Baseball for the Thinking Fan

Login | Register | Feedback

btf_logo
You are here > Home > Baseball Newsstand > Baseball Primer Newsblog > Discussion
Baseball Primer Newsblog
— The Best News Links from the Baseball Newsstand

Sunday, June 18, 2006

Beyond the Box Score: Beamer: Why DIPS does what it does?

DIPSomania, running wild…with John Beamer.

Voros McCracken is one of the most heralded names in sabermetrics. That is because Voros is the founding father of one of the most revolutionary axioms in baseball: namely, hurlers have no control over balls in play! As avid readers of BtB I am sure that this isn’t news to you; in fact, it is almost a basic tenet of our baseball knowledge. But in 2001, just 5 years ago, this denouement was, in the words of Mao Zedong, a Great Leap Forward. Since then various analyses have shown that Voros’ sweeping conclusions, while generally true, don’t always hold. Indeed over many seasons, and also for a certain type of pitcher (eg, Knuckleballers), pitchers can control BABIP. An excellent article by Tom Tippet, of Diamond Mind, details the argument well.

Following this astonishing conclusion, Voros developed a statistic called DIPS ERA (dERA) to measure a pitcher’s true ability in naked, defense independent terms: strikeouts, walks, HBP and home runs. This is all well and good but DIPS is actually a very complex beast. What I want to explore for the remainder of this article is how the DIPS formula actually works.

Repoz Posted: June 18, 2006 at 11:44 PM | 100 comment(s) Login to Bookmark
  Tags: sabermetrics

Reader Comments and Retorts

Go to end of page

Statements posted here are those of our readers and do not represent the BaseballThinkFactory. Names are provided by the poster and are not verified. We ask that posters follow our submission policy. Please report any inappropriate comments.

   1. Buzzards Bay Posted: June 19, 2006 at 12:44 AM (#2068040)
site management eclipsed the voice of an Auburn grad that i'd like to hear now on this topic//i have only been around since 12/05 and i have missed a lot on this topic
   2. Gaelan Posted: June 19, 2006 at 01:14 AM (#2068067)
Why are we still talking about this. DIPS isn't true, it wasn't true and it never will be true.
   3. bibigon Posted: June 19, 2006 at 03:26 AM (#2068188)
Why are we still talking about this. DIPS isn't true, it wasn't true and it never will be true.


Stuff like this is why we're still talking about this.
   4. Jeff K. Posted: June 19, 2006 at 03:30 AM (#2068194)
Buzzards Bay, here is a good primer (no pun intended) on DIPS.

I'm as friendly with Backlasher as anyone outside the Union, but his orginal viewpoints (which he should be given credit for) have become fairly mainstream around these parts. You'll get them in other posts.
   5. Jeff K. Posted: June 19, 2006 at 03:34 AM (#2068200)
I should say, the Wiki link is pro-DIPS. I see that #2 has already chimed in with the basic outline of BL's thoughts on the subject.
   6. G.W.O. Posted: June 19, 2006 at 09:45 AM (#2068278)
Without wishing to express an opinion on DIPS either way, Voros' original methodology (comparison over correlations between consecutive points of a time series) is well known (in the maths/statistics world) to be deeply flawed. Tangotiger's methodology is a lot more sound.
   7. JPWF13 Posted: June 19, 2006 at 11:04 AM (#2068282)
Why are we still talking about this. DIPS isn't true, it wasn't true and it never will be true.

Because the truth lies somewhere between Voros' proclamations and the claims of his detractors

it's not a two choice answer: either
A: DIPS is right and pitchers have no influence over BABIP OR
B: DIPS is compleletyly 100% wrong and of no value.

The simple fact is
A: Given enough of a sample size the range in pitcher BABIP is much smaller than it is for hitters.
B: Given a samll sample size (ie: 1 year) a single individual pitcher's BABIP can fluctuate wildly from one year to the next- that sample individual pitcher's OTHER stats- K/BB, k/PA, etc will not fluctuate as much
C: A pitcher's DIPS ERA in year one will correlate better with said pitchers actual ERA in year 2 than will his actual ERA from year 1- which means that DIPS may not be "true" (as espoused by Voros) but is certainly useful and worth looking at.
   8. Andrew Edwards Posted: June 19, 2006 at 12:03 PM (#2068284)
JPWF13 has it right.

Whether DIPS holds some philosophical property of idealized Platonic truth-ness is not especially interesting. The point is that, knuckleballers excepted, DIPS gives us a vastly better projection of future performance that most other easily-calculated numbers. This makes it important.

DIPS changed the way I look at pitching statistics - I skim for K/9 and BB/9 and HR/9 and BABIP and see whether the pitcher's ERA is justified by his K, BB, and HR, or is caused by abnormal BABIP.
   9. G.W.O. Posted: June 19, 2006 at 12:06 PM (#2068285)
DIPS gives us a vastly better projection of future performance that most other easily-calculated numbers. This makes it important.
Is it really much better than career ERA?
   10. Buzzards Bay Posted: June 19, 2006 at 12:18 PM (#2068287)
Thanks for the link Jeff K.
   11. NedFlandersFields Posted: June 19, 2006 at 12:27 PM (#2068290)
Is it really much better than career ERA?


GWO - I'd say that depends in part on the length of career. With a young pitcher, I'll look at the components (I always have) to get a flavor of likely continued success. (That's how I built a decent roto pitching staff - always went for low-walk low-homer pitchers.) Also with a pitcher later in career, you sometimes wonder whether a low ERA masks the beginnings of decline, such that the "cliff drop" may be nearer.
   12. Slinger Francisco Barrios (Dr. Memory) Posted: June 19, 2006 at 12:45 PM (#2068301)
hurlers have no control over balls in play

I'm just a bit tired of this canard. The most you might be able to say is that pitchers have <u>equal</u> control over balls in play.

The difference is, put my way, it's a bit easier to swallow.
   13. BDC Posted: June 19, 2006 at 12:55 PM (#2068306)
pitchers have <u>equal</u> control over balls in play

Or perhaps "once the ball's in play, the fielders chasing it have more control over it than the guy who happened to throw the pitch?" which seems unexceptionable in any school of thought.
   14. Dizzypaco Posted: June 19, 2006 at 12:57 PM (#2068307)
I agree with JPWF13.

Is it really much better than career ERA?

It depends on what you are trying to measure. If you are trying to evaluate a player's career, I look at ERA - I don't like using DIPS for evaluating performance. However, if I'm trying to predict what a pitcher is going to do in the future, I find DIPS useful.
   15. Matt Clement of Alexandria Posted: June 19, 2006 at 01:19 PM (#2068318)
The point is that, knuckleballers excepted, DIPS gives us a vastly better projection of future performance that most other easily-calculated numbers. This makes it important.
It also doesn't "work" for closers. Or for pitchers with only a small amount of big league experience. The evidence is pretty good - see Tango's DIPS work - that pitchers have real differences in BABIP, even pitchers with significant major league time. Clay Davenport showed that minor league pitchers who fail to make the majors have higher BABIP than future major leaguers, so difference exists at that level quite clearly.

Basically, I look at DIPS as a rough regression stat. Generally, any time you add a regression factor to raw baseball stats, you improve their predictive utility. DIPS (or FIP, which is a superior calculation) does its regression somewhat idiosyncratically. It regresses two parts of a statline entirely to the league mean - hits on balls in play and scatter/cluster of offensive events - and leaves the rest of the statline untouched.

The problem is that our tools for measuring hard contact or clutch pitching are themselves very rough, and so we don't lose a lot of utility in ignoring these statistics, and, in general, the utility added by regression factors outweighs what is lost. I think that saying it "works" or that it is "better" or "worse" than other methods doesn't really get at the issues (although there are cases where it simply doesn't work.) It works for a specific reason - regression factors add utility to projections - and that doesn't necessarily tell us anything about the game on the field (ie. whether pitchers have "control" of BABIP).

I use FIP as one of the tools I look at. I definitely don't just ignore differences in BABIP, but I do recognize that stat doesn't measure very well pitchers' skill in preventing hard contact, and so it should be considered carefully depending on the size of the sample. And I recognize that for relievers and pitchers with little MLB experience, it gets more complicated.
   16. Gaelan Posted: June 19, 2006 at 01:38 PM (#2068332)
C: A pitcher's DIPS ERA in year one will correlate better with said pitchers actual ERA in year 2 than will his actual ERA from year 1- which means that DIPS may not be "true" (as espoused by Voros) but is certainly useful and worth looking at.


See that's just it. It only looks good when compared to a useless stat. No one uses ERA to project performance and no one should use ERA to compare performance.

DIPS doesn't tell you anything we didn't already know. All the weak version of DIPS says, which is all anyone bothers to defend these days, is that pitchers don't have total control over balls in play. Well that's just a pointless insight that everyone has always known. Everyone who watches baseball knows that the pitcher can't control whether the next texas leaguer is caught. If pitchers had total control over balls in play then there wouldn't be any need for defense.

And it's not just knuckleballers who have control over balls in play. All pitchers exhibit control over balls in play. Some, very good, pitchers are able to do it year after year. We can call those pitchers Greg Maddux but there are others.

Moreover the supposition that just because that some pitchers aren't able to consistently control the results over balls in play that the ability doesn't exist is by not self-evident. First of all there is so much noise in the data I wonder what definite conclusions can really be drawn. And even if the statistical conclusion is valid all this demonstrates is that mlb pitchers live on the margins of success. A very small change in the ability to control location results in large differences in performance. This explanation of performance variability also has the benefit of conforming with our micro level experience of actually watching baseball. Which I guess is what really bugs me. It's a false idea that negatively affects the visual experience of watching baseball.

DIPS is a bad idea that needs to be put to bed. Sabremetrics is the search for objective knowledge about baseball. DIPS is not objective and it's not knowledge. It's a litmus test for charlatans and fools.

Now if all you want to say that a .200 on BIP is not sustainable well then I agree but you don't need DIPS to know that.
   17. Mike Emeigh Posted: June 19, 2006 at 01:40 PM (#2068335)
C: A pitcher's DIPS ERA in year one will correlate better with said pitchers actual ERA in year 2 than will his actual ERA from year 1


It is fairly clear that the skill of preventing H/BIP is an important factor that managers use (probably not directly - more likely a result of evaluating other "qualitative" skills that are reflected in H/BIP) in determining whether or not pitchers can succeed in the majors. As has been cited before, Clay Davenport found that in the minors, pitcher advancement is predicated to a large extent on prevention of H/BIP. In the majors, pitchers who face few batters - as a group - tend to have the highest BABIP, and this is true both within a season (as I found) and across a career (as Tom Tippett found). Furthermore, when you look at pitchers as a group (as opposed to as individuals) within a season, those with low BABIP in a season will tend to have a better-than average BABIP the following season, and those with high BABIP in a season will tend to have worse-than-average BABIP the following season.

The reason that DIPS ERA correlates better than actual ERA has to do with the nature of the skill compared to the variance of BABIP. The degree of Y-T-Y variance in BABIP is affected not only by pitcher skill but by other factors (quality of hitters faced, quality of fielders, park and other environmental considerations, and so forth). These other factors that affect BABIP are not explicitly addressed in the "standard" DIPS model (although Tango et. al. did address them in "Solving DIPS"), but because, taken together, they tend to be larger in the aggregate than the effect of the pitcher, the net result is that DIPS, which regresses BABIP (nearly) 100% to the mean, is *more correct* than actual ERA, which does no regression of BABIP. A model that explicitly factors in the other considerations - filtering out the *true talent* of the pitcher from the other factors rather than regressing 100% to the mean, which is what Tango et. al. have tried to do - is likely to be *more correct* than DIPS.

The argument that BL, I, and others have raised is that it is inappropriate to conclude from Voros's analysis that pitchers have no effect on the results from balls in play. The further point that I am raising here is that, while DIPS ERA is *more correct* than a model of pitcher performance that does no regression of BABIP, it is not an appropriate model of pitcher performance, either. The other point that I want to make is that a pitcher with a high rate of allowing hits on BABIP (I typically use .320) should be viewed with suspicion regardless of his DIPS ERA, especially if he's also allowing a high rate of LDIP.

-- MWE
   18. JPWF13 Posted: June 19, 2006 at 02:12 PM (#2068350)
Also with a pitcher later in career, you sometimes wonder whether a low ERA masks the beginnings of decline, such that the "cliff drop" may be nearer.

I've also noticed that some veteran pitchers can maintain pretty good k/bb and k/9 #s well towards the end, but they just get hammered so badly on balls in play (think Lima, Kevin Brown at the end) that their peripherals are rendered meaningless.
   19. Slinger Francisco Barrios (Dr. Memory) Posted: June 19, 2006 at 02:15 PM (#2068352)
Or perhaps "once the ball's in play, the fielders chasing it have more control over it than the guy who happened to throw the pitch?" which seems unexceptionable in any school of thought.

That makes perfect sense. What seems to be harder for folks to take is that the batter has far more to do with it than the pitcher.
   20. JPWF13 Posted: June 19, 2006 at 02:27 PM (#2068355)
What seems to be harder for folks to take is that the batter has far more to do with it than the pitcher.

That's been "known" for quite some time- and disputed just as long.
It was observed 30+ years ago that the difference between the best and worst hitters at batting average was greater than the best and worst pitchers at opp batting average, that the difference between the best and worst hitters at hitting homers was greater than the difference between the best and worst pitcher at allowing homers, that the difference between the best and worst hitters at walking or striking out was greater than the difference between the best and worst pitchers etc etc...

which meant that the BATTER not the pitcher exerted more of an influence on the outcome of an at bat.

and most people refused to believe it, even those who saw the evidence.
   21. GuyM Posted: June 19, 2006 at 02:34 PM (#2068359)
Post 15 makes an important observation: DIPS essentially does this regression:
BABIP 100%
HR 0%
K 0%
BB 0%
We know that all 4 of these are wrong. A better DIPS would regress all 4 factors the correct amount given a pitcher's BFP (with BABIP of course being regressed most heavily). Whether that would be much more accurate in practice than just regressing total R/G, I'm not sure, but it has to be more accurate than DIPS. David Gassko could do the same thing with DIPS 3.0, regressing each batted-ball-type (though again, the additional accuracy may not be worth the complexity).

That said, I continue to find it hard to see the utility of evaluating pitchers on a single-season basis, rookies excepted. The idea that we should regress Pedro or Santana or Mariano to the league mean at this point -- for any of the four variables --seems rather silly.

MWE provides a great summary, as usual (but Mike: when will we see your article from last summer??). I would just add two observations:

1) Within the reliever population, good pitchers seem to have a more clear BABIP advantage over weak pitchers than among starters. This is consistent with MWE's point that much of the DIPS phenomenon reflects the fact that MLB pitchers have been pre-selected for their ability to prevent HBIP. Since the range in talent among relievers is much greater than among starters -- the best/worst starters are probably #1 and #6 on your staff in terms of talent; best/worst relievers are #2 and #12 -- it would follow that the compression of BABIP ability is less extreme.

2) What gets called "small" differences in hit prevention ability by DIPS proponents are not in fact small. A 5% change in HR/9, K/9, or BB/9 is trivial, but a 5% change in BABIP (.015) makes a difference of .35 R/G. Differences in hit prevention talent are "small" in the sense they are hard to detect given the statistical noise, but they are not small in terms of a pitcher's success. If you look at the Davenport data, it becomes clear that in terms of what separates a MLB pitcher from a AAA pitcher, the ability to prevent hits on BIP is nearly as important as getting Ks or preventing BBs or HRs. This is the part of the story I never hear DIPS fans acknowledge or address.
   22. Andrew Edwards Posted: June 19, 2006 at 02:50 PM (#2068361)
No one uses ERA to project performance

Are you kidding? The list of free agent picthers who've signed huge contracts based on a fluky low-BABIP year is as long as my arm. You mean "no sabermetrician", which is fine, but come to my other point:

A model that explicitly factors in the other considerations - filtering out the *true talent* of the pitcher from the other factors rather than regressing 100% to the mean

Yes, this would be better. But I can't do a rough approximation of that in my head based on a standard ESPN stat line. I can with basic DIPS, though, roughly calculate how 'lucky' on BIP a pitcher has been and so get to a good (not perfect) sense of his future performance. Baseball journalists and managers, with few exceptions, don't have the statistical chops to 'get' more complex regressions, but they can be made to 'get' the suggestion "among MLB pitchers, look at K/9, BB/9 and HR/9 to get a decent idea of their true potential".

DIPS is not perfect, but it works pretty well. It is a hugely useful and meaningful 80/20 pitcher evaluation rule (vastly better than just 'past year performance') and to wave it off in favour of messy SPSS models that are theoretically better is straight ivory-tower snobishness.
   23. The Greatest Game on Earth Posted: June 19, 2006 at 02:52 PM (#2068365)
Clay Davenport showed that minor league pitchers who fail to make the majors have higher BABIP than future major leaguers, so difference exists at that level quite clearly.

I haven't read that research, but that smacks of possible selection bias. Perhaps some major leaguers became major leaguers because lucky minor-league BABIP rates made them appear better than they were? Such comparisons are shaky because players do not reach the majors on an objective basis.
   24. JPWF13 Posted: June 19, 2006 at 02:57 PM (#2068368)
That said, I continue to find it hard to see the utility of evaluating pitchers on a single-season basis, rookies excepted. The idea that we should regress Pedro or Santana or Mariano to the league mean at this point -- for any of the four variables --seems rather silly.


It may seem silly, but it's not, even pitchers like Pedro, Santana and Mariano can bounce around a lot from season top season-

ok, if a guy has 500+ career ip it would make more sense to regress them to their individual established mean
   25. G.W.O. Posted: June 19, 2006 at 03:09 PM (#2068369)
I think the problem here is reductionism (and Voros is the worst offender). There are interesting things to be said about BABIP variability, and using regressubg BABIP to detect fluke year. The problem lies when we turn it into a formula for DIPS, or a simplification like Voros' original claim, and pretend that that either the formula or the claim is the unvarnished truth.

It's an interesting fact that even Greg Maddux exert as much control on BABIP as one might think, and that the defense behind him are massively more important, but we can't expect interesting facts to map smoothly onto equations.
   26. GuyM Posted: June 19, 2006 at 03:13 PM (#2068371)
I haven't read that research, but that smacks of possible selection bias.

I'm sure a few pitchers get promoted based on lucky BABIP results, but it's totally implausible this could account for Davenport's results. First, he finds the same spread between MLB-bound pitchers and non-MLB pitchers at every single level of minor league ball. Second, as I said, the AAA/MLB difference in BABIP is nearly as great as for K/9, BB/9, or HR/9. If there were no difference in true talent, the BABIP difference would be much smaller, even given some selection bias.

even pitchers like Pedro, Santana and Mariano can bounce around a lot from season top season-

Of course they can. But people often jump from that fact to conclude "BABIP can't be a very important skill, if a guy is among the league leaders one season and near the bottom the next season." That is mistaken. Pedro's ability to prevent HBIP is at work every year, whether he's lucky or not (or has good or bad fielders behind him). In his lucky years he may be at .240 and in his bad years .310, but in both years he's 25 points or so better than an average pitcher with equivalent luck. Every single season, it's a big part of his success. There's nothing a pitcher or a GM can do about the luck -- it's part of the game. And no matter how much y-t-y variation there is, you still want the guy with the best ability to prevent HBIP.
   27. Matt Clement of Alexandria Posted: June 19, 2006 at 03:14 PM (#2068372)
It may seem silly, but it's not, even pitchers like Pedro, Santana and Mariano can bounce around a lot from season top season-
Right, but that's true of Ks, BBs, HR, runs allowed, and all that. It's why we use regression factors in projection, and increased use means increased predictive utility.

DIPS is just a sorta strange regression stat. The thing that makes it useful, to me, is that it's readily available and FIP is a ridiculously easy little formula. There's nothing essential about the regression factors in DIPS that makes them better than many other regression factors, but those are the ones we have easily google-able for a variety of reasons that have pretty much nothing to do with hte superiority of DIPS over those other stats.
I haven't read that research, but that smacks of possible selection bias. Perhaps some major leaguers became major leaguers because lucky minor-league BABIP rates made them appear better than they were? Such comparisons are shaky because players do not reach the majors on an objective basis.
I find this highly unlikely, because, as Mike said, the size of the difference in BABIP is about equivalent to the size of hte difference in other stats - K, BB, etc. The "selection bias" argument works for all those numbers, and so it doesn't really capture the issue unless you want to argue that future major leaguers are no better than their career minor league colleagues.
   28. base ball chick Posted: June 19, 2006 at 03:19 PM (#2068377)
Buzzards Bay Posted: June 18, 2006 at 08:44 PM (#2068040)

site management eclipsed the voice of an Auburn grad that i'd like to hear now on this topic//i have only been around since 12/05 and i have missed a lot on this topic

- did he get banned from the site?????
   29. Rally Posted: June 19, 2006 at 03:36 PM (#2068394)
ok, if a guy has 500+ career ip it would make more sense to regress them to their individual established mean

No, you still regress to the league average, just regress less as the pitcher has more batters faced. That applies as long as you are going just on statistics.

You can also regress Pedro to a sample that he represents, pitchers with great control and killer changeups or something.
   30. DCA Posted: June 19, 2006 at 03:45 PM (#2068398)
hurlers have no control over balls in play

I'm just a bit tired of this canard. The most you might be able to say is that pitchers have equal control over balls in play.

Or perhaps "once the ball's in play, the fielders chasing it have more control over it than the guy who happened to throw the pitch?"


This is why DIPS was so important. Nobody thinks (a), and when they say so it's just a poor choice of wording for saying (b). And I doubt anyone would complain about (c). But the thing is that it took DIPS for us to figure this out: pitchers control BIP a lot less than we thought they did -- even if there is still some control. All of the more detailed investigative work on the topic is a direct result of DIPS.

We don't think JJ Thompson was an idiot and doing active harm to the study of physics because his atomic model wasn't right. Clearly it's less important, but I think Voros' DIPS theory is the same kind of necessary incremental step. Those who would dismiss DIPS as Gaelen does in #2 would be going backward to Dalton, not forward to Rutherford. And that would be the real mistake.
   31. greenback wears sandals on his head Posted: June 19, 2006 at 03:46 PM (#2068399)
That applies as long as you are going just on statistics.

As I watch Mark Mulder's arm slowly fall off, this seems like a good point to raise.
   32. Steve Treder Posted: June 19, 2006 at 03:55 PM (#2068406)
But the thing is that it took DIPS for us to figure this out: pitchers control BIP a lot less than we thought they did -- even if there is still some control. All of the more detailed investigative work on the topic is a direct result of DIPS.

Exactly right. And the physics analogy is apt.

Despite its every flaw, McCracken's DIPS work represented a tremendous insight and breakthrough. It moved the ball a long way forward.
   33. Matt Clement of Alexandria Posted: June 19, 2006 at 03:58 PM (#2068409)
But the thing is that it took DIPS for us to figure this out: pitchers control BIP a lot less than we thought they did
I just don't think this is true. I think that DIPS showed that our current ability to measure the difference among pitchers to control BIP is less than we thought it was. Just a small amount of BABIP is a significant amount of runs, and so even differences that are hard to measure can be just as meaningful as differences that are easy to measure.
We don't think JJ Thompson was an idiot and doing active harm to the study of physics because his atomic model wasn't right. Clearly it's less important, but I think Voros' DIPS theory is the same kind of necessary incremental step. Those who would dismiss DIPS as Gaelen does in #2 would be going backward to Dalton, not forward to Rutherford. And that would be the real mistake.
Here I agree. I think BL stated it as "Voros as muse."

His investigation into regression-factor statistics led to a lot of really good work on the topic, even though that work subsequently showed that the original DIPS studies were quite flawed.
   34. Jeff K. Posted: June 19, 2006 at 03:59 PM (#2068410)
Completely agreed with point (c) from DCA and with Treder here. DIPS was way overstated, but it was important. Denying that is foolish.
   35. GuyM Posted: June 19, 2006 at 04:01 PM (#2068412)
This is why DIPS was so important. Nobody thinks (a), and when they say so it's just a poor choice of wording for saying (b). And I doubt anyone would complain about (c).

Except, "b" is also wrong, as documented in multiple posts here.

And "c" probably isn't correct either. The variation among team defenses in terms of true DER ability is probably less than the variation in pitchers' true BABIP ability. The "Solving DIPS" paper by Allen, Hsu, and Tangotiger concluded a pitcher's BABIP was 44% luck, 28% pitcher ability, 17% fielding, 11% park. I'm not sure that's exactly right, and they were tentative about the conclusion. I think most everyone would agree that luck is most important factor in any single year's BABIP, but it's by no means clear that defense is any more important than a pitcher's ability.

But the thing is that it took DIPS for us to figure this out: pitchers control BIP a lot less than we thought they did

I'd agree with that.
   36. Mike Emeigh Posted: June 19, 2006 at 04:15 PM (#2068420)
DIPS is not perfect, but it works pretty well. It is a hugely useful and meaningful 80/20 pitcher evaluation rule (vastly better than just 'past year performance') and to wave it off in favour of messy SPSS models that are theoretically better is straight ivory-tower snobishness.


I'm not suggesting that, at all. What I am suggesting is this:

1. DIPS works well for identifying flukish "good" seasons based on a low BABIP (Joe Mays, 2001).
2. You need more information to determine whether a "bad" season based on a high BABIP is flukish (Randy Johnson, 2003) or legitimately bad. Without such information, you're usually better off assuming that it was legitimately bad. With the Unit in 2003, he was pretty clearly hurting, and it stood to reason that he'd improve when he got healthy.

I don't have much problem with regressing a starting pitcher who posts a .250 BABIP in a given season 100% to league average, because we *know* (as much as we possibly can know) that no one has a true level of talent that good, and therefore league average is a better estimator than .250 (not ideal, just better). The problem occurs at the other end. We don't *know* what a guy who posts a .320 BABIP can do. We do *know* that to be consistently successful in the majors he has to do better than that, because there are few examples of pitchers with career BABIP at that level who have long careers, and thus it makes sense to project a regression to the mean. But we don't really know whether he actually will do that, or whether he'll be one of the larger group of pitchers with high BABIP who wash out of the majors.

Bottom line: you can assume a pitcher with a good season based primarily on a low BABIP was *lucky*, you should not assume that a pitcher with a poor season based primarily on a high BABIP was *unlucky*.

-- MWE
   37. JPWF13 Posted: June 19, 2006 at 04:17 PM (#2068424)
To tunr left a bit, I see Voros' initial proclamation that pitchers have zero ability to influence BABIP to be analogous to James' 1984/85 assertion that MLEs were EXACTLY as useful in predicting mlb performance as MLB performance was.

Both were extreme overstatements that unfortunately masked something important.

In the case of mles- the "truth" was that minor league stats when properly adjusted for park and league (and age- and years at a level etc) were very valuable and useful- not as good as mlb stats, but far more useful than previously believed. (previously they were seen as having no predictive value, the only reason a team or scout paid attention was to see if a player was progressing the way they wanted or thought a player should- they just didn't think of translating a AAA line into an mlb line- A, AA, AAA were seen as being so different from mlb in so many reaspects that linear comparisons were impossible).

In really gross terms, what DIPs does is this- take 2 pitchers, both going 15-10 with an era of 3.00. One has more Ks, less BB and less Hrs and a BABIP of .310, the other guy has a BABIP of .275- well the first guy is more likely to sustain that success

prior to DIPs, most observers (scouts, fans, statheads etc) would prefer the second guy (well the statheads would prefer the first guy for the high K rate)- why? obviously the second guy knows how to pitch, whereas the first guy is a thrower.
Look at old Baseball Digests from the 70s, or Sporting News, etc.
The Randy Jones/ Mark Fydrich's of the world were (until they cliff dived)were seen as somehow better than the harder throwing guys.
The guys who preferred the hard throwers were seen as throwbacks (sure the scouts, correctly, always preferred the hard throwers, but bizarrely the columnists who tend to defend scouts, tended to part company with the scouts on this one, and would wax eloquently on how baseball had an unfair bias agaisnt control pitchers and preferred throwers...)
   38. Mefisto Posted: June 19, 2006 at 04:18 PM (#2068426)
did he get banned from the site?????

Yes. I got the impression it's temporary, but nobody, to my knowledge, has said how long.
   39. Joyful Calculus Instructor Posted: June 19, 2006 at 04:23 PM (#2068431)
I've also noticed that some veteran pitchers can maintain pretty good k/bb and k/9 #s well towards the end, but they just get hammered so badly on balls in play (think Lima, Kevin Brown at the end) that their peripherals are rendered meaningless.

And I think pitchers who aren't completely healthy are the same way (think Randy Johnson 2003).
   40. Jeff K. Posted: June 19, 2006 at 04:24 PM (#2068432)
Yes. I got the impression it's temporary, but nobody, to my knowledge, has said how long.

One month, starting from last Tuesday.
   41. Rally Posted: June 19, 2006 at 04:52 PM (#2068453)
Yes. I got the impression it's temporary, but nobody, to my knowledge, has said how long.

One month, starting from last Tuesday.


Who got banned? What did he do?

Post 15 makes an important observation: DIPS essentially does this regression:
BABIP 100%
HR 0%
K 0%
BB 0%


Which is more accurate than using no regression, taking previous stats at face value.

What would be more accurate, at least in theory, is regressing hits 75%, HR 50%, Walks 25%, and K's 10% - or something like that, I don't remember the actual figures.

I tried something like this last year, and while theoretically sound, it did NOT predict future ERA any better than DIPS did. I was surprised by this result. So that's what DIPS is - simple, flawed, but gets the job done.
   42. The Balls of Summer Posted: June 19, 2006 at 05:01 PM (#2068462)
Who got banned? What did he do?

Backlasher. See the thread below.

The thread in question
   43. More Dewey is Always Good Posted: June 19, 2006 at 05:14 PM (#2068475)
prior to DIPs, most observers (scouts, fans, statheads etc) would prefer the second guy (well the statheads would prefer the first guy for the high K rate)- why? obviously the second guy knows how to pitch, whereas the first guy is a thrower.
Look at old Baseball Digests from the 70s, or Sporting News, etc.
The Randy Jones/ Mark Fydrich's of the world were (until they cliff dived)were seen as somehow better than the harder throwing guys.
The guys who preferred the hard throwers were seen as throwbacks (sure the scouts, correctly, always preferred the hard throwers, but bizarrely the columnists who tend to defend scouts, tended to part company with the scouts on this one, and would wax eloquently on how baseball had an unfair bias agaisnt control pitchers and preferred throwers...)


Of course, there are and were pitchers who carved out outstanding careers for themselves despite not having great peripherals. At some point, you've got to acknowledge that Jim Palmer's career wasn't one long fluke.

I also think there's a selection bias going on, as well - I think that pitchers with good "stuff", who can strike out a ton of guys, are going to get more opportunities to pitch over rough patches than guys who depend on balls in play.

I think DIPS is a neat little tool, but it's utility is limited - there simply are pitchers (and more than just a couple of knuckleballers) for whom it doesn't apply.
   44. Rally Posted: June 19, 2006 at 05:17 PM (#2068479)
Of course, there are and were pitchers who carved out outstanding careers for themselves despite not having great peripherals. At some point, you've got to acknowledge that Jim Palmer's career wasn't one long fluke.

I haven't looked too closely at Palmer's career with DIPS or anything, but I wonder if having people like Brooks Robinson, Mark Belanger, and Paul Blair had a lot to do with his success.

Not to say that he wasn't a great pitcher, but he sure had a lot of help on those great teams.
   45. Rally Posted: June 19, 2006 at 05:27 PM (#2068485)
Backlasher. See the thread below.

Damn that's a shame. Insult the Jim and you get banned. That's weak.

Oh well, if I get banned now too, see y'all in a month.
   46. GuyM Posted: June 19, 2006 at 05:29 PM (#2068486)
I tried something like this last year, and while theoretically sound, it did NOT predict future ERA any better than DIPS did. I was surprised by this result. So that's what DIPS is - simple, flawed, but gets the job done.

But how much better is DIPS than just regressing ERA to the mean? That's even more simple, and I suspect it gets the job done nearly as well.

In any case, how often do we need to predict next year's ERA/RA based only on last year's data? If that's the challenge, then sure, DIPS is fine. But if you've got at least a few years worth of data, you can do much better than DIPS. Certainly, "career DIPS" should be considered an oxymoron.
   47. Joey B. is counting the days to Trea Turner Posted: June 19, 2006 at 05:34 PM (#2068489)
Backlasher got banned before Kevin? I would never have seen that one coming.
   48. DCA Posted: June 19, 2006 at 05:36 PM (#2068492)
Certainly, "career DIPS" should be considered an oxymoron.

Why? Isn't that just "career pitching value of K, BB, and HR outcomes" reported on an ERA scale? The difference between that and ERA is park/league adjustment, defense, luck, BABIP skill, and situational pitching. And I expect for most pitchers would be pretty small.
   49. JPWF13 Posted: June 19, 2006 at 05:39 PM (#2068497)
Oh well, if I get banned now too, see y'all in a month.

nah, you have to call him Nellie...

I haven't seen a post from Kevin in awhile... (but then I don't read PED threads any more...)
   50. GuyM Posted: June 19, 2006 at 06:00 PM (#2068510)
Why? Isn't that just "career pitching value of K, BB, and HR outcomes" reported on an ERA scale?

Because with many years of data, you can do a much better estimate of a pitcher's true BABIP than just assume league average. Why would you want to know the combined effect of K/BB/HR, any more than K/HR/BABIP, or K/BB/BABIP, or any other combination? I guess I can imagine very specific reasons you might calculate career DIPS as part of a study, but I think it would be wrong to treat it as a better measure of a pitchers' "true talent" than, say, ERA+.

I think the nail in the coffin for the notion that there are only "small differences" among pitchers on preventing HBIP will come when someone does a careful analysis of the effect of age on BABIP. If you believe in DIPS, there should be at most a small effect of age on BABIP. But I don't believe that is true. Greg Maddux was .267 from age 22 to 33, .286 from 34 to 39. I suspect this is typical. Tango has done a study suggesting that BABIP generally rises with age. It's not definitive, and it's difficult to measure this given the selective sampling issues, but I think we'll eventually discover a pretty clear pattern. And we may even learn that (as suggested in post #18), a sharp rise in BABIP is a good indicator an aging pitcher has reached the end of the line (though I'm sure is easier to see in hindsight than at the time).
   51. Traderdave Posted: June 19, 2006 at 06:00 PM (#2068512)
Damn that's a shame. Insult the Jim and you get banned. That's weak.

This is Jim's site. Insult a man in his house, expect to be shown the door. Given all the hassles that BL caused Jim, I'd say Jim showed unusual patience.
   52. DCA Posted: June 19, 2006 at 06:17 PM (#2068528)
Why would you want to know the combined effect of K/BB/HR, any more than K/HR/BABIP, or K/BB/BABIP, or any other combination?

The same reason Voros did. There's a long history of grouping the three true outcomes together on the offensive side. Voros wanted to separate pitching outcomes into two bins: only the pitcher and hitter are involved and pitcher, hitter, and fielders are involved. K/BB/HR go together naturally. Maybe that product is a bit artificial, but it at least has some historical background and logical framework.
   53. Biscuit_pants Posted: June 19, 2006 at 06:29 PM (#2068537)
Insult a man in his house, expect to be shown the door.
True, but I do not think it applies here. Very few threads go by without insults somewhere and yet nothing is done until they are directed at him THAT is what he means by weak.
   54. GuyM Posted: June 19, 2006 at 06:37 PM (#2068540)
Bottom line: you can assume a pitcher with a good season based primarily on a low BABIP was *lucky*, you should not assume that a pitcher with a poor season based primarily on a high BABIP was *unlucky*.

This is a great observation (unless you're talking about Andy Messersmith or someone else with a demonstrated ability to post very low BABIPs). Perhaps you've already done this, MWE, but it would be interesting to look at the pitchers whose career ERA significantly exceeds their career DIPS ERA. My guess is that most or all have short careers, and flushed out because they couldn't prevent hits. These are essentially the short-career pitchers that Tippet found had high BABIPs. One example who kept getting opportunities, presumably because of his good DIPS stats, is Steve Woodard: 3.1 K/BB ratio and career FIP of about 4.37, but actual ERA of 4.94.

Some will argue that these pitchers were just the victims of bad luck and small sample sizes, but when you consider the desperate need for pitching by almost all organizations, and the opportunities pitchers have to redeem themselves in the minors and earn another promotion, that just isn't plausible. Clearly, many of these short-career pitchers just couldn't prevent hits on BIP. And if you can't do that, you can't pitch in MLB.

There's a long history of grouping the three true outcomes together on the offensive side.

Yeah, but just for fun. Players differ considerably in their BABIP and SLGBIP -- you can evaluate a hitter meaningfully using just the TTOs.
   55. Rally Posted: June 19, 2006 at 06:39 PM (#2068542)
Insult a man in his house, expect to be shown the door.

I wouldn't have reacted that way. He calls Jim "Nellie" because Jim shares a lastname with a famous singer. He'll twist anyone's name in a similar fashion. I'm sure you all remember Cadbury, Neosporin, and plenty of others. I don't know how anyone familiar with him can take that too seriously. He's kind of like Sawyer on Lost, though the BL has been at it longer than that show's been on. I know sometimes you'd like to smack him through the computer screen, but sometimes you just can't wait to see what he'll come up with next.

Not that anyone reads my blog anyway, but if you do and insult me, I'm fair game. I wouldn't ban anyone unless they go after my girlfriend or threaten to sacrifice my cat to Chris Truby.

To each his own though.
   56. Rally Posted: June 19, 2006 at 06:43 PM (#2068547)
But how much better is DIPS than just regressing ERA to the mean? That's even more simple, and I suspect it gets the job done nearly as well.

I think I tried that too. Maybe didn't use actual ERA, but used component ERA, with each category regressed exactly the same. (50% I think) It didn't beat DIPS.
   57. Ron Johnson Posted: June 19, 2006 at 06:50 PM (#2068554)
I think DIPS is a neat little tool, but it's utility is limited - there simply are pitchers (and more than just a couple of knuckleballers) for whom it doesn't apply.


James had a useful example in the new HA. The difference in career expected hits allowed between Bob Gibson and Tommy John almost certainly isn't random.

But I think "doesn't work" is an overbid. In the sense that the imperfect estimation of hits is (in general -- there will be a few exceptions) something like the 5th biggest source of error in projecting future performance.

(ability to control running game, DP support, HR rate, runners reaching on error(?) and luck [timing, clutch pitching -- you choose])

Always assuming a pitcher stays healthy.

And yes, I agree that with PBP data we can do better than DIPS. I think the works Studes and company are doing is just awesome -- and I've suggested that (once the issues with the data source are ironed out) the next thing to look at is the interaction between the count (rather that K rate) and balls in play.
   58. Joyful Calculus Instructor Posted: June 19, 2006 at 07:01 PM (#2068565)
Not that anyone reads my blog anyway

Post a link to it. Anyway, I don't want to read a whole 300+ post thread (unless it's the Everett or Onion thread) so could somebody tell me where on the thread are the remarks made and the banning made. Backlasher must have said something really bad. I've pissed off a ton of people without ever getting banned.
   59. GuyM Posted: June 19, 2006 at 07:02 PM (#2068568)
used component ERA, with each category regressed exactly the same. (50% I think) It didn't beat DIPS.

That states the question backward (I think). Of course DIPS is a better predictor than ERA alone. The question is: does DIPS beat ERA regressed to the mean? And if not, what good is it really?
   60. Rally Posted: June 19, 2006 at 07:08 PM (#2068573)
I don't want to read a whole 300+ post thread (unless it's the Everett or Onion thread) so could somebody tell me where on the thread are the remarks made and the banning made. Backlasher must have said something really bad.

Backlasher called Jim "Nellie". That's it. Nothing else was heard from BL, JC informed us that BL had been banned.

I think it all happens on page 2-3 of the thread. Just open it up and search for "backlasher" and you won't have to read too much of the thread.
   61. Barry`s_Lazy_Boy Posted: June 19, 2006 at 07:09 PM (#2068574)
Not that anyone reads my blog anyway

If you post a link to it, BL will forever rag about it, even when it has no relevance to the matter at hand.

I'm sure you're aware of this behavior.
   62. Rally Posted: June 19, 2006 at 07:17 PM (#2068590)
I won't bother posting a link. If you want to find it you can click on my screen name and follow the homepage link. But there's a reason I don't advertise, I don't post enough content. Nothing interesting since just before Jered Weaver was called up.
   63. Rally Posted: June 19, 2006 at 07:21 PM (#2068597)
I'm sure you're aware of this behavior.

Why certainly. Danny once posted a link to sim league stats by accident, which he immediately apologized for when it was pointed out. Of course in Backlasher's limited bag of tricks this was enough to discredit anything Danny ever said, until the end of time.

He can be a (JJ) Putz, but I don't think he should be banned. I know its not my call, just my 2 cents.
   64. Jack Keefe Posted: June 19, 2006 at 07:34 PM (#2068620)
Yes well I have a blog and I have not wrote much on it either ladley but I will link to it its called Jack Keefes Blog and if any 1 makes fun of it I will punch them in the kisser.
   65. Joyful Calculus Instructor Posted: June 19, 2006 at 07:37 PM (#2068626)
The name Nelly was used in both post 148 and 197. In post 148, BL criticized Jim for encouraging bashing of the anti-sabres while supporting the sabres. JC said he was banned for calling him Nelly, but then Jim actually posted that BL misrepresented his opinion. BL's critiques were somewhat out of line.
   66. Barry`s_Lazy_Boy Posted: June 19, 2006 at 07:44 PM (#2068634)
He can be a (JJ) Putz, but I don't think he should be banned. I know its not my call, just my 2 cents.

I agree that he shouldn't be banned, but I am hardly surprised after his constant baiting of Jim.

JC said he was banned for calling him Nelly

I'm sure the banning was a long time coming, and I doubt any single item caused it.

It's disheartening, cuz he had be a lot more rational lately. His reasoned, non-attacking mode can actually accomplish what he thinks he is: which is changing the line of thought by many on this board.
   67. Dan Szymborski Posted: June 19, 2006 at 09:33 PM (#2068778)
Backlasher got banned before Kevin? I would never have seen that one coming.

Me neither. I was quite suprised to find out that BL had been suspended.

Kevin hasn't been banned - if what I was told is accurate, he's boycotting until BL's back.
   68. Chris Dial Posted: June 19, 2006 at 10:13 PM (#2068813)
Kevin hasn't been banned - if what I was told is accurate, he's boycotting until BL's back.

JC has been out of site a while too.

And it's been nothing but a love-in since (not).

I think the physics analogy is particularly apt - even not exactly right, it changed the way people viewed pitchers, and created a whole new area of study.
   69. JPWF13 Posted: June 19, 2006 at 10:30 PM (#2068824)
Kevin hasn't been banned - if what I was told is accurate, he's boycotting until BL's back.

bad move, that might be seen as reason enough to extend the ban.
   70. Traderdave Posted: June 19, 2006 at 10:43 PM (#2068832)
JC said he was banned for calling him Nelly

The way that thread reads, I get the feeling there were some deleted posts. There may have been stronger words than Nelly, which does seem a petty reason to ban. But in any case BL had given Jim many reasons for banishment well before that.
   71. Dr. Vaux Posted: June 19, 2006 at 11:07 PM (#2068849)
What the hell dos 'Nelly' mean in that context anyway? You probably shouldn't answer that, though, since it could get you banned.

But I must say, if I haven't already, that nobody always agrees, so even though I disagree with the banning of Backlasher it doesn't change my appreciation of the site or Jim's hard work to provide it.
   72. GGC:BTF's Biggest Underachiever Posted: June 19, 2006 at 11:28 PM (#2068894)
I like Bl and understand the ban. But Nelly is in the long line of mock handles he used like "The Jim", "Trederdave", and "Cadbury".

JC has been out of site a while too.


I saw him on the Arod thread over the weekend. Andy's around, too.
   73. DCA Posted: June 19, 2006 at 11:42 PM (#2068916)
Who was Cadbury? I can't even think who that would be.
   74. HCO Posted: June 19, 2006 at 11:51 PM (#2068927)
LOL.
   75. greenback wears sandals on his head Posted: June 19, 2006 at 11:51 PM (#2068930)
JC Bradbury of Sabernomics fame.
   76. GGC:BTF's Biggest Underachiever Posted: June 19, 2006 at 11:55 PM (#2068935)
I like Bl <strike>and</strike> but understand the ban.
   77. jmp Posted: June 20, 2006 at 12:26 AM (#2068987)
In 143 of the thread BL wrote "I have found The Jim's skills lacking in even being able to understand the topic of conversation".

That seems to be a baseless insult. If it seems mild, that's only because it pales to BL's early work. I don't think that warrants a month suspension, but Furtado said that he preferred to keep discipline private, so we don't have all the facts.
   78. The Run Fairy Posted: June 20, 2006 at 12:43 AM (#2069016)
Jim sucks. I hate him.
   79. The Run Fairy Posted: June 20, 2006 at 12:45 AM (#2069023)
BL should just reregister under a different name. How could that ##### Jim stop him.
   80. Dr. Vaux Posted: June 20, 2006 at 12:56 AM (#2069043)
He could ban the whole ISP, and then we'd lose half of Massachussetts. That would slow things down for a while...
   81. Jeff K. Posted: June 20, 2006 at 01:15 AM (#2069097)
The way that thread reads, I get the feeling there were some deleted posts.

Not to further the off-topicness of this thread, as I wouldn't like to see it shut down, but no, there are no deleted posts. They're all there.

BL should just reregister under a different name. How could that ##### Jim stop him.

Believe me, if BL wanted to post, he would. I believe he is respecting the ban.
   82. SPICEY WITH A SIDE OF BEER ON A BABYYYYYYY Posted: June 20, 2006 at 01:26 AM (#2069128)
This is all Dan Szymzzlkdfjlasdfkborski's fault for being a dirty Polack and banning Backlasher.

It's in Revelations people.
   83. Joyful Calculus Instructor Posted: June 20, 2006 at 01:50 AM (#2069204)
Have posts ever been deleted? Can posts be deleted?

Why aren't people allowed to delete their own posts? It would clean up double posting for one thing.
   84. jmp Posted: June 20, 2006 at 02:04 AM (#2069245)
Key holders can delete posts, primarily for the removal of the most offensive of posts. Presumably we are not allowed to modify our posts because some users would take advantage of it by taking out something after they wrote it, making it difficult to follow a heated discussion.
   85. Gaelan Posted: June 20, 2006 at 02:12 AM (#2069268)
Let's get back on topic. If we take DIPS at face value then what we are forced to conclude is that Roy Halladay is lucky and that Josh Towers is unlucky. As any fan of the Blue Jays, hell anyone with eyes, will tell you that is not true. It's not within an apple falling from the sky of the truth.

Now I know that recent work using play-by-play data is trying to get at a more refined formula. The problem is that this isn't DIPS anymore. Play-by-play data acknowledges that pitchers have some control over, well, the play.

Even if I'm willing to accept Tango's number that pitchers have 25% control (and I am for the most part) over what happens. What surprises me is how many people talk about that as if it were an insignificant number. And that's without getting into the whole my name is Josh Towers and I throw batting practice bit (the MWE point).

The problem with DIPS, indeed the problem with all abstract statistical methodology, is that it has no way of identifying ahead of time whether any particular player is an outlier. Therefore while it may produce satisfactory results in the aggregate there is no way you can be confident in any particular finding. And since baseball decisions, as well as baseball discussions, are always about particulars it doesn't tell us anything we don't already know.
   86. Steve Treder Posted: June 20, 2006 at 02:25 AM (#2069295)
Even if I'm willing to accept Tango's number that pitchers have 25% control (and I am for the most part) over what happens. What surprises me is how many people talk about that as if it were an insignificant number.

I don't think it's an insignificant number. But I will say that it is a far, far smaller number than I had always -- for decades, in fact -- assumed it to be, without ever once challenging my assumption.

The DIPS findings blew me away. For the longest time I didn't accept them as being remotely true (and we had several of these discussions on this site). From where I was coming from after more than 30 years of intensely watching and studing baseball, the difference between 25% control over BABIP and 0% control over BABIP isn't nearly as significant as the difference between 25% and what I had thought it was.

Maybe you already knew all of this, and so the orginal DIPS presentation and everything that's followed in its wake didn't tell you anything you didn't know. If so, more power to you. It's told me a hell of a lot I didn't already know.
   87. limozeen Posted: June 20, 2006 at 02:30 AM (#2069304)
I just wrote an article on a new way of calculating DIPS. Please check it out at Beyond the Boxscore.

http://www.beyondtheboxscore.com/story/2006/6/19/222621/191
   88. Dan Szymborski Posted: June 20, 2006 at 02:37 AM (#2069316)
I should go bug Voros to pop up into this thread.

From talking with him over the years (he's yet another member of the Usenet Mafia), I get the sense that he's actually quite irritated about how many people have put words into his mouth and then got mad at him for those words.

Voros from the original thread, in January 2000:


What I did say was that pitching seems to be a minor
contributor compared to the other factors. As I mentioned before, hits on
balls in play for individual pitchers correlate VERY poorly from year to
year.

Now groundball pitchers seem to give up more hits than flyballers and
trick delivery pitchers seem to give up less hits than others, etc. It's
just that in general the effect pitchers have appears to be small.


That's a lot different than the suggestion that Voros is claiming that league-average $H is a magic number that all pitchers will gravitate towards.

Tippett found more predictability than Voros did at first (Voros had already taken a second look at additional pitchers by the time Tom did his SABR Presentation), but he also demonstrated that it was very small.
   89. JPWF13 Posted: June 20, 2006 at 11:29 AM (#2069497)
I get the sense that he's actually quite irritated about how many people have put words into his mouth and then got mad at him for those words.

But he did at one time make many if not most of the claims that people keep putting "into his mouth"
The very first article I ever read by Voros did in fact state, up front and center that pitchers had no effect on BABIP- of course towards the end of the article he started quibbling around the edges of that initial statement...

Part of the problem is poor essay construction, Voros (like many writers, myself included) innaccurately summarize what is to follow in their inirial paragraph

teh other problem is that Voros is being blamed/accused of still believing something he quite likely no longer adheres to
   90. greenback wears sandals on his head Posted: June 20, 2006 at 12:01 PM (#2069503)
The problem with DIPS, indeed the problem with all abstract statistical methodology, is that it has no way of identifying ahead of time whether any particular player is an outlier.

I've been doing this Internet discussion thing for a long time now, and I've yet to encounter a person who doesn't use some sort of "abstract statistical methodology." Some RBI and W/L-quoting folks may not recognize that's what they're doing, but in fact that is what they're doing. Black swan caveats aside, given everyone uses some sort of stats, doesn't it make sense to use the best available stats while recognizing the limitations of those stats?
   91. Gaelan Posted: June 20, 2006 at 03:27 PM (#2069628)
Black swan caveats aside, given everyone uses some sort of stats, doesn't it make sense to use the best available stats while recognizing the limitations of those stats?


Sure, but as has been explained by others more clearly than I have managed DIPS and other constructions aren't the best available stats. In many ways they are the worst because not only are they not true the pretension that it is based upon doesn't allow you to recognize it's limitations.

The best that can be done is to say here are this guys DIPS numbers take them for what they are worth. Which is not much.

Now it is true that every tool is only as good as the person that is using it and some people are going to make fewer mistakes than others. Here, at the vanguard of sabremetrics, the erroneous quality of DIPS is at least ackowledged and if some people find uses for it despite that well that's fine with me. However if you go trolling around the internet and read supposedly savvy internet baseball blogs you will see over and over again that pitchers have no control over balls in play. In many ways that ridiculous statement is the least contentious and least contended portion of the sabremetric orthodoxy. That's a pretty big problem if you ask me and it seriously reduces the overall quality of discussion and makes the "field" of sabremetrics look pretty silly. In fact, if I may continue my rant a little longer, it's why, without knowing anything about physics, the physics analogy isn't apt. It wasn't a breakthrough that was subsequently improved upon. It was a mistake from the beginning from which it's adherents keep retreating.

In fact, if anything, the insanity seems to be spreading. More and more I've been reading about how this or that hitters batting average on balls in play isn't sustainable. Now I don't think that anyone here necessarily believes that kind of thing but I believe it's the inevitable result of building statistical models upon the basis of false metaphysical premises. Whether DIPS conforms to the Platonic idea of truth, as someone put it, matters because the effect of knowledge, when disseminated, goes far beyond whether the immediate practical application works. It starts to enter people's consciousness as a building block of reality from which they construct their understanding of the world. This is true for both readers who never do any actual calculations themselves as well as actual practitioners. It's how science becomes a religion and it's why the purpose of true science must always be motivated by the quest for truth and should not be motivated by utility.
   92. greenback wears sandals on his head Posted: June 20, 2006 at 04:34 PM (#2069679)
Sure, but as has been explained by others more clearly than I have managed DIPS and other constructions aren't the best available stats.

What thread are you reading? I haven't seen anything proposed here that's superior to a DIPS-like construction. Even MWE, a frequent critic of DIPS, is in post 36 at worst ambivalent to the use of a piece-wise DIPS function.

However if you go trolling around the internet and read supposedly savvy internet baseball blogs you will see over and over again that pitchers have no control over balls in play.

I don't care much for philosophers whose primary goal is to build a philosophy that can't be misinterpreted by twentysomething know-it-alls. Furthermore you're not going to solve the problem at hand simply by abandoning statistical analysis. The know-it-alls will misinterpret something else. That's what they do.

As it is, the genie is out of the bottle. One of the safeguard against dogmatization is, in direct opposition to your post #2, discussion of the matter.
   93. Buzzards Bay Posted: June 21, 2006 at 12:04 AM (#2070097)
semantics and linear logic don't jive here
   94. Buzzards Bay Posted: June 21, 2006 at 12:26 AM (#2070129)
there is a quantum dynamic between a ball and a strike
what is the dynamic between a letter and a number in this application
   95. Gaelan Posted: June 21, 2006 at 02:01 AM (#2070331)
As it is, the genie is out of the bottle. One of the safeguard against dogmatization is, in direct opposition to your post #2, discussion of the matter.


Well to be perfectly honest I was being deliberately inflammatory in order to provoke a response. Sometimes threads disappear and die here and I didn't want that to happen.

Furthermore you're not going to solve the problem at hand simply by abandoning statistical analysis. The know-it-alls will misinterpret something else. That's what they do.


I'm not opposed to statistical analysis. I'm opposed to bad statistical analysis and I'm really opposed to statistical approximations that portray themselves as accurate.

The interesting thing is that the key point isn't really in dispute. Pitchers do have significant control over balls in play and this ability varies among pitchers.
   96. Rally Posted: June 21, 2006 at 02:20 AM (#2070358)
That states the question backward (I think). Of course DIPS is a better predictor than ERA alone. The question is: does DIPS beat ERA regressed to the mean? And if not, what good is it really?

All right, I've finally dug up some year old research.

My sample was all pitchers with 50 or more IP in 2003 and in 2004.

r - results compared to actual 2004 ERA were:

2003 ERA (nonregressed) = 0.32
Regressed 2003 ERA = 0.32
How much? I tried 50%, 25%, 75% - got the same figure for each
2003 component ERA: 0.33 - a slight improvement
2003 FIP = 0.44 (Tango's formula - I have not the patience to calculate DIPS the way Voros Does it)
Custom regression formula = 0.41 For a pitcher with 150 innings, hits in play are regressed 75%, homers 50%, walks 20%, and strikeouts 10%

That's the most theoretically sound way to do it. But in practice it didn't beat DIPS.
   97. G.W.O. Posted: June 21, 2006 at 02:41 PM (#2070620)
Voros: As I mentioned before, hits on balls in play for individual pitchers correlate VERY poorly from year to year.

Point-by-point cross-correlations of time series data is a very, very, very bad way of doing anything, except determine if your data has a lot of noise in it.
   98. Steve Treder Posted: June 21, 2006 at 03:26 PM (#2070668)
The interesting thing is that the key point isn't really in dispute.

No. The key point is the subject of tremendous dispute; witness this very thread and all of the discussion like it that's resulted from Voros's initial presentation.

Pitchers do have significant control over balls in play and this ability varies among pitchers.

This statement is only "not in dispute" so long as we all agree upon exactly what the definition of "significant" is, as well as all agreeing upon *how much* this ability varies among pitchers. And, of course, *why* what the causes are of whatever variation does occur.

These issues haven't been resolved. They remain open to lots of meaningful further inquiry. Voros's initial inquiries were most certainly not the last word, but they were a tremendously important and stimulating first word, if you will, in generating tons of fascinating study and discussion. Because before Voros came along and did his work, this vein of study and discussion wasn't being explored. We all had our assumptions -- I know I did -- but they weren't being challenged and tested.
   99. GuyM Posted: June 21, 2006 at 11:33 PM (#2071223)
ARM:
Does the amount of regression on the 2003 ERA vary based on each pitcher's individual IP/BF? If not, then I think it would have to be the same r as for 2003 ERA, no matter how much you regress.

I imagine that FIP beats the customized regression because it captures park factors, and because relievers and starters probably need to be regressed to different means (assuming it's not just a fluke due to a single year sample, which it easily could be). If the four rates are park adjusted and regressed properly, with a large sample that would have to be a little more accurate than FIP -- by definition, really. But I take your point, which is that any improvement on FIP (or DIPS) will probably be very small.
   100. Gaelan Posted: June 22, 2006 at 05:01 AM (#2072012)
No. The key point is the subject of tremendous dispute; witness this very thread and all of the discussion like it that's resulted from Voros's initial presentation.


Really? My impression was that the pro-DIPS camp acknowledges that it's not really true but doesn't care because it nonetheless seems to "work".

You must be Registered and Logged In to post comments.

 

 

<< Back to main

News

All News | Prime News

Old-School Newsstand


BBTF Partner

Support BBTF

donate

Thanks to
Martin Hemner
for his generous support.

Bookmarks

You must be logged in to view your Bookmarks.

Hot Topics

NewsblogOTP 20 November 2017: Sheriff’s official suspended 10 days over Cubs World Series sneak-in
(277 - 2:09pm, Nov 21)
Last: Barnaby Jones

NewsblogMore on WAR – Joe Blogs – Medium
(197 - 2:08pm, Nov 21)
Last: jmurph

NewsblogJoe Morgan Wants To Keep Steroid Cheats Out Of Baseball Hall Of Fame | MLB | NESN.com
(23 - 2:05pm, Nov 21)
Last: Steve Balboni's Personal Trainer

NewsblogJim Palmer on Mark Belanger and Omar Vizquel: The Hardball Times
(119 - 2:03pm, Nov 21)
Last: Rally

NewsblogOT - NBA 2017-2018 Tip-off Thread
(1477 - 2:01pm, Nov 21)
Last: madvillain

NewsblogThird time through the order isn’t a charm, and Mets are ready to deal with it | Newsday
(2 - 1:57pm, Nov 21)
Last: Adam Starblind

NewsblogTaking Back the Ballparks - Phillies Voting Thread
(6 - 1:55pm, Nov 21)
Last: Jose is an Absurd Doubles Machine

NewsblogProspectus Feature: Bill James vs. The Noise - Baseball Prospectus
(7 - 1:11pm, Nov 21)
Last: 6 - 4 - 3

Sox TherapyEnd of an Era
(6 - 11:51am, Nov 21)
Last: jacksone (AKA It's OK...)

Gonfalon CubsLooking to next year
(202 - 11:40am, Nov 21)
Last: Pops Freshenmeyer

NewsblogTeams set 40-man rosters ahead of Rule 5 Draft
(7 - 11:38am, Nov 21)
Last: DL from MN

NewsblogGabe Kapler's only season as a manager gave hints for how he will lead Phillies
(31 - 11:10am, Nov 21)
Last: Omineca Greg

NewsblogJudge and Altuve | Articles | Bill James Online
(61 - 11:10am, Nov 21)
Last: Rally

Hall of Merit2018 Hall of Merit Ballot Discussion
(245 - 10:51am, Nov 21)
Last: JoeD has the Imperial March Stuck in His Head

NewsblogFinding the best spot for Giancarlo Stanton
(3 - 10:23am, Nov 21)
Last: Sunday silence

Page rendered in 1.3090 seconds
47 querie(s) executed