SoxProspects News
|
|
|
|
Legal
Forum Ground Rules
The views expressed by the members of this Forum do not necessarily reflect the views of SoxProspects, LLC.
© 2003-2024 SoxProspects, LLC
|
|
|
|
|
Forum Home | Search | My Profile | Messages | Members | Help |
Welcome Guest. Please Login or Register.
|
Post by jmei on May 16, 2016 16:17:37 GMT -5
SIERA adjusts for high IFFB rates, which at least partially accounts for how most knuckleballers have historically outperformed the league-average BABIP (both Dickey and Wakefield were above-average in IFFB, for instance). It also adjusts for the fact that pitchers with high fly ball rates generally tend to have lower HR/FB rates (call it the Chris Young effect). That effect (which, again, is accounted for by SIERA) at least partially explains why Wright's AVG/SLG on fly balls is better than those other pitchers (who are ground ball guys).
His SIERA doesn't change much if you exclude those relief appearances, but I'm also not convinced it is analytically appropriate to omit them. When you have a small sample to begin with, you shouldn't start shaving away the data points that don't fit your hypothesis. I suspect most starting pitchers are worse on pitches 1-15 than beyond (teams generally put their best hitters towards the top of a lineup) and I have a hard time concluding that Wright is a better true-talent starting pitcher than relief pitcher based on what I suspect is less than 100 innings of data.
|
|
|
Post by mgoetze on May 16, 2016 16:58:15 GMT -5
SIERA adjusts for high IFFB rates, [...] It also adjusts for the fact that pitchers with high fly ball rates generally tend to have lower HR/FB rates Wright has the batted-ball profile that SIERA likes least: almost no IFFB and a fairly even mix of FB to GB. I don't really have anything to add...
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,882
|
Post by ericmvan on May 16, 2016 17:05:50 GMT -5
SIERA adjusts for high IFFB rates, which at least partially accounts for how most knuckleballers have historically outperformed the league-average BABIP (both Dickey and Wakefield were above-average in IFFB, for instance). It also adjusts for the fact that pitchers with high fly ball rates generally tend to have lower HR/FB rates (call it the Chris Young effect). That effect (which, again, is accounted for by SIERA) at least partially explains why Wright's AVG/SLG on fly balls is better than those other pitchers (who are ground ball guys). His SIERA doesn't change much if you exclude those relief appearances, but I'm also not convinced it is analytically appropriate to omit them. When you have a small sample to begin with, you shouldn't start shaving away the data points that don't fit your hypothesis, no matter how strong your claimed justification is, and even if doing so has already resulted in dramatically better projections. I suspect most starting pitchers are worse on pitches 1-15 than beyond (teams generally put their best hitters towards the top of a lineup) and I have a hard time concluding that Wright is a better true-talent starting pitcher than relief pitcher based on what I suspect is less than 100 innings of data. FIFY, since that's the actual argument you're making here. But there is in fact no rule of thumb that says to regard all data in a SSS as relevant. In fact, failing to look at the data and ask whether some of it might be irrelevant is a beginner's stat mistake. Eliminating his start in early 2013 is a no-brainer. Even I didn't think he was an MLB-quality pitcher at that point (although I was very high on him because he had been learning the knuckler faster than everyone I could find except for Hough). His AAA performance breakthrough was half a season away, and it would be 1 1/2 seasons before he got an MLB starting shot on merit. It's not remotely relevant to the question of how good a pitcher he is now. When you look at the remaining data, there were two oddities: he was much better as a starter, and he had one awful start that was a huge outlier. That start was on very long rest. Those two things may have been random, or they may have had a common cause which in fact makes perfect sense for knuckleball pitchers exclusively: being on a regular routine greatly improves the odds that he has proper "touch" on the knuckler. Now, at this point I have a hypothesis. I didn't trim data to fit a hypothesis; the hypothesis is that the very odd outlier data ought to be trimmed, for reasons that are entirely credible. Because I invented the hypothesis to make more sense of the odd data, rather than having it to begin with (as we did, e.g., with the improved strike zone command of Chavis and Ockimey), I did my off-season analysis (IIRC) both with and without the removal of the long-rest start, which I thought was the most open to skepticism. I probably should have done it a third way, including the relief appearances (and in fact I'm not sure whether I did, at least a bit). I didn't think it would be worth the time spent, though. Well, so far, he has pitched much better than even the trimmed data set would lead you to believe. It is, however, quite credible as a performance based on the trimmed data set as a baseline. It's hardly credible as one based on his entire MLB career. I don't see you rejecting the hypothesis about Shaw using scouting reports and his intelligence as a justification for why a surprising amount of his MLB performance may be for real and not random. I mean, if you were intellectually consistent about this, we'd all be sick to death about hearing from you that Shaw was due for an imminent collapse at the plate, as opposed to me being mildly frustrated by your continued resistance to Wright's obvious talent.
|
|
|
Post by mgoetze on May 16, 2016 17:07:26 GMT -5
OK, I will add one thing though. One guy who is really good at getting IFFB and actually does have a high FB/GB rate and consequently a significantly better SIERA than xFIP is Koji Uehara. Even he has worse results on FBs than Wright (.163 AVG .452 SLG).
Uehara career 2.89 xFIP 2.26 SIERA <-- note big difference here Wright career 4.33 xFIP 4.19 SIERA <-- note almost insignificant difference here
|
|
|
Post by jmei on May 17, 2016 6:07:50 GMT -5
SIERA adjusts for high IFFB rates, [...] It also adjusts for the fact that pitchers with high fly ball rates generally tend to have lower HR/FB rates Wright has the batted-ball profile that SIERA likes least: almost no IFFB and a fairly even mix of FB to GB. I don't really have anything to add... Sorry, I should have been more explicit. Wright does not have the batted ball mix that most pitchers (including most knuckleballers) use to sustainably outperform their peripherals. His career Soft/Med/Hard%s are essentially league-average, he doesn't prevent hitters from pulling the ball more (if anything, his pull% is higher than league-average), and his exit velocity (eyeballing it from here) does not seem notably better than league average. His minor league BABIP since he was traded to the Red Sox is .295 (290.1 IP). Even if you give him a knuckball bump, I remain unconvinced that his true-talent BABIP is much lower than .280 or so, and if that's the case, I think he'll perform closer to his observed SIERA than his observed ERA going forward.
|
|
|
Post by jmei on May 17, 2016 6:33:52 GMT -5
SIERA adjusts for high IFFB rates, which at least partially accounts for how most knuckleballers have historically outperformed the league-average BABIP (both Dickey and Wakefield were above-average in IFFB, for instance). It also adjusts for the fact that pitchers with high fly ball rates generally tend to have lower HR/FB rates (call it the Chris Young effect). That effect (which, again, is accounted for by SIERA) at least partially explains why Wright's AVG/SLG on fly balls is better than those other pitchers (who are ground ball guys). His SIERA doesn't change much if you exclude those relief appearances, but I'm also not convinced it is analytically appropriate to omit them. When you have a small sample to begin with, you shouldn't start shaving away the data points that don't fit your hypothesis, no matter how strong your claimed justification is, and even if doing so has already resulted in dramatically better projections. I suspect most starting pitchers are worse on pitches 1-15 than beyond (teams generally put their best hitters towards the top of a lineup) and I have a hard time concluding that Wright is a better true-talent starting pitcher than relief pitcher based on what I suspect is less than 100 innings of data. FIFY, since that's the actual argument you're making here. But there is in fact no rule of thumb that says to regard all data in a SSS as relevant. In fact, failing to look at the data and ask whether some of it might be irrelevant is a beginner's stat mistake. Eliminating his start in early 2013 is a no-brainer. Even I didn't think he was an MLB-quality pitcher at that point (although I was very high on him because he had been learning the knuckler faster than everyone I could find except for Hough). His AAA performance breakthrough was half a season away, and it would be 1 1/2 seasons before he got an MLB starting shot on merit. It's not remotely relevant to the question of how good a pitcher he is now. When you look at the remaining data, there were two oddities: he was much better as a starter, and he had one awful start that was a huge outlier. That start was on very long rest. Those two things may have been random, or they may have had a common cause which in fact makes perfect sense for knuckleball pitchers exclusively: being on a regular routine greatly improves the odds that he has proper "touch" on the knuckler. Now, at this point I have a hypothesis. I didn't trim data to fit a hypothesis; the hypothesis is that the very odd outlier data ought to be trimmed, for reasons that are entirely credible. Because I invented the hypothesis to make more sense of the odd data, rather than having it to begin with (as we did, e.g., with the improved strike zone command of Chavis and Ockimey), I did my off-season analysis (IIRC) both with and without the removal of the long-rest start, which I thought was the most open to skepticism. I probably should have done it a third way, including the relief appearances (and in fact I'm not sure whether I did, at least a bit). I didn't think it would be worth the time spent, though. Well, so far, he has pitched much better than even the trimmed data set would lead you to believe. It is, however, quite credible as a performance based on the trimmed data set as a baseline. It's hardly credible as one based on his entire MLB career. I don't see you rejecting the hypothesis about Shaw using scouting reports and his intelligence as a justification for why a surprising amount of his MLB performance may be for real and not random. I mean, if you were intellectually consistent about this, we'd all be sick to death about hearing from you that Shaw was due for an imminent collapse at the plate, as opposed to me being mildly frustrated by your continued resistance to Wright's obvious talent. I didn't include his 2013 appearances in my above numbers. The rest is your typical modus operandi-- find a blip in the data that doesn't fit your conclusion and fit a post hoc hypothesis that justifies omitting it. Wright has a mildly better ERA as a starter (3.29 versus 3.88) in a tiny sample (48.2 IP reliever, 104 IP starter), so let's throw out the reliever data! Never mind that Wright's peripherals as a reliever have been better (4.48 xFIP starter, 4.03 xFIP reliever) or that we're talking samples so small that factors like quality of competition or weather may have an outsized effect once you begin slicing the data down. Don't get me started on throwing out one of his worst starts because he had too much rest. What gives it away is that you never omit data that doesn't fit your ultimate conclusion (e.g., the good starts Wright has had on six days' rest). We will see with Wright (and Shaw, for that matter).
|
|
|
Post by mgoetze on May 17, 2016 6:53:36 GMT -5
His career Soft/Med/Hard%s are essentially league-average, I have yet to see anyone attempt to prove that Soft/Med/Hard% stats are worth the pixels they take up on my monitor, and indeed Wright is a great case to induce skepticism of those statistics. Even at a conceptual level, I can't see the advantage of lumping in groundballs in one velocity range with flyballs in a different velocity range, etc. As for his minor league BABIP, fair enough. As you said, we shall see.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,882
|
Post by ericmvan on May 18, 2016 13:06:26 GMT -5
I don't really have anything to add... Sorry, I should have been more explicit. Wright does not have the batted ball mix that most pitchers (including most knuckleballers) use to sustainably outperform their peripherals. His career Soft/Med/Hard%s are essentially league-average, he doesn't prevent hitters from pulling the ball more (if anything, his pull% is higher than league-average), and his exit velocity (eyeballing it from here) does not seem notably better than league average. His minor league BABIP since he was traded to the Red Sox is .295 (290.1 IP). Even if you give him a knuckball bump, I remain unconvinced that his true-talent BABIP is much lower than .280 or so, and if that's the case, I think he'll perform closer to his observed SIERA than his observed ERA going forward. Looking at the play logs of a pair of Wright's games with high Hard% and low BABIP (4/22 and 5/3), there's good reason to believe that the three-bucket system fails for him. Put it this way, most simply: if you divided the Hard bucket in half, into Hard and Really Hard, I believe he'd be among the league-leaders in the last bucket. He gives up a ton of balls coded as Fliner (Liner), which count as Hard but are usually outs. He gives up very few hard fly balls. For the same reason, his average exit velocity is not that helpful. What we actually care about is the distribution of exit velocities at the high end. For pitchers with an ordinary distribution, we can use the average velocity as a proxy. But Wright's distribution is, I think, noticeably more peaked. The pitch is easy to make contact with, but its real late break is just enough to consistently avoid very hard contact. That moves the balls in the tails that fall into the hard and soft buckets closer to the middle. I agree that his career .252 BABIP in his relevant starts is probably due for regression, but I'd guess his skill is in the .265 or .270 range.
|
|
|
Post by jmei on May 31, 2016 14:05:22 GMT -5
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,882
|
Post by ericmvan on May 31, 2016 14:38:20 GMT -5
And the key stat in that article, that he's in the 98th percentile for keeping exit velocity below 95, quantifies the assertion I made just above about their distribution. Which I made, BTW, without any data at all. It just followed from the behavior of the knuckleball when thrown correctly, which has been the basis of my championing Wright from the get-go. We tend to think of two lenses on the game: scouting and stats. There is a third lens, however, and that is science: mostly physics and psychology, but also biomechanics and a bit of material science. The only reason I got the job as the Chair of SABR's Science and Baseball Committee is that Alan Nathan was quitting and we were friends, and I seemed to be vaguely qualified ... but I'm beginning to think that I and the committee are actually a good fit!
|
|
|