Gary: Conversely, if a pitcher has a particularly low batting average on balls in play, they like to tell you it’s going to rise eventually. Well, to me that doesn’t make any sense. Certain guys hit the ball harder than other guys hit it. Certain pitchers induce more groundballs or more weakly hit balls than others. That’s part of what you’re trying to do. Am I totally off base with that?

I actually don't think Cohen is too far off-base here. Hasn't DiPS (at least the strong version of it) been pretty thoroughly debunked?

Meh, at least he didn't go to spreadsheets or mothers' basements, he actually thought about it. That way lies hope.

At least he sounds very rational and logical about unlike 99% of the other critics of sabermetrics.

Believers dont like it when you express skepticism about their religion.

I actually don't think Cohen is too far off-base here.

I agree. That's a pretty reasonable position there. For pitchers, "debunked" might overstate it somewhat, but BABIP is definitely a skill. What bothers me is when "sabermetrics" guys try to take the lesson of DIPS and apply it to hitters, which makes no sense. BABIP is clearly a skill for hitters, which is exactly what Cohen is saying here. Now, if a guy's BABIP is well above his personal "BABIP skill level", then it might be reasonable to expect some regression, absolutely. And this does appear to be the case, at least somewhat, with David Wright. Although he's kind of changed his profile the last couple of years - he's now a higher-BABIP, higher-K guy in 2009-10 than he was before then.

Also, as an aside, it seems to me that BABIP as currently constructed isn't the right thing to be looking at for batters anyway. It seems to me that you ought to be looking at on-contact BA/SLG, i.e., including home runs.

I actually think that's a pretty reasoned discussion of the nuances that are lost if you just look at BABIP as an indicator of non-repeatable "luck".

Hasn't DiPS (at least the strong version of it) been pretty thoroughly debunked?

That's news to me. Where was it debunked? I thought basically the argument was not that it was going to rise, but that BABIP has no predictive value from one year to the next because BABIP was not entirely in the control of the pitcher. I understand that there have been some refinements, but has that basic premise been refuted?

The point about BABIP with hitters, as I understand it, is not to compare them to the league BABIP but to their own BABIP.

Cohen seems right to me (and I hate Cohen). Discount sabermetrics has started using BABIP for hitters as if it were totally based on luck (or based on luck and line drive percentage). Reading through the last to Baseball Prospectus annuals, everyone who overperformed did so because his BABIP was high due to luck.

Ron: No I totally agree with you, I think that for the average hitter, to have a high average putting balls in play, it’s probably because they do have some lucky hits. But certain hitters, like Wright, hit the ball hard almost all the time.

Without getting into details about Wright as an individual (I'm sure Mets fans have fully hashed out the profile change or BABIP fluke argument) isn't this precisely true?

For instance, David Eckstein career BABIP .298 (in 5537 PAs)
Derek Jeter career BABIP .358 (in 10188 PAs)

The point about BABIP with hitters, as I understand it, is not to compare them to the league BABIP but to their own BABIP.

OK, but their "own BABIP" is not necessarily static, of which the case in point, David Wright, is a fine example.

has that basic premise been refuted?

Not at all. The basic premise stands as one of the most dramatic insights ever achieved regarding the intricate interaction of pitching and fielding, a remarkable step forward. The debate is entirely around the details. Nothing has been "debunked."

Where was it debunked? I thought basically the argument was not that it was going to rise, but that BABIP has no predictive value from one year to the next because BABIP was not entirely in the control of the pitcher. I understand that there have been some refinements, but has that basic premise been refuted?

Essentially, original DIPS said that if you were going to regress things either 0% or 100% toward league averages, then you want to regress K-rate and BB-rate 0% and BABIP 100%. But there's no good reason to limit your choices to 0% or 100% regression and the "right" answer for BABIP is something less than 100%. DIPs has also been refined to recognize that "BABIP skill" exists for various subsets of pitchers - knuckleballers; left-handed pitchers, I believe; pitchers with certain "trick" pitches (Mariano's cutter); Mike Emeigh has suggested that closers have below-average BABIP numbers; some research suggests that high-K pitchers tend to have below-average BABIP numbers; etc.

See, for example, this and this.

That's news to me. Where was it debunked? I thought basically the argument was not that it was going to rise, but that BABIP has no predictive value from one year to the next because BABIP was not entirely in the control of the pitcher. I understand that there have been some refinements, but has that basic premise been refuted?

The major discussion turning point, I think, was the "solving dips" thread at the old "Primate Studies" blog where Tango used to work. Here's a pdf culled from the major posts on that thread.

My understanding is that it is the case that single-year BABIP is not a particularly useful statistic, but this is not because pitchers don't differ significantly in BABIP skill - rather, it's a function of variance. A single pitcher season isn't a sufficient sample for understanding a pitcher's expected BABIP, and the signal gets swamped by the variance. Over larger samples, differences come through.

Another place to look this up is the delta-H number at Baseball Prospectus, on their player cards. That number is the difference between a pitcher's hits allowed and the pitcher's expected hits allowed based on his teammates BABIP (the team's DER, that is). For pretty much every great pitcher in history, that number is negative - the pitcher gave up fewer hits on balls in play than his teammates. Randy Johnson is the lone exception, iirc.

Nevermind.

BABIP has no predictive value from one year to the next because BABIP was not entirely in the control of the pitcher. I understand that there have been some refinements, but has that basic premise been refuted?

Some pitchers have more control over BABIP than others. The original ("strong") DIPS argument was that pitchers have absolutely no control over BABIP, that the only thing that pitchers have control over (at least at the MLB level) are peripherals, and everything else is the performance of the defense behind the pitcher and "luck".

But then cracks started appearing in that argument. First came an acknowledgement that knuckleballers are an exception to the rule. Then the acknowledgement that sinkerballers and screwballers can control BABIP at least to some extent. While it still applies to many, even most pitchers, there are enough exceptions to the rule that it cannot be applied across the board. So yes, DIPS has been debunked, or at least heavily modified from its original formulation.

Thanks all for the thoughtful and patient explanations.

While it still applies to many, even most pitchers, there are enough exceptions to the rule that it cannot be applied across the board. So yes, DIPS has been debunked, or at least heavily modified from its original formulation.

I would say that an insight that no one in history had made before, that after intense scrutiny is confirmed to apply to many, even most pitchers, has hardly been debunked. It should be acknowledged as the breakthrough achievement that it is, something that all of us now routinely use when none of us had before. None of us view pitching statistics the same way we did ten years ago. Our comprehension of what we're seeing has been permanently and meaningfully improved.

Not at all. The basic premise stands as one of the most dramatic insights ever achieved regarding the intricate interaction of pitching and fielding, a remarkable step forward. The debate is entirely around the details. Nothing has been "debunked."

I mean, the first DIPS article said, in plain English, that pitchers have no ability to control balls in play. That was the entire thesis of the first DIPS article. The second DIPS article said that was not the case. No one, anymore, thinks that's the case. "Debunked" is pretty fair.

I see DIPS as an idea that was deeply wrong, but wrong in a brilliant and creative way, which led to a huge amount of new, fascinating research. Without DIPS, we probably don't have a lot of the things we think of as contemporary sabermetrics. But it was wrong, and it got debunked very quickly.

I would say that an insight that no one in history had made before, that after intense scrutiny is confirmed to apply to many, even most pitchers, has hardly been debunked.

It doesn't apply to any pitchers. If handedness and k-rate have correlations with BABIP - as even DIPS 2.0 argues - then all pitchers differ and strong DIPS applies to no one. I would further suggest that individual pitchers clearly have different expected BABIPs simply based on their being individual pitchers, beyond handedness and k-rate. That's the conclusion drawn in "Solving DIPS".

The contemporary sabermetric consensus that we need to take regression seriously and account carefully for how hits happen in between pitching and defense wouldn't have come about without DIPS, or at least in the actually-existing history of sabermetric, Voros' research was central and formed a major locus of discussion.

single-year BABIP is not a particularly useful statistic, but this is not because pitchers don't differ significantly in BABIP skill - rather, it's a function of variance. A single pitcher season isn't a sufficient sample for understanding a pitcher's expected BABIP...

If this were true, then the phrase "debunked" would in fact be accurate, but I don't think this is true.

No one, anymore, thinks that's the case.

Unfortunately, I'm not so sure that's true. FIP is an extremely popular pitching metric that very explicitly excludes BABIP entirely. I do agree with the spirit of Steve's #17. Voros made an extremely significant contribution to the understanding of pitching skill at the major-league level.

Voros made an extremely significant contribution to the understanding of pitching skill at the major-league level.

He certainly did, and to breezily dismiss its enormous importance, to casually characterize it as having been "debunked," is simply not right.

The basic premise stands as one of the most dramatic insights ever achieved regarding the intricate interaction of pitching and fielding, a remarkable step forward.

I really never understood all the excitement. Anybody who ever watched a pitcher get blooped and seeing-eye-groundered to death more than once must have intuitively understood the basic premise long before Voros ever over-stated it. And James figured out in one of the early abstracts that high strikeouts and low walks are the best predictors of future success for pitchers.

Also, is "debunked" really fair if the guy who wrote the first article is also the guy who corrected it in the second?

The only problem I have with Cohen here is that he's putting words in the mouths of sabrmetric proponents. Nobody is out there claiming that BABiP isn't a skill for hitters.

And this does appear to be the case, at least somewhat, with David Wright.

Hard to know exactly what's going on with David (in a lot of respects I guess). Last year when he was "off," his BABIP was .394 compared to a career average just shy of .350. People (reasonably) thought that disparity portended bad things to come for his overall stats. But this year the BABIP is .410, which is more or less historic.

#10 - Yes, quite accurate. The problem is that it was presented as a rebuttal to the "sabermetric" view, when in fact it is very clearly supported by the research to date on BABIP, batted ball types, and so forth. Some players hit the ball harder than others. Yes, that's what we've been saying.

We essentially agree on this point, but our broadcasters are so compelled to dismiss objective analysis that they'll distort -- or just misunderstand -- the findings in order to make their case against it. That is precisely what I'm frustrated and disappointed by.

I mean, the first DIPS article said, in plain English, that pitchers have no ability to control balls in play. That was the entire thesis of the first DIPS article. The second DIPS article said that was not the case. No one, anymore, thinks that's the case. "Debunked" is pretty fair.

Not to get into an argument over semantics, but the definition of "debunked" is "the exposure of falseness or pretensions". While the degree to which the original DIPS theory was postulated has been altered, the theory itself has not been "exposed as false". The theory prior to DIPS said that pitchers had a lot (or complete) control on balls play; the original DIPS theory said that pitchers have no control. So it was a 0% or 100% binary. Since then, we have moved closer to an answer that lies between the two, closer to the original theory of 0%. I believe Matt Swartz and Eric Seidman found that pitchers control something like 15% of BABIP and almost all of that is dictated by DIPS skill (K/BB).

Not to get into an argument over semantics, but the definition of "debunked" is "the exposure of falseness or pretensions". While the degree to which the original DIPS theory was postulated has been altered, the theory itself has not been "exposed as false".

Exactly.

The fundamental insight was original, powerful, and lasting. It deserves our respect, not our minimizing.

Here's the Swartz/Seidman article:

http://www.baseballprospectus.com/article.php?articleid=10281

In any given season, the average starting pitcher who can keep his job will have his BABIP determined roughly 75 percent by luck, 13 percent by his team's defense/park, and 12 percent by his own skill.* How do we actually figure out the 12 percent that is skill, when we know that the variance in BABIPs and the limits of sample size imply that such a large fraction is luck? Fortunately, J.C. Bradbury found in 2005 that much of BABIP skill from pitchers can actually be explained by their defense-independent skills. In fact, about 86 percent of the pitcher portion of BABIP skill is explained by these statistics.

It is obviously not the case that higher strikeout rate causes lower BABIP. That's not even wrong, it's incoherent. How could strikeout rate or walk rate cause BABIP rate, to any meaningful degree?

It appears to be the case, rather, that the underlying skills that produce good three-true-outcomes numbers for pitchers also produce good BABIP numbers for pitchers. This is interesting, and it makes some amount of sense, because good pitchers are good, and missing bats and preventing strong contact are fundamental to Ks, HR prevention, and hit prevention. K/BB/HR skill does not "explain" or "dictate" BABIP skill. Underlying ability to miss bats, rather, leads (to different degrees, and differing among pitchers somewhat) to strikeouts, to preventing homers, and to preventing hits.

Seidman/Swartz are quite clear on this point, and argue it well. They are not claiming that BABIP skill is somehow secondary to K-rate - they're claiming that these are both related to underlying skills, in ways that are useful for producing their uber-stat.

They still use Bradbury's misleading "explained by" phrase - really, not misleading, simply wrong - but they then explain they don't mean "explained by", they mean something else entirely.

"I think even casual fans can grasp the basics of things like BABIP (when explained properly) and FIP ..."

I doubt that. Keep in mind that fewer than 4 in 10 Americans believe in, accept or understand biological evolution. About a quarter of Americans are Biblical literalists who believe that the Earth and all its flora and fauna were created over 6 days 5,770 years ago; and every animal, plant and man himself has never evolved. The rest are still trying to figure out email.

"Nobody ever went broke underestimating the intelligence of the American public." --H. L. Mencken

Simon is overly mean-spirited and overzealous in his post. But he does point out that BABIP != luck. It's the main part of BP's annual that bothers me (that and the cherry-picking of successes on the back cover--any publication that analyzes hundreds of items can report a similar success rate). BP *knows* that BABIP isn't merely due to luck--they talk about line drive rates and defense--but then, in the player comments, they often say that so-and-so is due for regression or a bounce-back season due to bad or good luck on balls in play. Maybe it's shorthand, but to me it's lazy and dishonest.

The fundamental insight was original, powerful, and lasting.

No it wasn't. The idea that pitchers can only do so much and that sometimes bloopers find holes and line drives find fielders is well known by anyone who has ever played. Voros simply stated (and indeed overstated) this truth in a very dramatic way.

No it wasn't. The idea that pitchers can only do so much and that sometimes bloopers find holes and line drives find fielders is well known by anyone who has ever played.

Baloney. DIPs isn't and never was merely a recognition of the obvious. The notion that everyone already understood the insight that DIPs provided is sour grapes revisionist history.

"All models are wrong, but some are useful." is a great old quote and is spot on here.

DIPS proved "wrong" as a statement of fact, but the model has been incredibly useful in understanding baseball. That's all you can ask out of model.

This is a nonsense article with a nonsense title. It would be more accurate to say: Gary Cohen Doesn't Understand Nuances of Certain Advanced Statistics, Distrusts Them.


Also, I completely agree with Treder in #33. DIPS was totally noninuitive, totally anti-tradition, it was basically as close to mind-blowing as any sabermetric advance can aspire to be. Yes, the case was overstated, but only barely.

The key word is 'believe' - that's the crux of any dualistic conflict between sheep and goats. Agnosticism's preferable to belief.

Simon is overly mean-spirited and overzealous in his post.

Yea, eff that guy.

Voros simply stated (and indeed overstated) this truth in a very dramatic way

I don't think that's true at all. Before Voros's work, if you'd have asked me or probably anybody which pitchers had the best BABIP last year or lifetime, we'd have said "the best pitchers." Voros drew attention to the fact that, particularly over small samples of time like a single season, that's not nearly the case. I don't understand a heck of a lot about the math, but it's a fascinating insight that I don't think anybody had had before in any kind of systematic way.

OMG, it's THE Eric Simon!!!

#32 -- If that were all that occurred, then no one would have given Voros so much credit when he first popularized his idea. Yet Bill James and many other respected baseball analysts immediately responded by crediting Voros with a great insight that up to that point had been missing. As many have said in this thread, the model has evolved with the benefit of hundreds of other smart sabermetricians looking at it from various angles and seeing where it could be improved. But that it is now different in certain respects than what Voros originally wrote does not change the fact that his "discovery" was widely credited by credible people as a great breakthrough. Almost everything in hindsight can be thought of as "a recognition of the obvious." But it was not obvious at that time.

#40 - The work that has gone on since Voros' overstated thesis is useful in quantifying something that we all knew was out there. People got real excited when they thought that he had proved something truly revolutionary, that pitchers had no control. Of course he had proved no such thing.

@41.

So, before Voros's work, people thought that pitchers were responsible for what pct of what happened on the field?

Voros was a hell of a lot closer than anyone else had it before. You not recognizing this is a problem with you, not with any theory produced.

[9]

Cohen seems right to me (and I hate Cohen).

I don't understand why someone would hate Gary Cohen.

Would you rather listen to such all-star announcers as Fran Healey, Tim McCarver (now, not mid 1980s version), Dave O'Brien, or Wayne Hagin?

Now, Howie Rose, on the other hand... :)

That's right. That's the way it is the more a discipline approximates science. Bill James himself has postulated a number of explanatory theoritical sets that he's since abandoned ecause they were discredited, and some other ideas of his he has revised substantially because of the input of others. I can't wait to see his corrections on WinShares--I have enough faith in his integrity to believe he is capable of being informed, and being informed often results in undergoing a change. He's not the only one capable of that. Actually, I find all that encouraging--it means the culture is open and receptive. As long as it stays that way, we'll learn things and then have a chance to come to state of the art explanations for those things.

"The idea that pitchers can only do so much and that sometimes bloopers find holes and line drives find fielders is well known by anyone who has ever played"

Or anyone who has ever watched. Mom's basement, pajamas, cheesy poofs, ding dongs, yadda yadda yadda

So, before Voros's work, people thought that pitchers were responsible for what pct of what happened on the field?

Some. Nobody had figured it out but it wasn't 100%.

Voros was a hell of a lot closer than anyone else had it before. You not recognizing this is a problem with you, not with any theory produced.

Good for him. Doesn't change that he got famous (well kinda) off a study that was statistically flawed and ultimately incorrect (at least in it's initial revision).

Everything that can be said on topic has been well expressed so I'll get this petty gripe off my chest. Why does Cohen say "...right FIELD line" and "...left FIELD line"? Is it some homage to rednecks?

Nobody had figured it

Yeah, that's kind of the gigantic point.

Those grapes sure are sour.

I don't understand why someone would hate Gary Cohen.

He's a very good play by play guy, but he moralizes, and he has the moral values of your average sportswriter, so he spends time complaining about flash or steroids or players being too latin.

Mike Emeigh has suggested that closers have below-average BABIP numbers

A cause could be the presence of defensive substitutes behind the closer. (I'm not familiar with Mike's work, maybe he hit on that.)

As for "DIPS debunked", we're arguing over semantics. Of course the literal conclusion of "pitchers have 0% control of BABIP" is not correct. But the messages of "pitchers don't have as much control as conventional wisdom supposes" and "one pitcher-season is a small sample" are undoubtedly true.

Would one say that Einstein debunked Newtonian physics by showing that there is no absolute universal time, ignoring that Newton correctly introduced the whole concept of universal gravitation?

Good for him. Doesn't change that he got famous (well kinda) off a study that was statistically flawed and ultimately incorrect (at least in it's initial revision).

Oh my stars. Clearly, Voros is history's greatest monster.

This is how things go.

I'm not familiar with many other fields, but this sounds exactly like how the process works in history.

1) Historian makes seemingly revolutionary claim
2) More historians read his work, assertions are challenged, conclusions qualified (with the original historian included in this proccess)
3) some sort of rough consensus is reached between what we thought we knew before, and what the revolutionary claims suggested.

Rinse, repeat.

Just because the original claims were challenged and the theory adapted doesn't mean that the historian didn't begin (and participate in) a process that furthered our knowledge in the given area. I think he/she ought to be given credit for that.

Just because the original claims were challenged and the theory adapted doesn't mean that the historian didn't begin (and participate in) a process that furthered our knowledge in the given area. I think he/she ought to be given credit for that.

Not just that - but in this particular case, the guy who did most of the challenging and refinement was Voros himself.

Not just that - but in this particular case, the guy who did most of the challenging and refinement was Voros himself.

Absolutely. He could hardly deserve any more credit and respect.

IMO, the big thing - and it is a big thing - that Voros did was to ask the question "Do pitchers have control over BABIP?" There are lots of sabermetricians who can do all the math that Voros did; it was his ability to see that this question had never been addressed formally that was important. This, BTW, is true of sabermetrics generally, if for no other reason than that sabermetrics is a branch of applied mathematics. In applied math, as opposed to theoretical math, it's almost as important to see what question to ask as it is to be able to answer it. They emphasized that in college for me by having open book tests. They didn't care whether I had memorized formulas. They wanted to know if I could figure out the (or a) correct approach to take in addressing whatever word problem was at hand. Voros didn't just do that; he figured out an important word problem that had never been asked before. That's big, no matter how the math comes out. That is, it's important even IF his conclusion is overturned. If that happens, then someone else had a better answer to the question. But that answer would never have come out if Voros had not asked the question in the first place. That credit cannot ever go away. - Brock Hanke

This will now be the third time I will post this set of usenet quotes on here (the BP article was in 2001):

"'However, I don't think you're wrong here. If the league average totals are adjusted to represent league average totals for GB, FB, LH, RH, SP and RP we might gain some accuracy in the DIP numbers. Also to night is that, anecdotally, I believe pitchers with trick deliveries (e.g. Knuckleballers) might post consistently lower $H numbers than other pitchers. I looked at Tim Wakefield's career and that seems to bear out slightly... I believe that GB/FB adjustments to league average figures would make the system more accurate.'

11/9/99

'The key from here is to refine the DIP measures (assuming now that they are a valuable tool) to be more accurate. We can assign different "league average" values for various pitchers depending on whether they are left handed or right handed, starter or reliever, groundball pitcher or flyball pitcher and other designations which could cause subtle shifts in the overall evaluations.'

12/7/99

'There may, however, be instances where a hits allowed total may be representative of a special ability of a certain pitcher, and the best candidates there would be trick delivery pitchers like knuckleballers and sidearmers. Still more work needs to be done there.'

2/29/00"

The whole point of the exercise was to re-evaluate how we were currently taking into account the performance of the fielders behind a pitcher when looking at a pitcher's statistics. Back in 1999, the way this was done by most everyone was that it wasn't done at all and everyone hoped it all evened out in the end.

The idea was to separate out the statitsics the defense couldn't affect to the ones they could, and then we'd _know_ that at least for the defense independent portion of a pitcher's line the defense was completely eliminated from the discussion. What became interesting was just how much pitching ability was down to the defense independent side of things (almost all of it for the vast majority of pitchers). Furthermore, the idea was to further refine things in the future to take into account subtle differences, but while trying to do so while remaining within the realm of defense independent statistics. So regressing BABIP would have been pointless for this purpose (it would be fine for a projection, and that's precisely what I did for my projections).


Remember in 1999, there was no fangraphs or PitchFX, ball in play data was limited and only very recent, and even something like Retrosheet was just then getting off the ground. It was very hard to get enough reliable data to conclude anything with regards to fly balls and ground balls (and line drives were either not available or completely unreliable). Furthermore, because the effects all around were so small, there just weren't enough years in the books to conclude much anyway. In the very first couple of studies, I even had to estimate BFP, because I didn't have it in my available database yet (the earliest version of the Lahman database didn't have this).

Indeed that's exactly where something like xFIP has gone, and further modifications have also gone. I also did the best I could with the data I could and produced a moderate upgrade in 2001, where I had actually confirmed my early speculations on knuckleballers, mentioned that there was a relationship between strikeouts and lower hits per balls in play, and again brought up the issues of how ground balls and fly balls affect things, and how in many cases the advantages and disadvantages of each cancel each other out (now in 2010, it looks clear that ground balls are superior because of the better way we now understand home runs, but this was harder to completely document back in 1999-2001).

One other note, anyone who says my conclusions were based on a "flawed study" is making a "flawed statement." Why? Because the case of the word "study" is greatly flawed. You haven't a clue how many different studies and different kinds of studies I did, all pointing in the exact same direction (IE, statistically significant, not particularly meaningful for almost all pitchers). There was a matched pairs study, there was a review of career statistics, there was the study I did involving non-pitchers attempts to pitch, there were of course about 30 or so of those year 1 compared to year 2 studies with a bunch of different parameters (every pitcher from 1993 on, every pitcher from 1991 on, only pitchers who changed teams, added dummy variables for knuckleballers and left handers, etc.). So the next time somebody tells you that the only study I did was one study involving 200 some pitchers from 1993-1999, I sure would hope you could correct them for me. They won't listen to me.

[Cohen's] a very good play by play guy, but he moralizes, and he has the moral values of your average sportswriter, so he spends time complaining about flash or steroids or players being too latin.

I have seen endless complaints about this on Gamechatter but I have never really noticed this tendency, or at least it has never bothered me. He is a superb play-by-play man. His relationship with Hernandez and Darling is wondrful. He knows the Mets better than just about anyone living, and yet never sounds like a homer. He is intelligent. He doesn't have catch phrases.

So he complained about Lastings Milledge a few too many times. He's one of the best in the business - enjoy him.

Many folks here are getting caught up in semantics, debating whether Voros's original thesis has been "debunked." Similarly, there have been a few here who maintain that because later research whittled away at Voros's discovery, that Voros "proved no such thing."

To me, this seems to miss the point and cast Voros in a far more negative light than he deserves. Before DIPS, it was presumed that pitchers had a significant role in controlling BABIP. Voros questioned this, but the fact later research has unearthed a small connection for certain types of pitchers shouldn't discredit the notion that for most pitchers, the relationship is minimal at best.

For that matter, it should not lessen the impact Voros's supposition had on baseball and the sabermetric community.

So, is the glass half-empty or half-full? That's essentially the argument going on here.

He's one of the best in the business - enjoy him.

Totally agree. To add to your points, Cohen pays attention to what's going on around the league (many others don't, or they think they do), and his knowledge of Mets' history is as if he's a walking media guide.

When it comes to other TV PBP guys- some are unabashed homers, some are stuck in the "batting average reigns supreme" era, some extol the scrappy scrubs. I heard Dick Enberg (who is still a fine play-by-play guy) recently call David Eckstein the Padres' MVP.

Cohen does get a little annoying with his moralizing, but IMO, the good far outweighs the bad. Gary/Keith/Ron is the best in the biz. For example, you won't hear them raving about how valuable Alex Cora and Jeff Francoeur are to the team, like Kevin Millar did on national TV over the weekend. That was embarrassing.

For that matter, it should not lessen the impact Voros's supposition had on baseball and the sabermetric community.

FWIW, this was never my intention. I was speaking specifically to what Cohen was saying -

if a pitcher has a particularly low batting average on balls in play, they like to tell you it’s going to rise eventually. Well, to me that doesn’t make any sense. Certain guys hit the ball harder than other guys hit it. Certain pitchers induce more groundballs or more weakly hit balls than others.

Cohen may very well be attacking a straw man, but I don't think Voros would disagree with Cohen here.

So, is the glass half-empty or half-full? That's essentially the argument going on here.

15%, but we knew that already.

#35, James' discussion of DIPS is interesting. Basically he says it's obvious once you think of it but not any the less valuable because of this. Then points out the difference in hits allowed on balls in play between Tommy John and Bob Gibson. In other words a very valuable general rule.

And high K pitchers tend to do somewhat better on straight BABIP because they tend to be flyball pitchers and flyballs in play tend to be converted into outs at a slightly higher rate than groundballs (they also seem to be somewhat better at inducing foul balls -- though this skill is generally swamped by the park which makes the analysis more complex).

It tends to come out pretty much in the wash because the flyballs that do drop in tend to be of somewhat higher value (again park factors complicate the analysis)

There was a matched pairs study, there was a review of career statistics, there was the study I did involving non-pitchers attempts to pitch, there were of course about 30 or so of those year 1 compared to year 2 studies...

Voros, did you ever put out the career statistics study in any form? You've mentioned it here a few times. Can you provide a link or citation? As you well know, several studies that looked at career stats tend to support a somewhat "weaker" version of DIPS. Did your study reach different conclusions?

*

Voros' stats for non-pitchers -- position players pitching in blowouts -- are interesting. If you compare them to the average MLB pitcher, you get a picture of what distinguishes a major league baseball player from a guy who is just a hard-throwing athlete. This is how non-pitchers compare to real pitchers:
Non-Pitchers
K/9: 4.0 (vs. 5.5)
BB/9: 5.9 (vs. 3.3)
HR/9: 1.26 (vs. .87)
BABIP: .298 (vs. .278)

Turn this into runs, and non-pitchers are roughly:
K/9: -0.44
BB/9: -0.76
HR/9: -0.56
BABIP: -0.57
It's not clear from this data that one of these 4 items should not be considered a relevant skill.

Good for him. Doesn't change that he got famous (well kinda) off a study that was statistically flawed and ultimately incorrect (at least in it's initial revision).

This actually made me laugh. And not with you.


Cohen IS a great announcer, one of the best, and I'm thankful we have him. The problem is however that he's so good with the game, I think it makes his insufferability when he DOES get on his Kilamanjaro-like soapbox particularly jarring and grotesque and is why people (and I am one of these people) complain about it.

Guy, what time period are you looking at in #64?

So, is the glass half-empty or half-full? That's essentially the argument going on here.

The glass would be optimally exciting at somewhere between 33 and 50% full.

Voros, thank you for reminding us all of the work you did. The anti-DIPS backlash around here is incredible.

The anti-DIPS backlash around here is incredible.

Yes. It's quite a combination of ignorance and pettiness.

AROM, it's Voros' data. 1946-2006.

The anti-DIPS backlash around here is incredible.

Because...there is always a fly in the ointment. (ducks flying object 47 times...at least!)

So he complained about Lastings Milledge a few too many times. He's one of the best in the business - enjoy him.

Personally, I find Lastings Milledge a lot more enjoyable than Gary Cohen.

So no, I won't.

Voros, thank you for reminding us all of the work you did. The anti-DIPS backlash around here is incredible.

agree, I pick on fangraphs and their war for pitchers because it's based upon dips instead of actual results, but it's not my intent to ever knock dips, which I feel is one of the most significant "discovery" in baseball research the past 20+ years. Just because some people use dips incorrectly or don't even try to understand what it says, doesn't mean they are right to completly knock it down.

The glass would be optimally exciting at somewhere between 33 and 50% full.

Take it to the soccer thread.

The anti-DIPS backlash around here is incredible.

Is this the equivalent of a Bat-signal for Backlasher?

I don't care what Cohen's views of sabermetrics are. Hell, I generally don't care what any announcers views of anything are. Cohen was very good as the #2 guy to Murphy - they were a great team. But he has become increasingly obnoxious since he became the #1 guy.

Voros' stats for non-pitchers -- position players pitching in blowouts -- are interesting. If you compare them to the average MLB pitcher, you get a picture of what distinguishes a major league baseball player from a guy who is just a hard-throwing athlete. This is how non-pitchers compare to real pitchers:
Non-Pitchers
K/9: 4.0 (vs. 5.5)
BB/9: 5.9 (vs. 3.3)
HR/9: 1.26 (vs. .87)
BABIP: .298 (vs. .278)

Turn this into runs, and non-pitchers are roughly:
K/9: -0.44
BB/9: -0.76
HR/9: -0.56
BABIP: -0.57
It's not clear from this data that one of these 4 items should not be considered a relevant skill.

Maybe I'm wrong, but I don't think anyone is trying to say it's not a skill. It's just that there is no significant difference between the 4th skill for most MLB pitchers versus the 1st, 2nd, and 3rd skill. The difference between a replacement-level MLB starter and a HOFer in general is going to be based on the first 3 skills. In the minors and at the amateur levels, there is a lot more variation in the BABIP skill.

One should hope that MLB non-pitchers have a higher BABIP than MLB pitchers. They don't have the skill to be MLB pitchers -- they generally last pitched in high school. Therefore they fall under the amateur pitcher category above -- they're going to give up more hits when contact is made. In fact, this difference is likely even worse than these numbers show -- when non-pitchers are in the game due to a blow-out situtuation, often the opponent's best hitters have been removed from the game for rest.

Maybe I'm wrong, but I don't think anyone is trying to say it's not a skill.

I was quoting Voros, who says that any differences in BABIP skill are generally not "relevant." (But to be clear, I'm not supporting or joining in on the bashing of DIPS/Voros in this thread. I agree with Treder's responses to that.)

It's just that there is no significant difference between the 4th skill for most MLB pitchers versus the 1st, 2nd, and 3rd skill. The difference between a replacement-level MLB starter and a HOFer in general is going to be based on the first 3 skills.

This is widely believed, but not really true. HOF pitchers are often above average in preventing hits on balls in play, and replacement pitchers are most definitely below average (as a group). There is lots of evidence of skill differentiation at all levels of pitching: AAA pitchers who eventually make the major leagues, for example, have better BABIP while in AAA than other minor league pitchers.

As an example, here is how Tom Seaver compares to average pitchers over his career, in terms of RAA:
BB +.21
K +.30
HR -.02
BABIP +.29
TOTAL: +.78
So Seaver's BABIP skill accounted for 37% of his advantage over an average pitcher, and was about equal in importance to his 3640 Ks.

I agree it's obvious that non-pitchers will have a higher BABIP. But it's not obvious that the gap between them and real pitchers is just as big on BABIP as it is on Ks or HRs. I don't think DIPS theory would predict that. Also, how plausible is it that major league pitchers differ from non-pitchers by almost identical amounts on all four dimensions, and minor leaguer pitchers also vary in this ability, but yet all MLB pitchers are at the exact same plateau? What other talent in the universe is distributed like that?

One should hope that MLB non-pitchers have a higher BABIP than MLB pitchers. They don't have the skill to be MLB pitchers -- they generally last pitched in high school. Therefore they fall under the amateur pitcher category above -- they're going to give up more hits when contact is made.

I seem to recall someone (maybe Vance Law Revue) saying that you or I would have the same BABIP as Randy Johnson if we faced major league hitters. I can't find the discussion. The search feature here is not helpful.

Anyways, I used to blame Voros for this extremism. I'm starting to get wiser. Gaelan may look down on me for being an apostate, but he doesn't know where I live.

Have to second Cardsfanboy on using FIP as the basis for WAR. According to Fangraphs, Liriano has had the best pitching season in the majors this year. But that's only because his .352 BABIP gets ignored as bad luck, while his .17 HR/G -- which involves way more luck -- gets treated as "real." The result IMO is neither a valid projection nor a valid measure of "real" performance.

Anyways, I used to blame Voros for this extremism.

I've always blamed Red Sox fans who don't take the blame for the poor behavior of all other Red Sox fans.

So, you're clear.

:-)

I've always blamed Red Sox fans who don't take the blame for the poor behavior of all other Red Sox fans.

So, you're clear.

:-)

Hey, I'm more diplomatic because I live in Connecticut. We're like Poland between Germany and Russia. So I get a little upset when I get lumped in with those guys comfortably parked in the rear echelon of Red Sox Nation.

Some pitchers have more control over BABIP than others. The original ("strong") DIPS argument was that pitchers have absolutely no control over BABIP, that the only thing that pitchers have control over (at least at the MLB level) are peripherals, and everything else is the performance of the defense behind the pitcher and "luck".

But then cracks started appearing in that argument. First came an acknowledgement that knuckleballers are an exception to the rule. Then the acknowledgement that sinkerballers and screwballers can control BABIP at least to some extent. While it still applies to many, even most pitchers, there are enough exceptions to the rule that it cannot be applied across the board. So yes, DIPS has been debunked, or at least heavily modified from its original formulation.

The theory prior to DIPS said that pitchers had a lot (or complete) control on balls play; the original DIPS theory said that pitchers have no control. So it was a 0% or 100% binary. Since then, we have moved closer to an answer that lies between the two, closer to the original theory of 0%. I believe Matt Swartz and Eric Seidman found that pitchers control something like 15% of BABIP and almost all of that is dictated by DIPS skill (K/BB).

So how would the DIPS percentages vary if we separated pitchers into five broad categories:

knuckleballers

sinkerballers

scroogiemeisters

all others (high k/bb ratio)

all others (low k/bb ratio)

I have to assume that a category breakdown like this has already been done. But where is it available?

you or I would have the same BABIP as Randy Johnson if we faced major league hitters

Mine would be a great deal lower, since every pitch I threw would either be hit over the fence or barely caught on the warning track.

- Not just that - but in this particular case, the guy who did most of the challenging and refinement was Voros himself.

Absolutely. He could hardly deserve any more credit and respect.

No offense to Voros when he reads this, but really? Should we give him the Nobel Prize? Or would that be a sign of disrespect nowadays?

And, really, Voros has no control over what his theories do when they're in play.

No offense to Voros when he reads this, but really?

Yes. Really.

I just encounter statements like "He could hardly deserve any more credit and respect" when they're said of people of ginormous accomplishment, humanitarian graciousness, and heroic purpose, like John Wooden, Mahatma Gandhi, and Scott Spiezio. Voros had some good ideas before he went too far, and his own generals tried to kill him, and I have no reason to disrespect him, but that praise struck me as hyperbole.

Again, no offense to Voros. DIPS changed the way we look at things, and united the perception that "the breaks [e.g., line drive outs and bloop hits] even out" with statistical reality, which is no small thing, either.

The anti-DIPS backlash around here is incredible.

It could be worse. Just take a look at some of the things Bill Staffa (of SkeeterSoft) has said about Voros (and Sabermetrics in general, FWIW).

Whatever.

you or I would have the same BABIP as Randy Johnson if we faced major league hitters

Mine would be a great deal lower, since every pitch I threw would either be hit over the fence or barely caught on the warning track.

Hey, it's probably hard to hit pitches that bounce 15 feet in front of the plate.

Steve, in case the inclusion of Steve Spiezio didn't tip it off, I'm pulling your chain a bit. I mean, I do agree that Voros' work was very important, but I do think there is a bit of hyperbole in your praise of him.

Unrelatedly, the actual discussion that was happening here before the diversion into what is and isn't "debunked" was good.

Maybe I'm wrong, but I don't think anyone is trying to say it's not a skill. It's just that there is no significant difference between the 4th skill for most MLB pitchers versus the 1st, 2nd, and 3rd skill.

If you select for a skill, as Bill James noted in the Win Shares book, and you are reasonably good at it, which most major league skill evaluators are, you will see less variance in that skill among the people you select than you will for other skills for which you don't select. If you don't account for that selection effect, you wind up exactly where "strong DIPS" winds up, with the argument that the skill matters little when in truth having the skill is essential to success.

-- MWE

the argument that the skill matters little when in truth having the skill is essential to success

Interesting. One sees this in commonplaces like "any major-league hitter can hit a 90-something-MPH fastball." It's quite true that if a big-league hitter knows that fastball is coming in over the plate without much movement on it, he can punish it. But a fraction of a second slower with the bat, and that hitter is back in the minors – or never leaves them in the first place.

So how would the DIPS percentages vary if we separated pitchers into five broad categories:

knuckleballers

sinkerballers

scroogiemeisters

all others (high k/bb ratio)

all others (low k/bb ratio)

Maybe once all the theological arguments have petered out I might be able to get an answer to this empirical question.

god this thread reminds me that people are petty.. I was making fun of my girlfriend because her family was petty in issues that concentrate on gauging intent instead of actual actions. but god, please people grow up.

Let us give some credit where it is due. I was first exposed to "thinking fan" baseball with James' 1991 Baseball Book, and the most memorable part about pitching theory/projection was James' constant focus on pitcher strikeouts and walks above all else. James observed that a 35-year-old striking out 7 men per 9 would last longer than a 25-year-old striking out 5 per 9. He stated it was almost impossible for a pitcher to find success with a strikeout rate in the low 4s or below. I tried to read whatever Saber literature I could over the following years and read the usenet reasonably regularly, but I never thought of or learned why James's observations were true, in spite of them being so basic. James didn't either. Voros came up with the answer, even if it required some later refinement, and that was a real accomplishment. Let's not pretend it was obvious or unimportant.

You know how you go to a museum sometimes, and some guy looks at a Picasso or Mondrian or whatever and says, "I could do that"? Well, why didn't you big guy? You could have millions of dollars and have a harem of starlets, but instead you're at the museum because the pest guy is fumigating your mother's basement to kill the roaches that live in your old pizza boxes on the floor. (I have no idea if Voros is living the high life with starlets, but he did get to work for the Red Sox.)

There is lots of evidence of skill differentiation at all levels of pitching: AAA pitchers who eventually make the major leagues, for example, have better BABIP while in AAA than other minor league pitchers.

Isn't there likely a selection bias at work here? The chances of a pitcher in AAA getting promoted are almost certainly influenced by the number of runs he allows, which in turn is influenced by his BABIP.

In other words, if you have two pitchers with identical FIP numbers in AAA, but one allows a BABIP of .250 compared to the other's .350, the guy with a .250 BABIP will allow fewer runs, and when the major league team has someone get hurt, the guy with the lower ERA is more likely to be promoted to take the spot in the rotation whether his BABIP is based on skill or not.

If that makes any sense.

If you select for a skill, as Bill James noted in the Win Shares book, and you are reasonably good at it, which most major league skill evaluators are, you will see less variance in that skill among the people you select than you will for other skills for which you don't select. If you don't account for that selection effect, you wind up exactly where "strong DIPS" winds up, with the argument that the skill matters little when in truth having the skill is essential to success.

This makes no sense to me. Can someone explain?

You know how you go to a museum sometimes, and some guy looks at a Picasso or Mondrian or whatever and says, "I could do that"?

I say that every time I listen to Gary Lewis and the Playboys, or read the statistics of Marc Sullivan.

This makes no sense to me. Can someone explain?

The example that Bill James used (if I remember correctly) was weight of offensive linemen in the NFL. If you do a correlation analysis of weight versus skill for NFL linemen, you'll get, at best, a very weak correlation and, in fact, you might actually see a negative correlation, because the only way you make the NFL as an under-sized offensive lineman is if you're exceptional at other skills. But there's absolutely a minimum weight below which you simply can't play the offensive line in the NFL.

The theory is that BABIP is the same thing. 75% of all plays result in balls-in-play. If you can't control BABIP, you simply can't survive as a pitcher in the major leagues. You'll get singled and doubled to death. So, that end of the BABIP-skill curve - the pitchers whose "true talent" is above a certain threshold (say, .330 or .350 or whatever) - get drummed out of MLB so quickly (if they make it at all) that they never qualify for any of the kinds of studies that Voros, et al. have done to try to identify BABIP variation. So, what you're left with is an artificially tightened cluster of BABIP-skills among MLB pitchers.

Piehole, I think Bill James used a football analogy (it may have been concerning something else though). If you look at all offensive linemen and compare their weight to their talent, whether measured by Pro Bowls, All-Pro, years starting....you might determine that weight has nothing to do with ability to be an offensive lineman. But everyone playing O-Line in the NFL has already been selected because of their weight. If you're under 280 lbs (maybe even 300 these days?), you don't even get the opportunity to play line in the pros. Among players big enough to play line, their weight has little correlation to their performance. But they have to be big enough to get that chance in the first place. Size is essential to playing O-Line in the NFL, although the average 350-lb lineman may be no better than the average 300-lb lineman.