BTW, the bases loaded # was the only number I used where the sample size was lower than 80 ABs. The 3 you pulled out to mock my point were all 15 or below.

The "luck" hypothesis in this case is remarkably contrived. It goes something like this "Look at this stat! see? Bad luck! Oh, here's another good one! see, how bad that luck is??

Sorry, but for most players, that in fact is the case. Really bizarre splits (with the exception of platoon splits, which are predictive) become normal splits far more often than not. You may find luck to be unsatisfying, but most of the time, that's what it is.

This is the same argument we would have in 2002, where people would say dismissing Damian Moss's season because of good luck was absurd.

Steve said this in #44

Projection systems and general findings are always best in expecting what will happen in a macro sense, but they don't necessarily override the particulars in individual cases.

I couldn't agree more. I love projections. I love them as a starting point, to then move up or down from in a subjective manner based on my own interpretations of whatever factors may exist that might not be accounted for strongly enough in the projection system. I realize that systems like ZIPS or PECODTA take a heck of a lot into account, but the "adjustments" still have to be smoothed out in order to make the system work on a league wide basis. It is only through subjective interpretation that we can make sense of it all on a player to player basis. Of course some are better at "making sense of it" than others. ;)

Thought a few of you might find interesting.

ooppsss sorry.

Try this one

One thing about Johnson's bad luck with runners on, though: he had put those runners on in the first place. I'm sure if you looked at his stats with the bases empty in innings where he allowed at least one baserunner, they'd be abysmal (and I yes, I realize that's necessarily a circular argument). The point, however, is that his performance from the stretch doesn't really counter the argument that those stats were the result of just suddenly losing it at some point in many of his starts. However, that in turn, doesn't really counter the argument that his losing it was injury-related.

shoewiz - it's important to remember that if you start with something like zips projections and then adjust them by hand (except in cases where zips will obviously be wrong, like francisco liriano 2007), you'll almost never do better than the projection system itself. there are a few articles about this from a few years ago - i wonder if they're still on the site any more.

shoewiz - it's important to remember that if you start with something like zips projections and then adjust them by hand (except in cases where zips will obviously be wrong, like francisco liriano 2007), you'll almost never do better than the projection system itself. there are a few articles about this from a few years ago - i wonder if they're still on the site any more.

I am not claiming to be able to forecast better than ZIPS. But I am sure that on some players I will have insight that will make my projection for that particular player more accurate than his ZIPS projection. But if I try to do that for every player in the league, of course ZIPS will kick my butt.

* Johnson’s BABIP with the bases empty was .239. With runners on, it was .369. This is almost certainly a matter of luck.

* So perhaps the big difference is that Johnson was getting hit up for extra bases more often with runners on base? Well, this is true: his home run rate was nearly twice as high with runners on (4.6%) than with the bases empty (2.5%). But the funny thing is that Johnson was actually doing a better job of keeping the ball on the ground with runners on. His groundout-to-airout ratio was 1.09 in those situatuons, as opposed to 0.78 with the bases empty. In other words, more bad luck.

People who actually watched Johnson pitch can verify that he simply sucked. It wasn't bad luck. It was bad pitching.

The only number worth looking at is 5.00.

I think it's pretty safe to assume then that the DBacks never had anyone actually watch Johnson pitch.

Here's some statistical evidence for you.
OPS against;
none on 1 or 2 outs: .524
none on: .593
one man on first: .854
runners on: .927
2+ runners on: 1.000
bases loaded: 1.167

Worst luck ever? You decide.


It's all very simple actually. He was being paid off by the mob to throw games. He did the best he could but the Yankee offense still found a way to get him 17 wins. Now he has to leave town to escape mob retribution.

Just out of curiosity. Why do you measure pitching effectiveness by OPS? Doesn't that just deliberately make no attempt to filter out luck.? At least people who look at things like xERA, DIPS, BABIP are making some attempt to separate skill from luck.

But you saw it with your own eyes, and of course your eyes know how to differentiate between bad luck and poor ability, so who can argue with that.

I used OPS against because it represents what actually happened. I used it to try to show that the pattern to his actual effectiveness was so outrageously unlucky that luck is the least likely of the possible explanations. I'll admit it could've been luck. It's theoretically possible to win the lottery 5 times in a row too. This is the kind of luck necessary to explain what was painfully obvious by observation. The whole point of filtering out luck is that it shouldn't be consistent or predictive of the future. In this case those of us who saw him pitch saw his "luck" become horrific about 25 times for the span of an inning or so then correct itself to an extreme degree. The man had consistent, predictable and frequent meltdowns. This was easily observable and discussed ad nauseum as it happened. Luck is a very, very, very unlikely way to explain such a pattern.

t's theoretically possible to win the lottery 5 times in a row too.

It's not comparable - it's unlucky in a single player, though not even in the same realm as a lottery, but it's not some horrifically unlikely event. In fact, it's likely that a few pitchers in the league would have that split, purely by random chance.

In this case those of us who saw him pitch saw his "luck" become horrific about 25 times for the span of an inning or so then correct itself to an extreme degree.

I have little doubt we're talking about some selective bias here. Our brains are hopeless at sorting out these kinds of long-term observations.

This is no different than 2005-2006 when people pointing out Curt Schilling's high ERA in 2005 as evidence he was done. Upon rebuttal that Schilling's extraordinarily high BABIP was impossible to be an ability, the response was "throw out those numbers! DIPS theory doesn't matter, I saw him pitch all season and Schilling is done!"

Reading this thread has been amusing. This is simply another excellent move by Brian Cashman who continiues to astound. Detractors of this deal who seem to be focused on ZIPS projections and the "luck" aspect of the men on base/bases empty splits are missing the forest for the trees: the Yankees were going to have to pay $20M (with luxury tax) for a 43 year old pitcher who is coming off back surgery. The "luck argument does not grab me at all, as I am another peson who actually watched the games and saw flat sliders being hammered all over the field. But in any case, ZIPs projections and splits simply do not take into account the risks asociated with back surgery. Brian Cashman had to take that into account and did. And the fact that he got some marginal value in return is impressive. I suspect Arizona could have had Johnson for nothing at all if they had just waited another month.

Of course, there is always the chance the Randy Johnson recovers from his surgery and has a good year. But there is almost certainly a far greater chance that he breaks down at some point and can't pitch at all. Or worse, performs poorly and forces the team that pays him all that money to stay with him as he struggles.

In Arizona's favor, I suppose, is that the desert is a better place for someone with back problems than New York.

Detractors of this deal who seem to be focused on ZIPS projections and the "luck" aspect of the men on base/bases empty splits are missing the forest for the trees:

What detractors?? It's almost unanimously agreed that the Yankees did well to get rid of a pitcher they didn't want by paying just $2 million and getting prospects in return. I don't think even the most rabid D-backs fans are calling this a terrible trade for the Yankees. At worst it is a win-win situation for both sides.

Brian Cashman had to take that into account and did. And the fact that he got some marginal value in return is impressive.

Really? Just eye-balling the projections for all the starting pitchers who changed teams this offseason (free agent signings and trades), Johnson has one of the best, if not the best, projection for 07. And he'll be signed for only 2 years at less money per year than Meche or Padilla, not to mention Zito or Schmidt. So getting some marginal value for that is impressive?

I suspect Arizona could have had Johnson for nothing at all if they had just waited another month.

This deal took over 3 weeks of discussions to complete. You think the Dbacks should've dragged it out until the end of January? I don't. It's nice that the Dbacks know what he'll cost in 07 and 08 so they can start addressing their arb-eligible players, and maybe begin discussing the general parameters of long term deals with some of the players they might want to keep (i.e., Davis, Valverde, Hudson, etc.) Saving a month of drawn out negotiations with NY was worth giving up the "marginal" players you're talking about.

Of course, there is always the chance the Randy Johnson recovers from his surgery and has a good year. But there is almost certainly a far greater chance that he breaks down at some point and can't pitch at all.

I'm sure you've quantified those and will give us the numbers. Include standard deviations.

it's likely that a few pitchers in the league would have that split, purely by random chance.

How are you figuring these odds? Those splits are not bad ones they are impressively horrid, why do you say otherwise? Can you find me just one season by one starting pitcher(150+ IP?) in the last few decades with such an extreme and progressively worsening spread of OPS against by situation? As I said before I looked those numbers up because they show there was a "good" Randy and a "bad" one and OPS against by situation was the best I could think of to progressively separate the two that was easily available. I wouldn't be too surprised if there were seasons where a pitcher had the same trend but I'd be rather shocked if there was another where it was so very extreme.

I have little doubt we're talking about some selective bias here. Our brains are hopeless at sorting out these kinds of long-term observations.

Millions of people saw it, dozens of articles written about it, even more discussions of it on here and elsewhere. The mans performance was thoroughly scrutinized and tracked by a huge amount of people. Considering the growth of the media and the internet and him being a high profile HOF pitcher in the media capitol of the world on probably the most popular sports team in the world it's possibly one of the most well documented and overly deliberated seasons by any athlete in history. I don't remember ever seeing anyone suggest that these frequent meltdowns weren't actually happening or that it was luck at the time. Are all of our brains really THAT hopeless?
Even if that's the case, are those situational OPS against splits really not direct evidence of a pattern of frequent meltdowns? What stats would you use to show such if it really did happen?
If it did happen is a consistent pattern of dominance broken up by short, extreme meltdowns really reasonably explainable by luck? I'm not some uneducated rabid anti-Johnson Yankee fan. I understand DIPS and selection bias and took 4 semesters of statistics at a good school. The assertion that such a pattern is luck seems astronomically small to me.

This is no different than 2005-2006 when people pointing out Curt Schilling's high ERA in 2005 as evidence he was done. Upon rebuttal that Schilling's extraordinarily high BABIP was impossible to be an ability, the response was "throw out those numbers! DIPS theory doesn't matter, I saw him pitch all season and Schilling is done!"

Dan, you cannot seriously be arguing that the difference in Schilling's performance from 2005 to 2006 was due to luck. This type of reasoning is why "traditionalists" tell statheads to watch the games and get their heads out of the spreadsheets.

In 2005 Schilling was coming off ankle surgery and the high of bringing a WS to Red Sox Nation. When he reported to camp he was out of shape and unable to repeat his delivery. He got knocked around and if he didn't improve he was toast. It wasn't selective memory that caused people to think Schilling stunk, he did stink(by his standards).

Now Schilling has enough "true talent" so that in 2006,after a winter of conditioning, he could almost get back to his 2004 form. Now maybe it looks to you like that is "regression to the mean". In a VERY large sense it is regression to the mean, but not because of luck. Due to injury and lack of conditioning, Schilling was not a good pitcher in 2005. It was reasonable to predict that given his prior history, Schilling would be good again i.e. regress to the mean. Schilling worked his **s off over the winter, got back into form and pitched much better in 2006. But concluding that the difference between 2005 and 2006 was luck, ignores what was happening on the field. If Schilling pitched the way he did in 2005, Schilling *was* done.

As for the Big Unit, the issues *are* similar. We know RJ was a great pitcher who was suffering from injury in 2006. His back has now been surgically repaired. His knee will always be a problem. The great unknown is whether RJ will recover from surgery to become the 2005 RJ (3.79ERA 114 ERA+) or will he be the 2006 RJ (5.00 ERA 88ERA+. My guess is that post surgery and in the warm weather, he will be closer to 2005. But that isn't because he was UNLUCKY in 2006, it will be because he will be a better pitcher in 2007.

Sometimes statistical principles do not explain what is happening on the field, they may actually mask it.

Fwiw, Keith Law appears to have joined the "detractors" of this deal; while most of his analysis is hidden behind the insiders wall on espn, the gist is summarized nicely in his first 3 paragraphs. In sum, the Dbacks didn't give up anything meaningful for the chance to see what RJ has got left in the tank...

Linky

So it's not possible that RJ was UNLUCKY in 2006 AND he was less effective because of his back?

Luck and health aren't mutually exclusive.

As to the merits of this deal, I agree with Levski that this is a good one for AZ. It's a bit of a risk, but RJ could be very good for the Snakes and they gave up nothing from their core and none of their top prospects.

As for the NYY, not bad from their POV. They shed salary, get a solid middle reliever and 2 arms and maybe a ss prospect they can deal. They didn't want RJ, he didn't want them and now they can go for Clemens and/or give pavano, kartens, hughes et al a legit shot at the rotation. I would have liked this deal a lot more for the Yankees if they had paid a little more for Owings. I've never seen either pitcher, but from everything I've read, Owings has much more upside than Ohlendorf.

Really? Just eye-balling the projections for all the starting pitchers who changed teams this offseason (free agent signings and trades), Johnson has one of the best, if not the best, projection for 07.

Yes, except for the minor triviality that you did not address: the fact that the the Unit had back surgery during the off-season.

So it's not possible that RJ was UNLUCKY in 2006 AND he was less effective because of his back?

Luck and health aren't mutually exclusive.

Good point,luck could have been a contributing factor. Nevertheless RJ served up far too many lasers and big flies.

I also think Randy will pitch better in the warm weather.

Yes, except for the minor triviality that you did not address: the fact that the the Unit had back surgery during the off-season.

I'm not sure why people view the surgery as a huge negative strike against him. I personally think of his back injury as the negative, and the surgery as a positive way to address it. He should be better BECAUSE of the surgery, not worse. I'm puzzled why people fail to see the difference. How effective he'll be with his repaired back is a valid question; the surgery itself, however, should not make him worse than he was in 06.

It's like saying Schilling would be better off pitching with his injured ankle, with the tendon stapled to his skin, like he did in the 04 playoffs, instead of having proper surgery to repair it. It doesn't make any sense.

Back surgery is always risky, particularly at age 43. Back surgery is something that should be resisted if at all possible; its not something where a successful recovery is automatic, particularly when you go and subject it to the same stresses that brought on the problem in the first place. A lot of ball players have recovered nicely from ankle or knee surgery. You'd be hard pressed to name a lot of ballplayers, particularly pitchers, who have come back 100% from back surgery.

You'd be hard pressed to name a lot of ballplayers, particularly pitchers, who have come back 100% from back surgery.

Randy Johnson?

And it's not like this is a sudden back injury that required surgery - Johnson's back has been bad for more than a decade. If the surgery was successful and repaired the herniated disk that's been bother him for quite awhile, I don't know why one would expect him to pitch better with a messed-up back than a successfully repaired (or, as is more likely for the back, mitigated or whatever a more appropriate word is) back.

What was the surgery he had, anyway? Fusion or what?

I also consider the back something of a sideshow. What about the knee? That was the problem in 2003, it was repaired with this Synvisc thing, which might have a time limit, and could affect the landings from the stretch.

I havent read every one of these threads, so I don't know if I've seen this specifically, does ZIPS compare ages, and does Johnson benefit specifically because of his age, as few bad pitchers make it to that age. In other words, if ZIPS thought he was a younger age when more pitchers would get worse, would his projections look different?

A couple of years ago, Pickering had some outlanding PECOTA projections, and I think a portion of the projection was based on his weight. The few players that played at his weight were very good, so Calvin got a boost he probably shouldn't have. Just wondering if RJ is benefiting in a similar way.

As I said before I looked those numbers up because they show there was a "good" Randy and a "bad" one and OPS against by situation was the best I could think of to progressively separate the two that was easily available. I wouldn't be too surprised if there were seasons where a pitcher had the same trend but I'd be rather shocked if there was another where it was so very extreme.


If you want to suport your argument that Johnson for some reason pitched worse with man on base, and this is somehow psychologically or mentally, caused why not show us his k-rate, BB-rate, command, and BABIP in the various situations.

Other people have done this showing that all the skill measures remain the same, only the BABIP change. I really don't quite understand your point or your supposed reasoning. It's as if you believe that you saw it, and some dumb/lazy writers wrote about it, therefore making meaningful statistical analysis null and void. Using OPS against a pitcher to measure skill is silly. It only measures effectiveness, and we already knew that he wasn't effective by the 5.00 ERA. What we need to determine is whether his lack of effectiveness was due a skill decline or bad luck. In fact it was due to both.

In 2005 Johnson's xERA was 3.05 and his actual ERA was 3.80. In 2006 those values are 3.83 and 5.00. So Johnson's overall ERA decline of 1.20 was caused by a .75 decline in skills, and a .45 decline in luck. Another indicatore of his luck decline is his strand rate which went from 72% in 2005 to 62% in 2006.

What stats would you use to show such if it really did happen?
If it did happen is a consistent pattern of dominance broken up by short, extreme meltdowns really reasonably explainable by luck? I'm not some uneducated rabid anti-Johnson Yankee fan. I understand DIPS and selection bias and took 4 semesters of statistics at a good school. The assertion that such a pattern is luck seems astronomically small to me.


Did you finally pass your stats course after your 4th attempt?

If you want to show a skills meltdown with men on base there needs to be decline in his k-rate or an increase in his BB-rate. Not a change in BABIP, which is the effect that causes the OPS pattern you have shown. A dramatic change in BABIP is luck, almost by definition.

A couple of years ago, Pickering had some outlanding PECOTA projections, and I think a portion of the projection was based on his weight.

I don't know how Pecota does it, but I put a cap on my weight factors. Yeah, a 230 pound player should hit more HR than a 165 punder, but there's a decreasing marginal gain here. I'm not going to extrapolate the HR gains to a 300 pounder.

Dan, you cannot seriously be arguing that the difference in Schilling's performance from 2005 to 2006 was due to luck. This type of reasoning is why "traditionalists" tell statheads to watch the games and get their heads out of the spreadsheets.

In 2005 Schilling was coming off ankle surgery and the high of bringing a WS to Red Sox Nation. When he reported to camp he was out of shape and unable to repeat his delivery. He got knocked around and if he didn't improve he was toast. It wasn't selective memory that caused people to think Schilling stunk, he did stink(by his standards).


Schilling's xERA for the past five seasons:

2002 - 2.31
2003 - 2.59
2004 - 3.05
2005 - 3.83
2006 - 3.23

Certainly Shilling pitched worse in 2005 than he had in the past. If you want to say that an xERA of 3.83 is "stinking", then that is up to you. A lot of pitchers would die for a "bad season" like that.

But he certainly didn't pitch at a skill level anywhere near a 5.92 ERA, which I think was Dan's point.

Certainly Shilling pitched worse in 2005 than he had in the past. If you want to say that an xERA of 3.83 is "stinking", then that is up to you. A lot of pitchers would die for a "bad season" like that.

But he certainly didn't pitch at a skill level anywhere near a 5.92 ERA, which I think was Dan's point.

When did we go through a wormhole into BPro circa 2001? The idea that we can identify "luck" with this sort of precision is not at all supportable statistically. We can identify tendencies in the aggregate, but these are not universalizable in the future, and they are most certainly not universalizable into the past.

Curt Schilling sucked in 2005. He was incredibly hittable becuase he was unable to command his splitter, which led to lots of hanging splitters, or, when he shelved that pitch, too many fastballs and too many hitters sitting on the fastball. The idea that he was "pitching at a 3.83 ERA skill level" pitcher is ludicrous, and a terrible misuse of statistics.

The idea that he was "pitching at a 3.83 ERA skill level" pitcher is ludicrous, and a terrible misuse of statistics.


Could it be that the statistics are only used properly when they corroberate an already formed opinion.

Schilling had an 87/22 K/BB ratio. That in not indicative of a 5.92 ERA skill level.

His BABIP from 2002 to 2005 - .310, .310, .300, & .390.

His strand rate from 2002 to 2005 - 71%, 76%, 73%, & 63%.

No skill or bad luck here. Just a pitcher who sucked!! To try to get any useful information from these stats about what really happened to Schilling is a "terrible misuse of them". Live and learn I guess.

From 2004 to 2005 his xERA went from 3.05 to 3.83, while his actual ERA rose from 3.26 to 5.92. In my feeble mind this would indicate that the total increase of 2.76 is made up of 0.78 skill decline and a 1.96 decline in luck.

I know that this is totally wrong, and if I had been a rabid Red Sox fan watching all of his games I would know the truth. But failing that, this explanation suits me just fine, which rather unexplicably was corroberated by his 2006 seasons.

I know that this is totally wrong, and if I had been a rabid Red Sox fan watching all of his games I would know the truth. But failing that, this explanation suits me just fine, which rather unexplicably was corroberated by his 2006 seasons.

No, it wasn't. Curt Schilling regained command of his splitter in 2006, and was a completely different pitcher.

Pretending that qualitative differences don't exist is a bald misuse of statistics.

I tend to think that with another 100 innings in 2005, Curt Schilling's HR rate would have gone way up, and his xERA would have headed up toward his ERA. He was pitching badly. The fact that it showed up in his BABIP is the only fluke.

In my feeble mind this would indicate that the total increase of 2.76 is made up of 0.78 skill decline and a 1.96 decline in luck.

That's a misunderstanding of regression. First, it has been shown by Tom Tippett and David Gassko and Voros McCracken that pitchers have consistent levels of BABIP skill. Second, you simply cannot divide up "skill" and "luck" that way - randomness is a factor in all statistics, and the tendency for heavy regression to produce better predictions in the aggregate does not and cannot mean that the gap between regressed stats and non-regressed stats is utterly meaningless.

And it's not like this is a sudden back injury that required surgery - Johnson's back has been bad for more than a decade. If the surgery was successful and repaired the herniated disk that's been bother him for quite awhile, I don't know why one would expect him to pitch better with a messed-up back than a successfully repaired (or, as is more likely for the back, mitigated or whatever a more appropriate word is) back.

Dan, it is possible to see worst and optimistic projections for RJ? When I look at RJ, I see a pitcher with an unusually high confidence interval in predicting his performance next year simply because we don't know what his health will be like next year. The surgery could very well be successful. Consequently, combined with a better windup/stretch split and pitching in the NL, RJ could have an excellent year for the D-backs. On the other hand, RJ's body will break down at a certain point preventing him from pitching well. Next season could very well be that season, thus his peformance could balloon next year. Or maybe some middle road scenario will happen next year (i.e. the surgery helps a little bit but there's still pain, other small injuries pop up) and he pitches to a tune of 5.00 ERA again. At this point, I just don't know what RJ's health will be like next thus I tend to think that there should be tremendous variation in his predictions for next year.

Did you finally pass your stats course after your 4th attempt?
If you want to show a skills meltdown with men on base there needs to be decline in his k-rate or an increase in his BB-rate. Not a change in BABIP, which is the effect that causes the OPS pattern you have shown. A dramatic change in BABIP is luck, almost by definition.

Nice one, but I specifically stated what I was trying to show already and it has nothing to do with what you are talking about. I was attempting to give evidence that shows there was a pattern to his actual effectiveness that was consistent, predictable and easily observable throughout the entire season in a way that totally defies an explanation based on luck. This is why, as far as I have seen, NO ONE who saw him pitch most of his games is in agreement that his main problem was bad luck.

I think of Schillng's 2005 as being the poster boy for what MWE reminds us at the beginning of every good DIPS thread: DIPS measures only those pitchers who have a demonstrated of being able to pitch in the major leagues.

Schilling in 2005 did not demonstrate that skill. Schilling in 2006 did. After 2005, there was a good reason to bet that he would have that skill in 2006: he demonstrated it repeatedly in prior years; he demonstrated the ability to not BB people in 2005; and he demonstrated the ability to SO people in 2005. Those all bode well for having the ability to be a major league pitcher - from observance you could see he didn't have that skill in 2005. However, when you do a study looking backwards at Schilling's season as part of 1500 other seasons, you'll see only that 2005 doesn't correlate with other seasons and someone may shout out "luck". Of course it doesn't correlate: if so, he'd be out of a job.

Similarly, either RJ did have that skill or did - enough Yankees fans have stated that he didn't, and I'd believe them. I'd bet, however, on a player who had that skill for so long having it in 2007, esp when we see that he had good SO & BB rates. That is a bet with some risk, but I'd still take it. But, if he doesn't - he'll be out of a job.

Schilling in 2005 gave up a ton of hits when they put the ball in play against him. Johnson didn't have the same problem until runners got on base against him.

If it was related to pitching from the stretch, its strange that while he was worse with a runner on first only, he didn't get really bad until they moved into scoring position. You'd think that if pitching from the stretch was killing him, he could use the windup more once runners had already advanced into scoring position.

With noone on, average batter vs Johnson looked like Cristian Guzman. With runners on first only, he was Carlos Lee. With RISP, he's Albert Pujols.

And then there's Ryan versus Carlton: there was the year they both passed Walter Johnson for the strikeout record and passed it back and forth between them - but you didn't know which one would hang on longer to claim the record in the end.

I don't remember anyone at the time thinking Carlton had a good chance to end up ahead of Ryan. Maybe that was just me, though. But IIRC the zeitgeist was it was just a matter of time before Ryan passed him for good.

That's a misunderstanding of regression. First, it has been shown by Tom Tippett and David Gassko and Voros McCracken that pitchers have consistent levels of BABIP skill.

While pitchers have consistent levels of BABIP skill...

1. Schilling's established BABIP is quite good.

2. They don't have a huge amount of control over BABIP - maybe 10% of BABIP variation is under control of the pitcher in the first place per Tippett's research.

3. Most of the BABIP variation that is under control of the pitcher is already accounted for in DIPS ERAs - 75% per Gassko. That leaves something like 2.5% of total BABIP variation to be the result of a pitcher's own skills beyond his peripherals.

The storyline that Schilling wasn't unlucky, but was actually a 5.69 ERA 'true talent' pitcher in 2005 is pretty absurd to me. When a guy doesn't have his control over his splitter, it will impact his peripherals, not only his BABIP. Now, it did impact his peripherals, but not the tune of a 5.69 ERA. Why people who acknowledge the otherwise statistical validity of current BABIP research refuse to seriously consider that a guy may have been unlucky in 93 innings is beyond me. It's not even like Schilling has a whole career of poor luck on balls in play. He has 93 innings.

The case for Johnson's situation not having much to do with luck is even more absurd.

I don't remember anyone at the time thinking Carlton had a good chance to end up ahead of Ryan. Maybe that was just me, though. But IIRC the zeitgeist was it was just a matter of time before Ryan passed him for good.

I remember it very differently. Carlton had the lead when he was still having excellent seasons with the Phillies. Many people were predicting that Carlton would outlast Ryan because he was the better pitcher - few thought Carlton would go downhill as quickly as he did. At the same time, no one was predicting that Ryan would have anything like the end of a career that he did.

The storyline that Schilling wasn't unlucky, but was actually a 5.69 ERA 'true talent' pitcher in 2005 is pretty absurd to me. When a guy doesn't have his control over his splitter, it will impact his peripherals, not only his BABIP. Now, it did impact his peripherals, but not the tune of a 5.69 ERA. Why people who acknowledge the otherwise statistical validity of current BABIP research refuse to seriously consider that a guy may have been unlucky in 93 innings is beyond me. It's not even like Schilling has a whole career of poor luck on balls in play. He has 93 innings.

I think it's more likely that Schilling was lucky in his other component stats, and in particular very lucky in his HR rate. I remember a huge number of warning track flies in 2005. I tend to think that because he was so lucky at keeping the ball in the park, he kept pitching in the zone more than he should have, and managed to build a good K/BB rate that was also probably unsustainable. Given that we know that regression of all stats creates better projections, it makes sense that in some cases it's the DIPS component stats that reflect random chance more than the BABIP and other stats which are regressed 100% in traditional DIPS.

I think it's more likely that Schilling was lucky in his other component stats, and in particular very lucky in his HR rate. I remember a huge number of warning track flies in 2005. I tend to think that because he was so lucky at keeping the ball in the park, he kept pitching in the zone more than he should have, and managed to build a good K/BB rate that was also probably unsustainable.

Regarding the sustainability of the K/BB rate - that's a different question from whether or not he was unlucky or not on balls in play in 2005.

However, I don't understand why you think he was lucky in his component stats to a significant degree. His HRs allowed per flyball rate was similar to 2004 - did he strike you as lucky on HRs in 2004?

His BB rate was worse than it's been in quite some time, and his K rate was pretty much in line with his norms. I don't see why you think he was lucky with these.

Given that we know that regression of all stats creates better projections, it makes sense that in some cases it's the DIPS component stats that reflect random chance more than the BABIP and other stats which are regressed 100% in traditional DIPS.

What now? Could you rephrase this, because I'm not sure I understand. I don't see how the general virtue of regression to the mean translates to it making more sense to regress DIPS component stats than BABIP sometimes. Or how this is one of those cases, should those cases exist.

I think of Schillng's 2005 as being the poster boy for what MWE reminds us at the beginning of every good DIPS thread: DIPS measures only those pitchers who have a demonstrated of being able to pitch in the major leagues.

Exactly. Also, what MCoA says is absolutely correct, - Schilling's splitter was not a major league pitch in 2005. This didn't just show up in his BABIP; he also had his lowest K rate and highest walk rate post-2000. An 87/22 K/BB is "good" in a neutral context, but for Curt Schilling, they were a drop-off, and an indicator that something wasn't right.

Could it be that the statistics are only used properly when they corroberate an already formed opinion.


In this case, no.

Exactly. Also, what MCoA says is absolutely correct, - Schilling's splitter was not a major league pitch in 2005. This didn't just show up in his BABIP; he also had his lowest K rate and highest walk rate post-2000. An 87/22 K/BB is "good" in a neutral context, but for Curt Schilling, they were a drop-off, and an indicator that something wasn't right.

First off, this isn't true - Schilling in 2005 posted the best K rate he's had in the AL. In was better than 2004, or 2005.

There is this curious tendency among some to find a singular reason why the basic trends of BABIP might not apply perfectly to a given pitcher, and to use that singular indicator as a reason to ignore the all of what we know about BABIP.

Just because his DIPS ERA may not have been a perfect representation of his talent doesn't mean he wasn't unlucky. The fact that an 87/22 K/BB ratio is a very poor one for Schilling is a sign that something is wrong with him, I agree. However, something being wrong with him doesn't mean that his true talent BABIP allowed is now .371.

Given that we know that regression of all stats creates better projections, it makes sense that in some cases it's the DIPS component stats that reflect random chance more than the BABIP and other stats which are regressed 100% in traditional DIPS.



What now? Could you rephrase this, because I'm not sure I understand. I don't see how the general virtue of regression to the mean translates to it making more sense to regress DIPS component stats than BABIP sometimes. Or how this is one of those cases, should those cases exist.


His tactic has now degenrated to confusing the opposition with nonsensical phrases. It's sort of like waving a white flag, subliminally, without ever having to admit that his whole premise is silly.

Hold up guys, we tend to keep our flame wars in the newsblog! :-)

Levski, shoewizard, robert s., et al, I'm sure this was covered somewhere in one of these Johnson threads, but does Arizona think that it has a shot next year? I thought that they were still in rebuild mode, but the NL West might as well be the Mexican Winter League to me. I just don't follow it that much.

Could it be that the statistics are only used properly when they corroberate an already formed opinion.

Schilling had an 87/22 K/BB ratio. That in not indicative of a 5.92 ERA skill level.

His BABIP from 2002 to 2005 - .310, .310, .300, & .390.

His strand rate from 2002 to 2005 - 71%, 76%, 73%, & 63%.

No skill or bad luck here. Just a pitcher who sucked!! To try to get any useful information from these stats about what really happened to Schilling is a "terrible misuse of them". Live and learn I guess.

From 2004 to 2005 his xERA went from 3.05 to 3.83, while his actual ERA rose from 3.26 to 5.92. In my feeble mind this would indicate that the total increase of 2.76 is made up of 0.78 skill decline and a 1.96 decline in luck.


Somebody just got his Baseball Forecaster.

First off, this isn't true - Schilling in 2005 posted the best K rate he's had in the AL. In was better than 2004, or 2005.

No, it was his lowest. K/9 is biased.

K/BFP:

2001 .287
2002 .311
2003 .288
2004 .223
2005 .208
2006 .219

That said, I agree that there was probably a little bit of bad luck in Schilling's 2005 BABIP.

I don't know why we need to argue if Schillings 2005 BABIP was bad luck or crappy pitching.

It should have been obvious to anyone that .390 was non-repeatable, and it turns out, it wasn't repeated.

I don't know why we need to argue if Schillings 2005 BABIP was bad luck or crappy pitching.

Can't speak for anyone else, but I'm just bored at work.

It should have been obvious to anyone that .390 was non-repeatable, and it turns out, it wasn't repeated.

Right, but it wasn't repeated due to qualitative changes in his pitching from 2005 to 2006. That is, Schilling's splitter started breaking again. He threw more hanging splits in 2005 than he had in probably his entire career up to that point, probably because of mechanical problems dating from his ankle surgery. As Josh argued above, of course Schilling's 2005 was unsustainable - he wouldn't be pitching in the major leagues if he was actually that bad. What I'm arguing is that qualitative changes were necessary in order for Schilling to avoid repeating his 2005.

Given what we know about sample sizes and regression, all stats compiled over 90 IP are vague estimates of "true talent". So, certainly, there was a lot of randomness in that .390 BABIP. But there was a whole lot of randomness in all his other numbers as well, and the focus on only one number to the exclusion of all the rest is problematic. As I've said, I expect Schilling's HR rate would have spiked with time if he kept pitching that way - to bibigon, I'm arguing that given how badly Schilling was pitching, his 2005 HR/FB rate was shockingly low. He was lucky to have a HR/FB rate at his typical levels, and I doubt he could have maintained it over another 100 IP.

At base, what I'm arguing is that Curt Schilling in 2005 was a bad pitcher, not a good one. The tendency to universalize aggregate findings like DIPS is problematic becuase it can cause us to miss all the interesting stuff that happens at the margins - all the low-IP, short-career pitchers who don't fit the model well, and all the other outliers, the injured, aging pitchers like Schilling, the Charlie Houghs and Tom Glavines. I'm arguing that Curt Schilling in 2005 was, based on my qualitative observation, backed up by the observations of many Sox fans I know, a bad pitcher. His DIPS suggests he was actually a good pitcher, but I think that DIPS, though it does demonstrate an aggregate trend within a certain population, in this individual case does not account for major factors in Schilling's struggles.

I don't know why we need to argue if Schillings 2005 BABIP was bad luck or crappy pitching.

Sure, but I think the point that we're getting at is whether Schilling was a good or bad pitcher in 2005. His results were very bad. His DIPS is quite good. That seems to be a real disagreement, and one that may have useful applications to thinking about other issues in pitcher statistics and projection.

There is this curious tendency among some to find a singular reason why the basic trends of BABIP might not apply perfectly to a given pitcher, and

Bibigion I liked your posts quite a bit above but I believe taking any "basic trend" and applying to a given pitcher is commonly called the ecological fallacy (i.e. taking an aggregate number like a trend and applying to an individual) so there's a good reason not do this. And on a related noted, I have a few questions about DIPS and LIPS which I love to hear your answer (or anyone else familiar with DIPS and LIPS). Until a few days ago, I didn’t know anything about DIPS and LIPS so let me apologize ahead of time if my interpretations and questions are misguided. According David Gassko’s recent article on DIPS and LIPS, Rodrigo Lopez was an extremely unlucky pitcher. According to LIPS, Lopez gave up about 1.19 run average more than what is expected due to bad luck. As an O’s fan, I watched Lopez quite a bit so I can see where MCoA is coming from regarding Schilling. Lopez was awful in 2006 and my initial thought when I ran across that statistic is that it doesn’t account how hard hit the ball is put into play (the model is underspecified in a sense). While one could interpret Lopez’s extremely high rate of ball put into play as hits as bad luck, I would interpret it as batters were just, for a lack of a better word, raking the crap out of the ball against Lopez thus Lopez wasn’t a unlucky pitcher because he was giving up laser beams left and right due to making bad pitches (a controllable element). I’m not denying some of the hits that fell in weren’t due to back luck (e.g. maybe a gust of wind floated a few line drives over an OF’s head), but I suspect Lopez was also just getting knocked around hard which is more responsible for his demise than some unlucky breaks. Similarly, I noticed Wang and Halladay were high the lucky list. Is it really luck if they purposively induce a lot of weakly hit groundball since both pitchers use heavy sinkers? Is there a way to measure the how hard a hit a pitcher typically gives up? Or this element already controlled for using other variables? If you know a cite that cover this ground, I would love to read it.

157 - the problem is that your idea is difficult to prove there, that those things were the results of skill rather than luck. I see no evidence to suggest that it is true that guys like Wang and Halladay actually have the skill to induce lots of weakly hit groundballs, as opposed to the more general idea that they have the ability to induce groundballs, and that they were somewhat lucky on the results of those balls.

Obviously it's not luck if they have that skill, but it needs to be demonstrated first that they actually indeed do have that skill on a level which exceed our current understanding of how BABIP works. If they do have that skill on such a level, then we would expect it to such up in a macro analysis of the data - which so far, it has not. (Beyond the general understanding of how much impact a pitcher has on BABIP that is.)

the problem is that your idea is difficult to prove there, that those things were the results of skill rather than luck. I see no evidence to suggest that it is true that guys like Wang and Halladay actually have the skill to induce lots of weakly hit groundballs, as opposed to the more general idea that they have the ability to induce groundballs, and that they were somewhat lucky on the results of those balls.

Fair enough. I suspected as much. "Weak" groundball is definitely a subjective evaluation thus I imagined it would be difficult to measure whether a pitcher like Wang induces a lot of dribblers to infielders for easy put outs. But isn't there some measure of how many groundball a pitcher induces? Do pitchers who induce a lot groundballs typically produce better DIPS and LIPS outcomes?

f they do have that skill on such a level, then we would expect it to such up in a macro analysis of the data

Is this necessarily true? If the skill is constrained to a small number of players - and there is also truly a large "luck"-based fluctuation in the results, wouldn't it be very difficult to tease out this skill by using macro analysis?

Do pitchers who induce a lot groundballs typically produce better DIPS and LIPS outcomes?

Here's DSG's take on batted ball types and DIPS. And
here'sthe followup.

I think the system incorporated in the second article is now called LIPS.

You may also be interested in this thread.

Jesus, Mary and Joseph! Schilling was only healthy enough to pitch 93 innings in 2005. It is absolutely stunning that anyone would seriously argue that the injuries that kept him from taking the mound didn't also impact his effectiveness when he did pitch. He wasn't unlucky in 2005, unless you just mean that being hurt is bad luck.

Still the last few posts got me thinking some. Anybody know where I can find data on GB BABIP vs FB BABIP? Both for the league and individuals.

</i>fixed?

And the Randy Johnson 2006 season is just more of the same. Look for solid improvement in 2007.

Care to set a friendly wager on that?

All this seems implausible to me...

It seems implausible to you because you keep trying to put it in the DIPS101 box. Major League caliber pitchers differ little in their true ability to prevent hits on balls in play. MCoA's argument is that in 2005, Schilling was not major league caliber, due to injury. The peripheral stats do not disprove this hypothesis. So just how bad does a BABIP have to get before it's not hard luck anymore?

I see no evidence to suggest that it is true that guys like Wang and Halladay actually have the skill to induce lots of weakly hit groundballs, as opposed to the more general idea that they have the ability to induce groundballs, and that they were somewhat lucky on the results of those balls.

My memory of the Tibbetts research is that knuckleballers and "trick" pitchers like Rivera consistently have lower BABIPs. Does the fact that Wang throws a 95 mph heavy sinker constitute evidence that Wang may have the skill to induce lots of weakly hit ground balls and thus a lower BABIP? I think so, even knowing that sinkerballers generally do not have a lower BABIP. I think his sinker is like Rivera's cutter.

So just how bad does a BABIP have to get before it's not hard luck anymore?

Isn't that a circular argument? He wasn't an MLB caliber pitcher because his BABIP was so high, which was because he wasn't an MLB quality pitcher.

Or, just how good does a K:BB ratio have to get before it's not good luck anymore?

the fact that Wang throws a 95 mph heavy sinker constitute evidence that Wang may have the skill to induce lots of weakly hit ground balls

No because there is no way to measure "heavy" sinker. Plenty of players throw a sinker but how do you objectively measure who's sinker is heavier?

I think so, even knowing that sinkerballers generally do not have a lower BABIP. I think his sinker is like Rivera's cutter.

I also very much think that Wang was able to induce a lot of fieldable GB outs. I don't buy that Wang was lucky (or Lopez was unlucky) but bibigon is right, there's no proof showing that.

Care to set a friendly wager on that?

Any takers?

Care to set a friendly wager on that?

What are the terms? RJ had an 88 ERA+ last year. Would you consider a 100 ERA+ to be "solid improvement?"

I can buy that due to his injuries, Schilling's 2005 true talent babip was higher than normal.

But there's still a huge element of luck in there. No way his true talent was .390. Glendon Rusch is about as hittable as they come, and he's around .335 for his career. There may be some pitchers whose true talent is .390 or above. I would be one of them. A batting practice machine would be one of them. But there's no way a guy good enough to post a 4/1 K/W ratio is one of them.

Are even non-pitchers taking the mound during blowouts that bad?

What are the terms? RJ had an 88 ERA+ last year. Would you consider a 100 ERA+ to be "solid improvement?"

Well, considering that a 100 ERA+ for Arizona pitchers last year would have been 4.76 ...

My thinking is that the projection of RJ presented at the top of this thread is unrealistically optimistic. I expect he will neither pitch as many as 216 innings, strike out as many as 206 batters, nor post a 3.63 ERA.

So what would you say to something like this: Randy Johnson will fall short of that projection in all three categories by at least 15%; that is, he will pitch no more than 184 innings, strike out no more than 175 batters, and post an ERA no lower than 4.17. That's my proposition.

If I'm right on all three counts, whoever takes me up on the wager will have to sponsor three $5 bb-ref player pages for a year. If I'm right on two, two $5 pages, and I'll sponsor one. If I'm right on one, you sponsor one $5 page, and I'll sponsor two. If I'm right on none of them, I'll sponsor three $5 bb-ref pages.

How about that?

Levski, shoewizard, robert s., et al, I'm sure this was covered somewhere in one of these Johnson threads, but does Arizona think that it has a shot next year? I thought that they were still in rebuild mode, but the NL West might as well be the Mexican Winter League to me. I just don't follow it that much.

I belive Josh Byrnes's sole motivation this year has been his desire to have the most left handed starters in the NL West. And win 90 games.

Isn't that a circular argument? He wasn't an MLB caliber pitcher because his BABIP was so high, which was because he wasn't an MLB quality pitcher.

Or, just how good does a K:BB ratio have to get before it's not good luck anymore?

That's not the argument I was trying to make, and I never said that the K:BB ratio was good luck. What I am trying to say is that if ability to control the three true outcomes really is independent of BABIP ability, then quoting DIPS stats can never address the question of whether a pitcher has the minimum level of BABIP ability required to succeed in the major leagues. Now, if you think that I have the same level of BABIP skill that Greg Maddux has, then you won't like this version either. But if you grant that DIPS only holds for the highest levels of the game, and that it holds precisely because everyone capable of pitching at those levels is very, very good at preventing hits on balls in play, then it seems at least plausible that a pitcher could suddenly lose his BABIP skill without losing his DIPS skills.

There were "some" comments saying that the unnatural BABIP might be correlated to some mythical luck factor. Isn't it an easier argument to say that it was related to tendencies? To elucidate..
There was a comment earlier in the thread which said RJ allowed a very high % of runners to score from 3B. What if that is because he didn't trust his control of the slider and pounded the zone with the fastball? And since his FB had lost a little zip, he became predictable AND hittable, which might lead to those obscene numbers...
Or it could be that he was tipping his pitches from the stretch last year. BABIP is not entirely whimsical. Just because we don't know all the factors affecting it doesn't mean we attribute all changes in BABIP to luck.

Levski, shoewizard, robert s., et al, I'm sure this was covered somewhere in one of these Johnson threads, but does Arizona think that it has a shot next year? I thought that they were still in rebuild mode, but the NL West might as well be the Mexican Winter League to me. I just don't follow it that much.

I think that the Dbacks feel they have a chance to compete for the division but they were not willing to give up any of the top prospects or young players already in the majors to make a "run at it" this year. Instead, they looked to make incremental moves to get better, within budget, without sacrificing any of the future core.

Mission accomplished.

The tradeoff is that some of the young pitchers in the organization won't get a shot in the majors this year because of the presence of Livan and Davis and Johnson, and more than likely they will end up losing a couple of decent pitchers over the next couple of years because they just were not able to give them enough starts in the majors to "prove themselves" before they ran out of options.

This is the only part of the Johnson deal as it has finally come about that bothers me. Giving up Ohlendorf and Jackson and Gonzalez is not going to hurt this teams future. And the money has been structured in such a way as to not hamstring the organization at all financially. But the cost to compete now is impeding the potential development of 4-5 guys that could potentially have been the core of a cheap rotation for a long time.

Then again, maybe Josh Byrnes determined none of those guys are likely to be very good and so he is going to start over trying to build up the organizational pitching depth through the draft and trades.

Or maybe he is just saying screw it, TINSTAAPP ... and he will use position players as trade bait and payroll flexibility to keep bringing in veteran pitching over the next 4-5 years.

I am not really sure what the game plan is are far as future rotation options.

So what would you say to something like this: Randy Johnson will fall short of that projection in all three categories by at least 15%; that is, he will pitch no more than 184 innings, strike out no more than 175 batters, and post an ERA no lower than 4.17. That's my proposition.

So the cutoff is a 114 ERA+? I think that's a substantial imporvement (basically rebounding to his 2005 run prevention level). I think he'll be right around there, but I wouldn't bet he'd be better than that. I also won't bet on a 43 year old staying healthy all year. I thought you were going to bet against him making "solid improvement," but you seem to be conceding that.

But if you grant that DIPS only holds for the highest levels of the game, and that it holds precisely because everyone capable of pitching at those levels is very, very good at preventing hits on balls in play, then it seems at least plausible that a pitcher could suddenly lose his BABIP skill without losing his DIPS skills.

I agree that it's plausible; lots of things are. I'm more interested in the probable. This same argument (that he wasn't major league caliber) could be made about every established pitcher that has an uncommonly high BABIP one year. Either you think those arguments are generally probably wrong, or your DIPS theory has no practicality on the high side.

So the cutoff is a 114 ERA+?

That's 15% worse than his ZiPS projection.

I also won't bet on a 43 year old staying healthy all year.

Ya think? Me neither.

Which is my whole point. The argument that I'm reacting to is the one that assumes a projection for a 43-year-old coming off injury is just as reliable as the projection for anyone else.

I thought you were going to bet against him making "solid improvement,"

That's exactly what I'm doing. Workload is a crucial component of every pitcher's contribution, as I seem to recall you and I discussing. ;-) Any rate stat alone, ERA+ or anything else, is far from sufficient to assess a pitcher's performance. So what I'm asserting is that not only is the ERA (and by extension ERA+) rate stat in that projection overly optimistic, so is the workload. And what the hell, I'm throwing strikeouts in there too just to make it interesting.

So what would you say to something like this: Randy Johnson will fall short of that projection in all three categories by at least 15%; that is, he will pitch no more than 184 innings, strike out no more than 175 batters, and post an ERA no lower than 4.17. That's my proposition.

Steve, I'll take you up on that if you use K/9 - 15% instead of total K as the third leg. Or K/BB - 15%. Otherwise, total K is just a proxy for innings. That said, 184 innings of 4.17 ERA seems pretty close to a median projection if Johnson is ready to start the season. I've read he might not be, but that was a while ago.

That's exactly what I'm doing. Workload is a crucial component of every pitcher's contribution, as I seem to recall you and I discussing. ;-) Any rate stat alone, ERA+ or anything else, is far from sufficient to assess a pitcher's performance. So what I'm asserting is that not only is the ERA (and by extension ERA+) rate stat in that projection overly optimistic, so is the workload. And what the hell, I'm throwing strikeouts in there too just to make it interesting.

I was under the impression the discussion was about whether Johnson would regain effectiveness, with health being an underlying assumption in the discussion. I'm saying I would expect him to be substantially better (when actually pitching) than he was in 2006, which would mean his numbers w/RISP were largely a fluke. More to the point, I'd wager that if he's healthy enough to throw 150 innings, he should have an ERA+ at least 15 points higher than last year. His health is an entirely different question--one related to value, but not one I thought was related to the discussion here.

Also, why do you care what the average ERA in Arizona is if ERA+ is league and park adjusted?

I agree that it's plausible; lots of things are. I'm more interested in the probable. This same argument (that he wasn't major league caliber) could be made about every established pitcher that has an uncommonly high BABIP one year. Either you think those arguments are generally probably wrong, or your DIPS theory has no practicality on the high side.

I, and other Red Sox fans, have offered specific and distinct reasons why Schilling was so bad. In particular, he was coming off ankle surgery, and this injury caused him to lose command of his splitter, and he spent 2005 either getting killed on hanging splitters or throwing too many fastballs, as that was his only consistent pitch. Curt Schilling, by observation, was a qualitatively different pitcher in 2005 from 2004 or 2006. I don't think that Schilling was bad because he had a high BABIP, I think he was bad because he was a one-pitch pitcher who threw way too many hanging splits. These problems happened to show themselves more in BABIP than in other component stats, but he was bad regardless.

The "X pitcher that I saw suck umpteen times last year sucked he wasn't unlucky" syndrome can get played out for many teams. See Lopez, Rodrigo. See Vazquez, Javy. The only reason that there is so much more controversy here is because it's the Yankees and it's the Big Unit. People believe solid research and theory as long as it doesn't contradict what "I SAW WITH MY OWN EYES". I understand. My eyes told me that Javier Vazquez sucked with men on base. Every time he had a man on I knew it was just a matter of time until someone hit a rope somewhere. Thankfully, my mind tells me Vazquez and RJ were unlucky and will be better next year. It's interesting.

It really all comes back to the scouts vs. stats arguments except we're the scouts!!

People believe solid research and theory as long as it doesn't contradict what "I SAW WITH MY OWN EYES".

I never said I don't think that, in the aggregate, one-year BABIP for an established MLB pitcher is a weak predictor on next-year BABIP. This has been shown, and it's a useful finding. What I take issue with is, first, the false absolutizing of this finding, which has been shown to be incorrect by all the best research on the topic. Second, the larger problem is the false universalizing of this finding, to argue that a particular outlier must actually have a league-average BABIP skill in a certain year, becuase this trend has been discovered in a certain aggregate. That's the ecological fallacy.

I dispute that I am behaving as irrationally as you say. I'd argue, in fact, that it's clearly irrational to ignore observational data and apply an aggregate trend universally to every individual. Obviously, we have to give credence to aggregate trends, and only question the relevance of the application in certain situations, when we have good reason - I've offered those on this thread, and I still find them convincing.

I, and other Red Sox fans, have offered specific and distinct reasons why Schilling was so bad. In particular, he was coming off ankle surgery, and this injury caused him to lose command of his splitter, and he spent 2005 either getting killed on hanging splitters or throwing too many fastballs, as that was his only consistent pitch. Curt Schilling, by observation, was a qualitatively different pitcher in 2005 from 2004 or 2006. I don't think that Schilling was bad because he had a high BABIP, I think he was bad because he was a one-pitch pitcher who threw way too many hanging splits. These problems happened to show themselves more in BABIP than in other component stats, but he was bad regardless.

If Schilling had a normal BABIP (luckily, as you would say) and an ERA under 4, do you think many Red Sox fans would have thought he was a bad pitcher?