Do you have adblock enabled?
If you can read this, either the style sheet didn't load or you have an older browser that doesn't support style sheets. Try clearing your browser cache and refreshing the page.

(Oddball Sports)   Mets' R.A. Dickey could become the first Triple Crown of Pitching winner to never throw a pitch above 80 mph   (oddballsportsblog.com) divider line 91
    More: Cool, R.A. Dickey, Dickey, Mets, Rare Ltd., knuckleball pitch, Pedro Martinez, Tim Wakefield, Justin Verlander  
•       •       •

1354 clicks; posted to Sports » on 21 Jun 2012 at 5:40 AM (2 years ago)   |  Favorite    |   share:  Share on Twitter share via Email Share on Facebook   more»



91 Comments   (+0 »)
   
View Voting Results: Smartest and Funniest

Archived thread

First | « | 1 | 2 | » | Last | Show all
 
2012-06-21 03:04:38 PM  

Dafatone: Where are you getting these numbers from, fangraphs? Can't find swing and miss percentages by pitch type (maybe I'm not looking hard enough) but I thought his swing and miss knuckleball percentage was way above that. Perhaps that was percent of swings that miss. Do you have evidence that these stats aren't particularly good or anything like that.


Brooks Baseball has swing & miss by pitch type, but their numbers differ from his.
 
2012-06-21 03:09:58 PM  

Super Chronic: But that's not such a terrible thing, because good knuckleballers largely owe their success to the rarity of it. If the world were awash in knuckleballers and batters experienced them more frequently, they'd all become more hittable.


Again, no. The knuckleball is the only pitch you can never adjust to. Knuckleball pitchers are not "figured out"; they implode. Every other pitch in baseball is deliberately aimed. The knuckleball is chaotic -- the pitcher's technique is all about limiting the randomness to a part of the strike zone, but within that area no one knows where it's exactly going to go.

DeWayne Mann: There's a few theories on why the knuckleball screws up BABIP. One of the most fun ones I've heard is that the batter may (consciously or not) pick up on where the catcher sets up on a standard pitch. But if the catcher has no clue where the pitch is going, as with the knuckler, that's an advantage the batter no longer gets.


It may have nothing to do with the catcher. A number of hitters are known to try to identify pitches by the ball's spin. You can certainly identify a knuckler that way, but in this case it doesn't tell you much. As mentioned above, in every other pitch in baseball, the pitcher makes a very deliberate effort to dictate where and how and when the pitch will move, and that's done by inducing a very deliberate, rapid spin. That is how most pitches work. So a veteran may have never even seen a particular pitcher before, but if he reads a scouting report and sees a certain spin he can think "oh this is a slider" and try to hit it. The key word is "try"; the point in throwing hard is to limit the time to look & react (at some point above 90mph it is in fact physically impossible, so scouts don't love velocity for nothing). In other cases they anticipate a pitch, especially if the ball is too fast or the pitcher is prone to patterns. You can anticipate ANY pitch, even Mariano Rivera's cutter. Hitting it is something else, of course, but people do know pretty well HOW the pitch will move. The knuckleball is the exception. It will drop, of course, as there's no backspin, but how it moves AS it drops is literally up to God.
 
2012-06-21 03:13:55 PM  

DeWayne Mann: Dafatone: Where are you getting these numbers from, fangraphs? Can't find swing and miss percentages by pitch type (maybe I'm not looking hard enough) but I thought his swing and miss knuckleball percentage was way above that. Perhaps that was percent of swings that miss. Do you have evidence that these stats aren't particularly good or anything like that.

Brooks Baseball has swing & miss by pitch type, but their numbers differ from his.


Actually, they appear pretty similar for 2012. What xpisblack left out of his breakdown, for whatever reason, is that 19% of Dickey's knuckleballs get fouled off. Only 18.25% are even put into play, with over half of those being ground balls.

So, (approximately) half of Dickey's knuckleballs get swung at, with a roughly even chance of being fouled off, missed, or put in play. Of the ones put in play, just over half are ground balls. That's pretty good.
 
2012-06-21 03:15:37 PM  

DeWayne Mann: You're the jerk... jerk: We have 500+ innings of him as a knuckleballer, so it isn't like we have nothing. I would love to see it continue, but I would bet against it.

That's not a lot of data. Let's put it this way: we have 300 IP that says Hellickson is a .237 BABIP pitcher. Feel comfortable about that?

Or, if 300 IP is too few (but 500 is ok), the last 500 IP of Jonathan Sanchez's career have had a .268 BABIP, yet his career is .288.


Completely different argument. I am saying that we have 400 IP of a pitcher with a BABIP only slightly less than league average, this year we have 100 IP of him well below league average. I am more likely to trust that last 400IP than the recent 100IP. Not saying it is definitive, but that is where I would place an even money bet.

Would you place an even money bet that he will not have a .280ish BABIP?
 
2012-06-21 03:54:07 PM  

Dafatone: Actually, they appear pretty similar for 2012. What xpisblack left out of his breakdown, for whatever reason, is that 19% of Dickey's knuckleballs get fouled off. Only 18.25% are even put into play, with over half of those being ground balls.

So, (approximately) half of Dickey's knuckleballs get swung at, with a roughly even chance of being fouled off, missed, or put in play. Of the ones put in play, just over half are ground balls. That's pretty good.


And this is how I learned that, even though I swear I selected 2012, for some reason I was still looking at the career data. Problem solved now.

You're the jerk... jerk: Completely different argument. I am saying that we have 400 IP of a pitcher with a BABIP only slightly less than league average, this year we have 100 IP of him well below league average. I am more likely to trust that last 400IP than the recent 100IP. Not saying it is definitive, but that is where I would place an even money bet.

Would you place an even money bet that he will not have a .280ish BABIP?


Greg Maddux had a .302 BABIP for the first 435.2 IP of his career (when the average BABIP was like, .280). We now consider Maddux as one of the guys who might've had some control over his BABIP, and his career mark is .280.

400 IP is not anywhere near enough to determine a pitcher's BABIP talent level.

Is it likely that anyone would have a BABIP down near .240? Of course not, and it doesn't help that the Mets have a pretty bad defense. But we have no idea what Dickey's BABIP should be, even with those 400 IP.
 
2012-06-21 04:05:13 PM  

DeWayne Mann: You're the jerk... jerk: Completely different argument. I am saying that we have 400 IP of a pitcher with a BABIP only slightly less than league average, this year we have 100 IP of him well below league average. I am more likely to trust that last 400IP than the recent 100IP. Not saying it is definitive, but that is where I would place an even money bet.

Would you place an even money bet that he will not have a .280ish BABIP?

Greg Maddux had a .302 BABIP for the first 435.2 IP of his career (when the average BABIP was like, .280). We now consider Maddux as one of the guys who might've had some control over his BABIP, and his career mark is .280.

400 IP is not anywhere near enough to determine a pitcher's BABIP talent level.

Is it likely that anyone would have a BABIP down near .240? Of course not, and it doesn't help that the Mets have a pretty bad defense. But we have no idea what Dickey's BABIP should be, even with those 400 IP.


I have no idea if you are picking out outliers or not. But assuming it is a random sample, 10% of Greg's career is going to predict his BABIP pretty well. It turns out that it doesn't. But it should do a pretty good job of it.

I don't know what the number is needed to predict true BABIP for a pitcher (we can figure this out if needed), but given that he was at about .280 before and we can expect about .280 then I would put even money that he is above .270 for the rest of the year.
 
2012-06-21 04:10:45 PM  

You're the jerk... jerk: I don't know what the number is needed to predict true BABIP for a pitcher (we can figure this out if needed), but given that he was at about .280 before and we can expect about .280 then I would put even money that he is above .270 for the rest of the year


Ordinarily I'd agree with you, but in this particular case the change in how he pitches has been pretty remarkable. He seems (to me, at least) to have added dimensions to his knuckleball that just weren't there before, even fairly recently. So while I don't think he'll be keeping *this* up (as much as I hope he does), I also think that there isn't a player in baseball whose past stats mean less.
 
2012-06-21 04:15:11 PM  

dragonchild: Super Chronic: But that's not such a terrible thing, because good knuckleballers largely owe their success to the rarity of it. If the world were awash in knuckleballers and batters experienced them more frequently, they'd all become more hittable.

Again, no. The knuckleball is the only pitch you can never adjust to. Knuckleball pitchers are not "figured out"; they implode. Every other pitch in baseball is deliberately aimed. The knuckleball is chaotic -- the pitcher's technique is all about limiting the randomness to a part of the strike zone, but within that area no one knows where it's exactly going to go.


I disagree. It's not that you can never adjust to them under any circumstances. While it is generally correct to say no two knuckleballs are alike, batters nevertheless struggle to adjust to them largely because they practice with a steady diet of more predictable pitches that are "deliberately aimed." They learn things about timing and pitch recognition, and then develop habits that go completely out the window when they face the knuckleballers. If the randomness of a knuckleball were the norm, then hitters would take completely different approaches at the plate and learn to adapt to knuckleballs in spring training.
 
2012-06-21 04:22:03 PM  

FreakinB: You're the jerk... jerk: I don't know what the number is needed to predict true BABIP for a pitcher (we can figure this out if needed), but given that he was at about .280 before and we can expect about .280 then I would put even money that he is above .270 for the rest of the year

Ordinarily I'd agree with you, but in this particular case the change in how he pitches has been pretty remarkable. He seems (to me, at least) to have added dimensions to his knuckleball that just weren't there before, even fairly recently. So while I don't think he'll be keeping *this* up (as much as I hope he does), I also think that there isn't a player in baseball whose past stats mean less.


I want to believe, I really do. But if we include RA Dickey the list of players who became superstars at 37 or older would be as follows:
RA Dickey
Now if we expand this to include all sports it would be:
RA Dickey
 
2012-06-21 04:30:35 PM  

You're the jerk... jerk: I have no idea if you are picking out outliers or not. But assuming it is a random sample, 10% of Greg's career is going to predict his BABIP pretty well. It turns out that it doesn't. But it should do a pretty good job of it.

I don't know what the number is needed to predict true BABIP for a pitcher (we can figure this out if needed), but given that he was at about .280 before and we can expect about .280 then I would put even money that he is above .270 for the rest of the year.


I picked Maddux because I knew his BABIP at the beginning of his career was WAY above average. One of those fun facts that are shoved in my brain.

But otherwise...why would you think 10% of a career predicts BABIP pretty well? I'm not even sure 50% of a career predicts the other half well. Tom Tango recommends looking at at least 6 full seasons worth of data for a pitcher. Baseball Prospectus says 8. And even then, we're talking "yeah, we have an idea of what's going on" not "we know exactly what's going on."
 
2012-06-21 04:37:51 PM  

Super Chronic: If the randomness of a knuckleball were the norm, then hitters would take completely different approaches at the plate and learn to adapt to knuckleballs in spring training.


You can't adapt to random. That's the whole point. It's not like there's no information on the knuckleball, but what's out there is very limited and already common knowledge. There is no additional expertise to be gained by watching a hundred, a thousand or even a million knuckleballs.

I roll a couple of dice (like in craps). You try to predict what'll come up. Yes, you can do better than just blindly guess numbers, but the first and ONLY lesson you can learn is the most likely outcome (7). That's better than nothing, but that's it. You can play the odds and that's all you can do. There is no pattern to adjust to in randomness, no tendencies, no tells, no hints. Unless I'm actually cheating, you can't watch how I throw the dice or look in my eyes or watch the dice really closely or do anything else to improve your raw mathematical chances of guessing what the probable outcome will be. That's how randomness works, and that is the defining characteristic of the knuckleball -- "adjusting" to it is a very short learning curve.

This isn't up for disagreement; physicists have studied the pitch and know how it works. The pitch would not be made any less effective if everyone threw it; you're demonstrating a textbook case of confirmation bias.
 
2012-06-21 04:46:03 PM  

You're the jerk... jerk: I want to believe, I really do. But if we include RA Dickey the list of players who became superstars at 37 or older would be as follows:
RA Dickey
Now if we expand this to include all sports it would be:
RA Dickey


Well, obviously, this depends on the definition of "superstar." And 37 is tough just because it's rare for anyone to play that late anyway. But I can give you all sorts of guys who were far better in their mid to late 30s than the rest of their career. Dennis Martinez might be the best example: his first ASG was at 35, and then he had 3 more.

And you missed a really, really obvious one: Satchel Paige.
 
2012-06-21 04:46:19 PM  

You're the jerk... jerk: But if we include RA Dickey the list of players who became superstars at 37 or older would be as follows:
RA Dickey


Jamie Moyer's best three (consecutive) seasons were when he was 38-40. He had some good seasons prior to that, but was generally known as a decent junkballer playing in a pitcher's park. His career really took off during the '01 season.
 
2012-06-21 05:00:11 PM  

dragonchild: Super Chronic: If the randomness of a knuckleball were the norm, then hitters would take completely different approaches at the plate and learn to adapt to knuckleballs in spring training.

You can't adapt to random. That's the whole point. It's not like there's no information on the knuckleball, but what's out there is very limited and already common knowledge. There is no additional expertise to be gained by watching a hundred, a thousand or even a million knuckleballs.

I roll a couple of dice (like in craps). You try to predict what'll come up. Yes, you can do better than just blindly guess numbers, but the first and ONLY lesson you can learn is the most likely outcome (7). That's better than nothing, but that's it. You can play the odds and that's all you can do. There is no pattern to adjust to in randomness, no tendencies, no tells, no hints. Unless I'm actually cheating, you can't watch how I throw the dice or look in my eyes or watch the dice really closely or do anything else to improve your raw mathematical chances of guessing what the probable outcome will be. That's how randomness works, and that is the defining characteristic of the knuckleball -- "adjusting" to it is a very short learning curve.

This isn't up for disagreement; physicists have studied the pitch and know how it works. The pitch would not be made any less effective if everyone threw it; you're demonstrating a textbook case of confirmation bias.


Okay, first of all, cut the crap with the "confirmation bias" thing (a common tactic in Fark threads, accusing people who argue against you as having a cognitive defect). Especially since you've completely misused the term, as I haven't even gotten into data (the term, FYI, refers to seeking out data in support of one's hypothesis). If you want to do it right, accuse me of having the Dunning-Kruger effect, that's the one that really gets 'em.

Now that we're done with that: I'm quite aware that physicists have studied it and know how it works, but you're taking meaning from it that isn't there. The movement of a knuckleball is not completely random; it would be random if it were not aimed at all, or aimed at an unreasonably large target (say, a barn). If throwing a knuckleball were like rolling dice then you or I could learn how to do it and immediately become R.A. Dickey or Tim Wakefield. And no batter would be better than any other batter at hitting knuckleballs: all variance would be noise. Neither of those things are true.
 
2012-06-21 05:01:27 PM  

DeWayne Mann: But otherwise...why would you think 10% of a career predicts BABIP pretty well? I'm not even sure 50% of a career predicts the other half well. Tom Tango recommends looking at at least 6 full seasons worth of data for a pitcher. Baseball Prospectus says 8. And even then, we're talking "yeah, we have an idea of what's going on" not "we know exactly what's going on."


This is why I sometimes don't like advanced metrics, or at least how they're incorrectly applied. BABIP's a good one. It's a great tool for saying "If this guy throws the way he has for the past five years, here's where he'll most often come out the next year." But players actually change, learning new pitches, taking different approaches, etc. So they become different players, effectively, making past performance much less relevant.
 
2012-06-21 05:03:35 PM  

dragonchild: You can't adapt to random. That's the whole point. It's not like there's no information on the knuckleball, but what's out there is very limited and already common knowledge. There is no additional expertise to be gained by watching a hundred, a thousand or even a million knuckleballs.


Yes you can. Or at least practice against it.

Set up a knuckleball-throwing pitching machine, or get Dickey to throw 100 pitches to some major league guy. You think he won't be taking better swings on pitch 100 than pitch 1?
 
2012-06-21 05:14:03 PM  

Dafatone: DeWayne Mann: But otherwise...why would you think 10% of a career predicts BABIP pretty well? I'm not even sure 50% of a career predicts the other half well. Tom Tango recommends looking at at least 6 full seasons worth of data for a pitcher. Baseball Prospectus says 8. And even then, we're talking "yeah, we have an idea of what's going on" not "we know exactly what's going on."

This is why I sometimes don't like advanced metrics, or at least how they're incorrectly applied. BABIP's a good one. It's a great tool for saying "If this guy throws the way he has for the past five years, here's where he'll most often come out the next year." But players actually change, learning new pitches, taking different approaches, etc. So they become different players, effectively, making past performance much less relevant.


It's still better than the old method of "If this guy throws the way he has for the past five years, he'll most likely...I dunno. But hey, at least he LOOKS like a winner!"

Obviously, there are a ton of variables in play. But the more of them we can trim away, the better.
 
2012-06-21 05:58:48 PM  

DeWayne Mann: You're the jerk... jerk: I have no idea if you are picking out outliers or not. But assuming it is a random sample, 10% of Greg's career is going to predict his BABIP pretty well. It turns out that it doesn't. But it should do a pretty good job of it.

I don't know what the number is needed to predict true BABIP for a pitcher (we can figure this out if needed), but given that he was at about .280 before and we can expect about .280 then I would put even money that he is above .270 for the rest of the year.

I picked Maddux because I knew his BABIP at the beginning of his career was WAY above average. One of those fun facts that are shoved in my brain.

But otherwise...why would you think 10% of a career predicts BABIP pretty well? I'm not even sure 50% of a career predicts the other half well. Tom Tango recommends looking at at least 6 full seasons worth of data for a pitcher. Baseball Prospectus says 8. And even then, we're talking "yeah, we have an idea of what's going on" not "we know exactly what's going on."


The confidence interval at 95% for 10% of a population is less than +/-5%. Pretty reasonable to use that for results (you cannot treat .280 BABIP as 28% since the real range is like .240 to .350).

Paige didn't become a star at 41 because he got better at 41 but because the world became slightly less racist.
 
2012-06-21 06:12:38 PM  

You're the jerk... jerk: The confidence interval at 95% for 10% of a population is less than +/-5%. Pretty reasonable to use that for results (you cannot treat .280 BABIP as 28% since the real range is like .240 to .350).


You appear to be making the following assumption:

"Pitcher A has 4000 IP and a BABIP of .280. Therefore, his BABIP talent level is .280"

But that's not true in the slightest. It's more like "his BABIP talent level is between .270 and .290."

There's way, way too much that goes into BABIP other than who the pitcher is.

You're the jerk... jerk: Paige didn't become a star at 41 because he got better at 41 but because the world became slightly less racist.


And Dickey completely changed his pitching style. It's like sometimes things change.
 
2012-06-21 06:31:59 PM  
Starting pitcher for the NL in the All Star game?

would love to see him against the best the AL has to offer


Let's Go Mets
 
2012-06-21 07:03:56 PM  

Dafatone: The knuckleball sets up the fastball, which is effective only because hitters are looking for the knuckleball. It's not like Dickey would have been better off throwing more fastballs last year.


Meh. Some knuckleballers use it to set up their fastballs, but some do the reverse-- they use the fastball to set up the knuckler. Some of them rely on the knuckleball as their go-to pitch, and throw fastballs to keep batters on their toes and remind them to look for other things or when the wind is acting up. Some only use the fastball against batters who hit knuckleballs relatively well or when weather conditions interfere. The data on knuckleball pitchers who use the knuckleball to set up the knuckleball (as Dickey appears to be doing this year) are pretty sparse.

In his case, not the abstract, the 174 fastballs Dickey's thrown this year, have included 73 called strikes (2 for the third strike), 40 balls, 23 fouls (5 in useless 2-strike situations), 9 swing-and-misses (3 for the third strike), and the remaining 26 have been in play. This is clearly an effective use of his two types of fastball. Even in full counts, when most pitchers value control over deception, his knuckleball-fastball proportion is 38-8. So yes, his knuckler is more reliable than some other pitchers', but it requires a lot of help from batters (see below), and his fastball has actually been his more effective pitch, in an admittrely much smaller sample (and almost certainly because his knuckelball has worked so well so thus far).

Last year, he threw 76.4% knuckleballs; in 2010, he threw 83.4% knuckleballs, much more like his 2012 numbers. In 2011, he relied more on his fastball than in either of the other Mets years, making it a bit aberrant. He's been primarily a knuckleball pitcher since 2010, and his best knuckleballs, paradoxically, are the ones with the least movement. That's the advantage he's found in throwing them harder: he gets less movement, so they look more like fastballs and stay closer to the zone, but they still dance. That's rare; time will tell how well it works in the long run, but so far it's still possible that the novelty may be as or more responsible for his success as the skill of the pitching. I hope not-- it's always fun to watch a good knucklballer.

This year, he's thrown 364 knuckleballs in 2-strike counts, compared to 439 with 0 strikes (309 with a completely fresh count, suggesting he reasonably considers it an odds-on strike pitch); he's relying on the knuckleball as his out pitch, and relying on his apparent control of its flutter far more than his other pitches, mainly relying on his fastball-- not just using it, but relying on it-- against (I've been told, but I'm by no means certain) batters who have historically hit knuckleballs well. Or that's how it looks; again, I haven't watched him too closely, so I could very easily be wrong here.

The breakdown numbers, by the way, are directly from pfx, not fangraphs, and my point in citing them was that it's not as though his knuckleball is a gimme strikeout pitch. Of his 364 2-strike knuckleballs pitched this year (of, by the way, 1178 total knuckleballs thrown):
93 have been called balls
17 have been called strikes
254* have elicited swings. Of those 254:
96 have been fouls
79 have been whiffed
79 have been in play (42 ground balls)
*(pfx actually lists it as 253, but that's probably a mistake given the other tallies)

Based on these results, the 2-strike value (which I'm using as a way to evaluate the pitch he considers his fallback out-getter-- of course, his primary out-getter is also the knuckleball, just because of its ubiquity) of his knuckleball is to prompt attempted contact: far more batter-exclusive outcomes are called balls than strikes. He needs batters to swing at his knuckler with 2 strikes; of non-foul (which, with 2 strikes, are more harmful to the pitcher than to the batter), swung-on outcomes, exactly half have been in play and half have resulted in strikeouts. So that's his go-to pitch, but it requires quite a lot of participation from the batters. In 2-strike situations, by the way, he's thrown 23 fastballs (13 4-seamers and 10 sinkers) for 6 balls, 2 called strikes (each from a sinker), 3 whiffs (each from a sinker), 5 fouls, and 7 balls in play, including one HR. The one 2-strike changeup he threw this year was called ball 2.

In other words, as long as he can keep making batters think he's not throwing a knuckleball with 2 strikes, he might be able to maintain something like this success. But he has set a clear pattern now, in that he has thrown a knucklleball with 2 strikes in 364 out of 388 cases, so if batters should start sitting on the 2-strike fastball and lay off anything that moves in any way other than a simple drop, he could become a less effective pitcher. And that's the sort of adjustment that teams can make around the ASB-- granted, of course, it's possible that all of those swings would have otherwise been called strikes, but that seems unlikely.

Dafatone: A whole bunch of advanced metrics and then we're going with wins?


First, I mistakenly thought I had covered that with my dig about the Mets' collapsing and its toll on his win stat, which is part of the triple crown. Wins have to be mentioned in a discussion of his chance to win the triple crown (as mentioned in the article) because they are part of the achievement in question. I'm not using wins to evaluate the pitcher, only to assess his chances of winning the triple crown.

And the assessment of Dickey as the best pitcher on the Mets after his callup in May 2010 came directly from this article from 27 August 2010. I used it because it shows contemporary opinion, not retrospect. The point, again, was to demonstrate how in-season evaluations can begin to look different in hindsight.

I should have been clearer about that.
 
2012-06-21 09:18:43 PM  

DeWayne Mann: You're the jerk... jerk: The confidence interval at 95% for 10% of a population is less than +/-5%. Pretty reasonable to use that for results (you cannot treat .280 BABIP as 28% since the real range is like .240 to .350).

You appear to be making the following assumption:

"Pitcher A has 4000 IP and a BABIP of .280. Therefore, his BABIP talent level is .280"

But that's not true in the slightest. It's more like "his BABIP talent level is between .270 and .290."

There's way, way too much that goes into BABIP other than who the pitcher is.

You're the jerk... jerk: Paige didn't become a star at 41 because he got better at 41 but because the world became slightly less racist.

And Dickey completely changed his pitching style. It's like sometimes things change.


I don't disagree with your first statement at all. But I think that a random sampling will give us a good overview of where we would land. There is a lot of noise in the data, but that can be dealt with. (Note that 500 innings tends to be a pretty good predictor of future ERA We can run this for BAPIB, I would be shocked if we found something different.


If this were a non-knuckleballer we wouldn't be having this debate as the evidence points to very little skill difference for ML pitchers with regards to BABIP, we would just assume that he would gravitate to the norm. What you are advocating is that while we have data, albeit insufficient to tell the whole story, we should ignore it for the current aberrations. My original post pointed out he has improved in k/9 and bb/9 so I expect him to be better. Just not this much better, his BABIP will almost certainly regress.
 
2012-06-21 09:57:10 PM  

You're the jerk... jerk: I don't disagree with your first statement at all. But I think that a random sampling will give us a good overview of where we would land. There is a lot of noise in the data, but that can be dealt with. (Note that 500 innings tends to be a pretty good predictor of future ERA We can run this for BAPIB, I would be shocked if we found something different.


No, 500 IP is not a pretty good predictor of ERA. According to that chart & piggybacking off the comments (since they worded it better than I could've), if someone's ERA is 4.00 over 500 IP, then there's a 66% chance his future ERA will be between 3 & 5, and a 95% chance it'll be between 2 & 6. That's pretty much the standard distribution of ERAs anyway. I mean, here's an ERA predictor I just made up:

def NameERA(String name):
return numberOfWords(name) + 2

I now do a "pretty good" job of estimating any pitcher's ERA with 0 IP required! Also, NO ONE SIGN RYAN ROWLAND-SMITH.

You're the jerk... jerk: If this were a non-knuckleballer we wouldn't be having this debate as the evidence points to very little skill difference for ML pitchers with regards to BABIP, we would just assume that he would gravitate to the norm. What you are advocating is that while we have data, albeit insufficient to tell the whole story, we should ignore it for the current aberrations. My original post pointed out he has improved in k/9 and bb/9 so I expect him to be better. Just not this much better, his BABIP will almost certainly regress.


My point is that we have NO IDEA what the "norm" is for him (and those previous 400 IP tell us very, very little about it). Will he regress? Probably. It's generally safe to assume any player will regress no matter what the question is. But you're trying to make a case using insufficient data.
 
2012-06-21 10:51:02 PM  

Super Chronic: Okay, first of all, cut the crap with the "confirmation bias" thing (a common tactic in Fark threads, accusing people who argue against you as having a cognitive defect).


Well now, as long as you're going to bring up other people and some personal baggage. . . You think the more people tell you something, the more likely it is they're wrong? If people are constantly calling you out as having a cognitive defect, did you ever consider the possibility that maybe they're on to something? Or can the whole world call you an idiot and you'd tell yourself to your dying day that they're all crazy and you're right?

Mind you, I don't believe any of that; truth is not democratic. But are you sure this is where you want to take this? Or can we go back to talking about baseball?

Super Chronic: The movement of a knuckleball is not completely random; it would be random if it were not aimed at all, or aimed at an unreasonably large target (say, a barn).


By this same flawed logic, it'd be possible to get dice rolls totalling -4.78 or 99,475 in craps because it's "random". Random does not mean unconfined; you have to be batshiat insane to guess a craps roll will NOT be between 2 and 12. Same thing as locating a knuckleball; properly thrown it will more or less wind up somewhere over the plate or maybe even a portion thereof. Me calling it "random" doesn't mean it'll suddenly defy the laws of physics, swoop upwards out of the park, jump through a circus hoop, cook your dinner and then lodge itself in Richard Gere's asshole. The trajectory is confined but unpredictable, and unfortunately for the batter, hitting a round ball with a cylindrical bat requires precision.

Dafatone: Set up a knuckleball-throwing pitching machine, or get Dickey to throw 100 pitches to some major league guy. You think he won't be taking better swings on pitch 100 than pitch 1?


He can certainly adjust to the pitcher -- how often he throws strikes, if he tries to pitch inside or outside on a given count, whether and by how much he changes speeds. Pitchers try to be as unpredictable as possible so even Wakefield had a few tricks up his sleeve. But while they can introduce variation way beyond single pitch, single speed, single location, they are human. So within that context it's perfectly reasonable for a batter to have SOME learning curve -- at least against Dickey.

My point, which is quite simple, is that beyond a certain point it's impossible to adapt to the pitch because of how it works. I really don't understand why this has people so riled up. I mean yeesh, Wakefield was pretty well figured out; the dude threw the soft knuckleball almost exclusively for years in the same park and almost to his last day in uniform the key factor in whether or not you could hit him was whether his flutterball was dancing in the zone or not. As in, it wasn't up to the hitter. It's not like he went out of his way to vary the speed; if batters could adjust you'd think they'd be smashing that pitch the 4th time through the lineup. They only smashed it when it stopped dancing.
 
2012-06-21 10:53:29 PM  

DeWayne Mann: My point is that we have NO IDEA what the "norm" is for him (and those previous 400 IP tell us very, very little about it). Will he regress? Probably. It's generally safe to assume any player will regress no matter what the question is.


Ah, but what's truly exciting is because we have no idea, there's a chance (however small) he won't regress. Methinks part of this excitement stems from the fact that even skeptics know, on some gut level, that there's a small chance we're witnessing something historical. And that's precisely because we don't have a perfect set of data to compare Dickey's run to. Only time will tell, of course, but it'd be worth it to sear these games into memory for that chance.
 
2012-06-21 10:58:07 PM  

DeWayne Mann: You're the jerk... jerk: I don't disagree with your first statement at all. But I think that a random sampling will give us a good overview of where we would land. There is a lot of noise in the data, but that can be dealt with. (Note that 500 innings tends to be a pretty good predictor of future ERA We can run this for BAPIB, I would be shocked if we found something different.

No, 500 IP is not a pretty good predictor of ERA. According to that chart & piggybacking off the comments (since they worded it better than I could've), if someone's ERA is 4.00 over 500 IP, then there's a 66% chance his future ERA will be between 3 & 5, and a 95% chance it'll be between 2 & 6. That's pretty much the standard distribution of ERAs anyway. I mean, here's an ERA predictor I just made up:

def NameERA(String name):
return numberOfWords(name) + 2

I now do a "pretty good" job of estimating any pitcher's ERA with 0 IP required! Also, NO ONE SIGN RYAN ROWLAND-SMITH.

You're the jerk... jerk: If this were a non-knuckleballer we wouldn't be having this debate as the evidence points to very little skill difference for ML pitchers with regards to BABIP, we would just assume that he would gravitate to the norm. What you are advocating is that while we have data, albeit insufficient to tell the whole story, we should ignore it for the current aberrations. My original post pointed out he has improved in k/9 and bb/9 so I expect him to be better. Just not this much better, his BABIP will almost certainly regress.

My point is that we have NO IDEA what the "norm" is for him (and those previous 400 IP tell us very, very little about it). Will he regress? Probably. It's generally safe to assume any player will regress no matter what the question is. But you're trying to make a case using insufficient data.


One of us will have to run the analysis checking the BABIP for first 400 IP vs the career for the R squared. There is so little variation in BABIP and it is generally distributed normally so my guess is that is solidly positive. Obviously there are going to be instances when this is wrong.
 
2012-06-21 11:04:18 PM  

DeWayne Mann: My point is that we have NO IDEA what the "norm" is for him (and those previous 400 IP tell us very, very little about it). Will he regress? Probably. It's generally safe to assume any player will regress no matter what the question is. But you're trying to make a case using insufficient data.


Here's a really, really stupid analogy I just came up with.

Pretend I know nothing at all about you. But you randomly tell me that you're going to run in a marathon, and you have a goal of 2 hours, 30 minutes.

I might suspect that you will not reach this goal. Because that's a tough goal, and most people can't do it. And as far as I know, you're "most people."

Then you tell me you've run 20 marathons before. So I check, and I learn that your last two marathons had times of 2:45, but I can't find out what any of your other times are. I still might not think that you'll reach your goal, but it's certainly more likely.

So, on the day of the race, you run the first 13 miles in an hour.

How likely is it that you'll reach your goal, and what of the above data should you use to determine that?

I hope the parallels are obvious, but just in case:

"You" are Dickey. "People" are pitchers, and "most people" are normal pitchers. "People who have run marathons" are knuckleballers. The time is their BABIP: we'd expect "most people" to have higher times than "people who have run marathons" just as we'd expect "normal pitchers" to have higher BABIPs than knuckleballers.

The goal is a ~.250 BABIP for the season. The previous two "marathons" resulted in ~.270 BABIPs. The other 18 "marathons" are the fact that, from before, we have "10%" of the data. And so far, he's slightly ahead of the pace for the goal.

Like I said, it's dumb. But does it at least show why I'm skeptical of the "400 IP!" claim?
 
2012-06-21 11:17:33 PM  
And while I was typing up that stupidly long post for no reason, 2 more posts appear!

dragonchild: DeWayne Mann: My point is that we have NO IDEA what the "norm" is for him (and those previous 400 IP tell us very, very little about it). Will he regress? Probably. It's generally safe to assume any player will regress no matter what the question is.

Ah, but what's truly exciting is because we have no idea, there's a chance (however small) he won't regress. Methinks part of this excitement stems from the fact that even skeptics know, on some gut level, that there's a small chance we're witnessing something historical. And that's precisely because we don't have a perfect set of data to compare Dickey's run to. Only time will tell, of course, but it'd be worth it to sear these games into memory for that chance.


Yeah, this is essentially what I'm saying. We can say Hellickson's gonna regress because we can compare him to a couple of dozen thousand other pitchers. We can compare Dickey to...um.

Well...

...

Yeah.

You're the jerk... jerk: There is so little variation in BABIP


WHAT? Seriously, either you phrased this wrong, I'm reading it wrong (which are sorta two sides of the same coin), or you're just making stuff up as you go along.

There's little variation in league BABIPs. But among individual players, the variation is MASSIVE. That's the entire reason anyone uses BABIP as a stat.

I mean, again, here are 5 of the best pitchers ever:

i1154.photobucket.com

Little variation? How many times can you find three consecutive points that average together to be the guy's overall BABIP? There are few, sure. But there are also a lot that are no where close.
 
2012-06-21 11:24:15 PM  
It just seems to me like saying "I want to learn the average temperature outside my house. So I'm going to take the temperature for the first 36 days of the year, and average that together. It's 10% of the year, but, meh, that's enough of a sample."
 
2012-06-21 11:49:00 PM  
Doing some research:

From Tom Tango:

It's still our best guess that if a pitcher has a (1b+2b+3b)/BIP rate of .320 and the league is .300, then a pitcher's "true" talent, based on the BIP, is .305 (80% regression towards the mean, or 1-r). This applies to pitchers with 500 to 1200 PAs.

In other words, with 1200 PA (somewhere around 300 IP), his "best guess" is that the pitcher's BABIP talent is probably not close to his past BABIP, and you're really better off just using the league average.

But, and I feel I have to point this out yet again: All of the these BABIP studies have little or nothing to do with knuckleballers, who have generaly been observed to have below league average BABIP-talents. There's very, very little data on knuckleballers, though. Essentially, trying to project them is a fool's errand.
 
2012-06-22 12:05:22 AM  
Without stats to back me up, just 40 years of watching baseball I would guess this guy will regress back to his norm in the second half of the season. I recall just a couple years ago a COL pitcher who was 15-1 at the All Star break and we had to endure the "can he get to 30 wins" articles. I clearly remember telling someone around that time that he would be lucky to get to 20 wins and he ended the season 19-8. Why? Because players numbers don't change that drastically. At least not without help (Brady Anderson). He will regress to his averages by the end of the season or this year will be the anomaly.

I will say this right now...he will not get to 20 wins and his ERA will be over 3 by the end of the season.
 
2012-06-22 12:08:45 AM  

bluenote13: Without stats to back me up, just 40 years of watching baseball I would guess this guy will regress back to his norm in the second half of the season. I recall just a couple years ago a COL pitcher who was 15-1 at the All Star break and we had to endure the "can he get to 30 wins" articles. I clearly remember telling someone around that time that he would be lucky to get to 20 wins and he ended the season 19-8. Why? Because players numbers don't change that drastically. At least not without help (Brady Anderson). He will regress to his averages by the end of the season or this year will be the anomaly.

I will say this right now...he will not get to 20 wins and his ERA will be over 3 by the end of the season.


This post gave me so many strokes.
 
2012-06-22 12:40:04 AM  

dragonchild: Super Chronic: Okay, first of all, cut the crap with the "confirmation bias" thing (a common tactic in Fark threads, accusing people who argue against you as having a cognitive defect).

Well now, as long as you're going to bring up other people and some personal baggage. . . You think the more people tell you something, the more likely it is they're wrong? If people are constantly calling you out as having a cognitive defect, did you ever consider the possibility that maybe they're on to something? Or can the whole world call you an idiot and you'd tell yourself to your dying day that they're all crazy and you're right?


No, but I do believe that someone who tosses out the term "confirmation bias" in a context where no confirmation bias has been suggested is pretty damn stupid.
 
2012-06-22 12:51:19 AM  

dragonchild: By this same flawed logic, it'd be possible to get dice rolls totalling -4.78 or 99,475 in craps because it's "random". Random does not mean unconfined; you have to be batshiat insane to guess a craps roll will NOT be between 2 and 12. Same thing as locating a knuckleball; properly thrown it will more or less wind up somewhere over the plate or maybe even a portion thereof. Me calling it "random" doesn't mean it'll suddenly defy the laws of physics, swoop upwards out of the park, jump through a circus hoop, cook your dinner and then lodge itself in Richard Gere's asshole. The trajectory is confined but unpredictable, and unfortunately for the batter, hitting a round ball with a cylindrical bat requires precision.


Oh, and as for baseball: no. You're explaining too much. Yes, two rolled dice will result in something between 2 and 12. For the knuckleball analogy to be apt, it would have to be somehow... constrained... what's the expression...

dragonchild:

Again, no. The knuckleball is the only pitch you can never adjust to. Knuckleball pitchers are not "figured out"; they implode. Every other pitch in baseball is deliberately aimed.


... ah yes, that's it, "deliberately aimed"!

Sorry buddy, but your logic sucks. What you're saying is that all swings at knuckleballs are just flailing for a random chance to hit something, and they won't get any more successful no matter how much practice a batter has and how many looks the batter has had at the pitches. And although it looks that way sometimes, with the best knuckleballers, that's not how it works. It's a skill, and it's a skill that can be acquired. But most batters don't have much of an opportunity to acquire the skill. Which brings me back to the original point that the relative scarcity of knuckleballers is an asset.

Now show data that no batter is better than any other batter at hitting the knuckleball, or we're done.
 
2012-06-22 02:18:09 AM  

dragonchild: truth is not democratic


"Quid veritas?"

Objective truth is exceedingly hard to find-- most is at best extrapolated from presumptions or potentially flawed reports. Even mathematical truth is flawed-- it works because we tell ourselves it works and because we need it to work, but exceptions and caveats pile up pretty quickly in serious study.

Human truth is, in fact, precisely democratic. We need observation, confirmation, repetition, reason, and agreement in order to describe something as true; sounds like an idealised form of democracy to me. Reality is consensus, truth is democratic, and nothing is as it appears-- which means that appearances have to suffice, and interpreted reports are sometimes the best we can do.
 
2012-06-22 07:10:09 AM  

Super Chronic: Now show data that no batter is better than any other batter at hitting the knuckleball


When you know full well that there aren't enough knuckleball pitchers out there to get a statistically acceptable sample size of any batter in the majors? Either you're incapable of reason or trolling, so yeah, we might just finally agree on something -- we're done. There are bigger fish to fry in the world and I kinda liked this discussion more back when it was fun.
 
2012-06-22 07:15:10 AM  

DeWayne Mann: And while I was typing up that stupidly long post for no reason, 2 more posts appear!

dragonchild: DeWayne Mann: My point is that we have NO IDEA what the "norm" is for him (and those previous 400 IP tell us very, very little about it). Will he regress? Probably. It's generally safe to assume any player will regress no matter what the question is.

Ah, but what's truly exciting is because we have no idea, there's a chance (however small) he won't regress. Methinks part of this excitement stems from the fact that even skeptics know, on some gut level, that there's a small chance we're witnessing something historical. And that's precisely because we don't have a perfect set of data to compare Dickey's run to. Only time will tell, of course, but it'd be worth it to sear these games into memory for that chance.

Yeah, this is essentially what I'm saying. We can say Hellickson's gonna regress because we can compare him to a couple of dozen thousand other pitchers. We can compare Dickey to...um.

Well...

...

Yeah.

You're the jerk... jerk: There is so little variation in BABIP

WHAT? Seriously, either you phrased this wrong, I'm reading it wrong (which are sorta two sides of the same coin), or you're just making stuff up as you go along.

There's little variation in league BABIPs. But among individual players, the variation is MASSIVE. That's the entire reason anyone uses BABIP as a stat.

I mean, again, here are 5 of the best pitchers ever:

[i1154.photobucket.com image 640x379]

Little variation? How many times can you find three consecutive points that average together to be the guy's overall BABIP? There are few, sure. But there are also a lot that are no where close.


Sorry, I meant to say "Little variation in career BABIP"
 
2012-06-22 07:37:58 AM  

DeWayne Mann: Doing some research:

From Tom Tango:

It's still our best guess that if a pitcher has a (1b+2b+3b)/BIP rate of .320 and the league is .300, then a pitcher's "true" talent, based on the BIP, is .305 (80% regression towards the mean, or 1-r). This applies to pitchers with 500 to 1200 PAs.

In other words, with 1200 PA (somewhere around 300 IP), his "best guess" is that the pitcher's BABIP talent is probably not close to his past BABIP, and you're really better off just using the league average.

But, and I feel I have to point this out yet again: All of the these BABIP studies have little or nothing to do with knuckleballers, who have generaly been observed to have below league average BABIP-talents. There's very, very little data on knuckleballers, though. Essentially, trying to project them is a fool's errand.


So I typed up a response to this and my computer ate it. But essentially the r squared in their for BABIP is .2 (good find, thanks!) so with two seasons worth of data we can't predict BABIP as well as I expected. But even if it were .9 you you would say it doesn't apply to knuckleballers. Maybe that is true, but I am guessing knuckleballers behave somewhat like regular pitchers and this underlies all my assumptions. There have been some "bad" knuckleballers like Steve Sparks who have done worse than league average for BABIP, we just remember the good ones.
 
2012-06-22 07:44:01 AM  

DeWayne Mann: DeWayne Mann: My point is that we have NO IDEA what the "norm" is for him (and those previous 400 IP tell us very, very little about it). Will he regress? Probably. It's generally safe to assume any player will regress no matter what the question is. But you're trying to make a case using insufficient data.

Here's a really, really stupid analogy I just came up with.

Pretend I know nothing at all about you. But you randomly tell me that you're going to run in a marathon, and you have a goal of 2 hours, 30 minutes.

I might suspect that you will not reach this goal. Because that's a tough goal, and most people can't do it. And as far as I know, you're "most people."

Then you tell me you've run 20 marathons before. So I check, and I learn that your last two marathons had times of 2:45, but I can't find out what any of your other times are. I still might not think that you'll reach your goal, but it's certainly more likely.

So, on the day of the race, you run the first 13 miles in an hour.

How likely is it that you'll reach your goal, and what of the above data should you use to determine that?

I hope the parallels are obvious, but just in case:

"You" are Dickey. "People" are pitchers, and "most people" are normal pitchers. "People who have run marathons" are knuckleballers. The time is their BABIP: we'd expect "most people" to have higher times than "people who have run marathons" just as we'd expect "normal pitchers" to have higher BABIPs than knuckleballers.

The goal is a ~.250 BABIP for the season. The previous two "marathons" resulted in ~.270 BABIPs. The other 18 "marathons" are the fact that, from before, we have "10%" of the data. And so far, he's slightly ahead of the pace for the goal.

Like I said, it's dumb. But does it at least show why I'm skeptical of the "400 IP!" claim?


No this is good, because when I see this I would still predict he finishes at around his previous times. I understand it isn't a certainty, but I am just looking at what is likely. As I stated before, I just asked myself would you take an even money bet that his BABIP is above X the rest of the year? And looking at the information I have if X was below .280 and my wife would let me gamble I would.
I guess in this world there are two types of people, those who are willing to make extrapolations from incomplete data
 
2012-06-22 12:48:45 PM  

You're the jerk... jerk: Sorry, I meant to say "Little variation in career BABIP"


That makes a bit more sense, but it's related to the fact that the league average tends to vary very little.

You're the jerk... jerk: There have been some "bad" knuckleballers like Steve Sparks who have done worse than league average for BABIP, we just remember the good ones.


Well, Voros' original work said "Also to [note] is that, anecdotally, I believe pitchers with trick deliveries (e.g. Knuckleballers) might post consistently lower $H numbers than other pitchers" and Tom Tiplett accidentally found the same thing when (poorly, IMO) trying to disprove Voros.

Most likely, this is an argument over the definition of "knuckleballer." If a guy throws one knuckleball a year, do we expect him to post lower BABIPs than someone who doesn't? No. Smoltz & Ryan Franklin both threw them from time to time. Do they count? No clue.

Obviously, Sparks mostly threw knucklers, so that doesn't really apply. But he actually did post below average BABIPs (with one major exception). Here's a graph of only him:

www.fangraphs.com

That being said, I did come across a reference to a knuckleballer I had completely forgotten about, and one who is (somewhat) worth mentioning in relation to RA Dickey, because he threw a "fast knuckler" with an average velocity of 86. Here's Jared Fernandez's BABIP graph:

www.fangraphs.com

guh...of course, that graph covers 100 IP, and the biggest season was a mere 50 IP, so it's not really helpful for anything we're doing. but hey, at least he exists.

You're the jerk... jerk: No this is good


Someone who doesn't think my terrible analogies are the worst thing ever? IMPOSSIBLE.

You're the jerk... jerk: I guess in this world there are two types of people, those who are willing to make extrapolations from incomplete data


OH NO YOU DIDN'T FINISH THE POST AND I DON'T KNOW WHAT THE SECOND TYPE OF PERSON IS.

I'm gonna assume it's "people who like cheese."

Mmmmmm, cheese.
 
2012-06-22 01:07:03 PM  

DeWayne Mann: he threw a "fast knuckler" with an average velocity of 86


No he didn't.

Sorry, I didn't get a lot of sleep last night. COMPLETELY misread that chart. His average knuckler was like 70.

I'm dumb.
 
Displayed 41 of 91 comments

First | « | 1 | 2 | » | Last | Show all

View Voting Results: Smartest and Funniest


This thread is archived, and closed to new comments.

Continue Farking
Submit a Link »
On Twitter





In Other Media


  1. Links are submitted by members of the Fark community.

  2. When community members submit a link, they also write a custom headline for the story.

  3. Other Farkers comment on the links. This is the number of comments. Click here to read them.

  4. Click here to submit a link.

Report