high run environment (which will “help” a pitcher when the adjustment is made to WAR for FIP-WAR) also has a low HR factor (which “helps” the FIP before the adjustment is done)

    Comment by Tom — August 29, 2012 @ 5:52 pm

  36. Dave, you wrote that “In looking at all 3,951 pitchers who have thrown at least 100 innings in the majors since 1963, the correlation between the FIP-based WAR and RA9-wins is .96.”

    What is the correlation between FIP-based WAR and just FDP?

    Comment by Eric — August 29, 2012 @ 6:58 pm

  37. For what it’s worth … Palmer NEVER gave up a grand slam … which says something about situational pitching and run prevention

    Comment by philcastle — August 29, 2012 @ 9:52 pm

  38. Tits article!

    Comment by bcp33bosox — August 30, 2012 @ 5:43 am

  39. FDP is an outstanding addition to the analytic repertoire, and I’m ecstatic to see it advanced to cover some of the blind spots in FIP and its like. I’ll add, Dave, that your exposition and choice of examples for this post are of exemplary clarity; this post will be a classic in the Fangraphs archives.

    FIP patently captures the _most important_ things a pitcher does in aggregate. BBs are bad; Ks are good. Producing them (or not) are primary skills. Changes in the rate of production correlate strongly with a pitcher’s results, over time and hence in aggregate. Situational pitching is a considerable part of performance, however, with real impact on runs, and so on wins. What kind of contact a pitcher yields beyond G/F also matters greatly to game and career outcome. FIP has been somewhat deaf to those qualities, but FDP looks to get the sense of them better.

    Dave: “Other pitchers do this by simply altering the way they pitch with men on base, increasing the amount of only-semi-harmful walks they allow in order to reduce the amount of very-harmful home runs they allow.” And by getting the groundball with men on base. FDP surely speaks to many kinds of outcomes, but above all we are likely getting a window on groundballs with men on base. Producing these is a _skill_, but GB% alone doesn’t fix sufficiently who can do this. Even flyball pitchers can delierately work for the GB in a game situation for instance. And it isn’t just a matter of ‘pitchers pitching better,’ since with men on base the batters to a (varying) degree may offer at pitches that they might not otherwise, something a pitcher can exploit. With men on base, a GB can matter _more_ than a K, which only nets one out when several may be needed. Keeping the ball in the infield generally _with men on base_ hasn’t been broken out well, but we may get the windo on it here. At last.

    A non-analytic aspect of FDP I particularly like is that it correlates with ‘visual inspection’ of performance. For Tom Glavine, for instance, FIP and hence WAR have never particularly put a shine on him, but as you highlight Dave, FDP isolates just why he was so valued by the teams and towns he pitched for. And the counter-example of Nolan Ryan is telling too. The knock on him until his late career excellence was that his results never matched his stuff. He walked many; he didn’t control the running game at all; there was some knock on him that he watched his stats more than the scoreboard or the standings for much of his career. He spent a lot of time pitching with men in scoring position, and didn’t shut own opponents’ scoring nearly as much as his K rates might imply. We see that in what FDP presents. And this stat allows those who, for whatever reason, want to compare the results of Ryan Franklin to Brandon Morrow to have a meaningful conversation about performance impossible in terms of FIP.

    A good step forward.

    Comment by Balthazar — August 30, 2012 @ 6:15 am

  40. To do this, we decided to calculate the linear weight value of singles and doubles — the difference between a double and a triple is almost certainly not due to something the pitcher did,

    For extra-base hits, shouldn’t you use a weighted average of the linear weight values of doubles and triples? E.g. if hitters generally hit four times as many doubles as triples the value should be (4 * w2B + w3B)/5. Otherwise you’re going to underestimate the value of preventing extra-base hits.

    Comment by Paul Clarke — August 30, 2012 @ 6:47 am

  41. And despite the lack of shine on Glavine’s fWAR, he still had close to 70.

    Comment by Matthew Cornwell — August 30, 2012 @ 10:04 am

  42. After extracting the defensive runs saved by his fielders, Palmer’s FDP is still one of the best all-time.

    Comment by Matthew Cornwell — August 30, 2012 @ 10:07 am

  43. You are correct that the run value should be the league-average blend of the two.

    For example, when I use wOBA, it’s 1.25ish for 2B, 1.60ish for 3B, but 1.3 for 2B+3B. You’d take about 10 parts 2B and 1 part 3B.

    Comment by Tangotiger — August 30, 2012 @ 1:31 pm

  44. Yes, that’s how it’s being calculated.

    Comment by Dave Cameron — August 30, 2012 @ 2:04 pm

  45. Ah, from this and David Appelman’s article it sounded like you were just using the value of a double. Thanks for the reply (Tango too).

    Comment by Paul Clarke — August 30, 2012 @ 3:26 pm

  46. You forgot about this guy:

    I’d love to see an article on the O’s defense of 1965-1975. They were the Murderer’s Row of fielding. I know defensive runs saved is a broad stroke and imperfect, but the O’s had three guys ranked in the top 12 all-time in fielding runs on the same team at the same time, plus an elite 2B in Grich. It’s like having Ruth, Bonds, and Musial in the same lineup. There’s a reason four guys won 20+ in 1971.

    Comment by GoToWarMissAgnes — August 30, 2012 @ 11:08 pm

  47. I did a quick scan, and a couple interesting things:

    Catfish Hunter is off the charts here compared to most other pitchers I could think of (didn’t check Palmer, though). That would seem to be expected.

    Tommy John had a slight positive impact. Which one would expect – he tended to make a lot of errors himself, which probably cancelled out some of the other good work he did stranding runners, controlling the running game, getting double-plays, etc.

    Comment by AsupporteroftheAs — September 7, 2012 @ 1:13 am

  48. Dave, I’m trying to replicate RA9-WAR and can’t quite make it work. I know you’re supposed to use runs allowed in place of adjusted FIP as the basis for RA9-WAR—but are you also supposed to use runs allowed when calculating the pitcher’s self-created run environment (as described here Or is this considered a constant (i.e., the “RtW” referenced in BIP-wins) for the pitcher across the board after it’s calculated from his FIP? Why or why not?

    Comment by Nathaniel — June 27, 2014 @ 1:08 am

Leave a comment

Line and paragraph breaks automatic, e-mail address never displayed, HTML allowed: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

Close this window.

0.096 Powered by WordPress