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.
Poll: Who Should Be in Next Year's Rotation? VOTE FOR FIVE!
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,924
|
Post by ericmvan on Sept 11, 2015 19:57:50 GMT -5
Sorry, I always take the bait, but I don't know that you can project a guy solely based on literally the best eight-start stretch of his career while also assuming that he's one of the rare guys with a significant true-talent ERA/xFIP split (based on all of 450 career innings, a decent chunk of which was spent coming out of the bullpen). What part of "if he can maintain a 98 xFIP-" didn't you get? The question I was trying to answer is how good he will be if he keeps pitching like this. That is anything but a foregone conclusion, and in fact I have not changed my fifth vote in the poll from Wright to Kelly, although I'm considering it. But if he does keep this up, how good is he? Not a 2.59 (IIRC) ERA guy, that's for certain. I think my conclusion, #3 starter, and probably a pretty good one, is in the ballpark. As for the xFIP - ERA difference, I've got data that shows that in at least some of his ERA (<)< xFIP stretches, his pitch selection with RISP was dramatically different than with non-RISP, and was almost certainly responsible for a sequencing skill. Basically, in his great run with the Cards he was a FB / change guy until guys got into scoring position, where he would nearly junk the change and throw a lot more sliders and curves, and be much more effective. The presumption is that he thought his breaking stuff could be very effective if not over-exposed, and so he saved it for "pitching in a pinch." That's the title of Christy Mathewson's book on pitching, so the idea of pitching differently when it was most important goes back a long way, and kind of blows up the idea that all situational splits are random. (At some point I'll go back and look at Dice-K's 2008, where he had insane situational splits.) And at some point I'll continue the Kelly data-gathering (alas, BrooksBaseball doesn't give pitch effectiveness by RISP / non-use, just usage, but I'm planning to e-mail Dan at season's end and tell him he has to add that!). But it's really clear to me that he's spent his whole career varying his pitch-mix approach, over stretches of starts, and sometimes situationally as well, and that the result has been some BABIP skill and (together or separately) some situational skill. The BABIP skill is close to indisputable, although it may not be part of his current run.
|
|
|
Post by jmei on Sept 12, 2015 10:53:00 GMT -5
Correlation does not imply causation, especially in small, cherry-picked samples. You are starting with his ERA < xFIP stretches and then looking backwards at what he might have done differently in those starts, which is the sort of p-hacking which leads to stuff like this. I don't think you do so intentionally, but I think you consistently read patterns into stretches of randomness by first noticing a statistical anomaly (often in a small sample) and then trying out hypotheses until you find one that fits. That's the definition of data dredging, and when you go about it backwards like that, the burden of proof is a lot higher than just trotting out a p-value or otherwise pointing at some correlations and calling it a day. Plus, past results aren't any guarantee of future results (for instance, if he maintained his optimal pitch selection patterns going forward, hitters may well pick up on it, and losing that element of surprise might eliminate any BABIP advantage he gains).
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,924
|
Post by ericmvan on Sept 12, 2015 12:17:07 GMT -5
Correlation does not imply causation, especially in small, cherry-picked samples. You are starting with his ERA < xFIP stretches and then looking backwards at what he might have done differently in those starts, which is the sort of p-hacking which leads to stuff like this. I don't think you do so intentionally, but I think you consistently read patterns into stretches of randomness by first noticing a statistical anomaly (often in a small sample) and then trying out hypotheses until you find one that fits. That's the definition of data dredging, and when you go about it backwards like that, the burden of proof is a lot higher than just trotting out a p-value or otherwise pointing at some correlations and calling it a day. Plus, past results aren't any guarantee of future results (for instance, if he maintained his optimal pitch selection patterns going forward, hitters may well pick up on it, and losing that element of surprise might eliminate any BABIP advantage he gains). But I didn't do any data dredging; I had the a priori hypothesis that his pitch mix differed between RISP and non-RISP and confirmed it on the first try. I severely don't have time to dredge. I'm actually pretty good at all that. I usually do have an idea of a cause in mind. On the rare occasions where I find something so interesting that I do some dredging, what I find has to make sense to me and then has to predict something else. In 2007 I noticed that every time Mike Lowell was hot, Drew (who was hitting in front of him) was cold, and vice versa, and in fact Drew flip-flopped about three days after Lowell did, as if it took opposing pitchers a few days to decide whether Lowell was protecting Drew. They were both streaky as hell, and the correlation was eerie. That gave me the idea that Drew was hurt by protection, that if pitchers went after him aggressively, they'd be surprised at how good the results were. That was already consistent with his being average against the top 50% of pitchers and killing the bottom 50% (it was actually 40-60 or 60-40, but you get the idea). But it also gave me the hypotheses that he should have better years when the rest of his team's lineup was weaker, which turned out to be true, and that he should hit better hitting 5th than 2nd, which turned out to be true, and that he should be significantly better with RISP and 1B open than in base-out situations where pitchers challenge hitters (like man in 1st), which turned out to be way true. Your second point is very true, and is probably the reason why he has had to keep shuffling his approach frequently. But I think that if hitters catch up to his pitching backwards, his xFIP is going to rise. And, BTW, there may not be a pitcher on the planet that Brian Bannister could help more than Kelly. And Wright is probably the opposite.
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,924
|
Post by ericmvan on Sept 12, 2015 12:23:37 GMT -5
Thought: I doubt there's enough data to figure out which specific hitters get most messed up by facing a knuckleball pitcher, but I bet there's enough to figure out which types of hitters do, and hence which opposing lineups are most susceptible.
You can use Steven Wright as your long man in the pen and give him 10 or so carefully chosen starts, in the first games of series, when either everyone in the rotation can use an extra day of rest, or one pitcher can use a skipped start. And really, really benefit.
|
|
|
Post by thelavarnwayguy on Sept 12, 2015 17:39:06 GMT -5
Would Leake be considered a major FA signing or a small time? Kazmir? To me both would be major:
1) Price/ Leake/ Kazmir 2) Porcello 3) Buchholz 4) Rodriguez 5) Kelly 6) Miley ( thinking he could work better out of the pen than Kelly ) 7) Johnson 8) Owens
Trade Wright for starter depth which is controllable.
For the record, I'm not digging any of them past the 1 slot. Never thought Porcello was a good extention. I like Buchholz but we can't count on him. Rodriguez I like as much as any of them going forward. Don't think Owens or Johnson are major league starters in 5 years.
|
|
|
Post by jimed14 on Sept 13, 2015 14:27:57 GMT -5
Can Rich Hill be added to the poll?
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,924
|
Post by ericmvan on Sept 13, 2015 23:26:51 GMT -5
Can Rich Hill be added to the poll? He might well be included in some kind of poll or survey about the potential starters that I'll probably cook up when the season ends. I want to poll people on both what needs to be done to the rotation (zero, one, or two additions) and the best way to fill the vacancies (free agency versus trades) as well as take a temperature reading on specific candidates both internal and external. The first two are easy to combine, but how best to do the third one is unclear to me. I would love, for instance, to know not just how many people want to trade Buchholz, but who they have in mind to take his rotation spot.
|
|
|