A crystal ball is for baseball front offices what the philosophers' stone was for medieval alchemists: probably the single most desirable tool out there. And there have been a lot of predictive tools, including tools that can spit out a whole slash line of MLB future for a 19 year old prospect in high A. I think that some of them, at least at that level, are borderline silly.
I took a look at potentially creating at a tool that was a bit more complicated that ERA/FIP or xFIP differential that can tell a couple of things at the same time: a. how good has someone been and b. how good is someone going to be in the near future. I did not want to predict W-L, ERA, IP and such. That is silly in my book. So I run the thing through 2009 to 2010, 2010 to 2011, 2011 to 2012 season differentials of individual pitchers to look for accuracy as far as improvement and decline went, and I got about 80% accuracy for pitchers who started a baseline of 100 innings the previous season.
Not that bad, but the algorithm still needs refinement for the lower inning pitchers; ideally I would like it to work at 50, so one could be able to potentially extrapolate September call up performances for the next season. Also, there is one thing that math cannot do, and that is take into account whether a young pitcher improves a particular pitch or learns another before the next season. At this point, I would say that it is not really great to predict young pitchers' performance. So it is not ready for release. One thing that I feel pretty confident about is that it is pretty good to predict mid-late career pitchers' performance. Free agents do fall into this category, so with the winter meetings coming up, I felt that I could present the predictions about starting pitcher performance in 2014. This includes Ricky Nolasco and Phil Hughes, the two Twins' free agent signees.
Mainly a bookmark, so I can check again after the season to see how it did vs. actual performance, but I thought that it might be fun to share.
Here is the list (with a lot of incompletes, as I indicated)
I am indicating the Twins' signees in bold and potentially good targets yet unsigned in italics and underlined. According this crystal ball the Twins did pretty well...
Alfredo Aceves (31) - not enough in 2013
Bronson Arroyo (37) - mid rotation - Consistent Decline
Scott Baker (32) - not enough in 2013
Erik Bedard (35) - end of rotation - Consistent Improvement
Travis Blackley (31) - not enough in 2013
A.J. Burnett (37) - top rotation - Consistent Improvement
Chris Capuano (35) - mid rotation - Consistent Improvement
Chris Carpenter (39) - not enough in 2013
Bruce Chen (37) - end of rotation - Consistent Decline
Bartolo Colon (41) - mid rotation - Consistent Decline
Scott Feldman (30) - end of rotation - Consistent Decline
Gavin Floyd (31) - not enough in 2013
Jeff Francis (33) - not enough in 2013
Freddy Garcia (37) - not enough in 2013
Jon Garland (34) - not enough in 2013
Matt Garza (30) - mid rotation - Consistent stay the same
Chad Gaudin (31) - not enough in 2013
Roy Halladay (37) - not enough in 2013
Jason Hammel (31) - end of rotation - conflict: same or improvement
Aaron Harang (36) - mid rotation - conflict: decline or improvement
Dan Haren (33) - top rotation - Consistent Improvement
Roberto Hernandez (33) - mid rotation - Consistent Improvement
Phil Hughes (28) - mid rotation - Consistent Improvement
Ubaldo Jimenez (30) - mid rotation - conflict: decline or improvement
Josh Johnson (30) - not enough in 2013
Jair Jurrjens (28) - not enough in 2013
Jeff Karstens (31) - not enough in 2013
Scott Kazmir (30) - top rotation - Consistent Improvement
Hiroki Kuroda (39) - mid rotation - Consistent Decline
John Lannan (29) - not enough in 2013
Colby Lewis (34) - not enough in 2013
Ted Lilly (38) - not enough in 2013
Paul Maholm (32) - end of rotation - Consistent Improvement
Shaun Marcum (32)- not enough in 2013
Jason Marquis (35) - replacement level - Consistent Decline
Daisuke Matsuzaka (33) - not enough in 2013
James McDonald (29) - not enough in 2013
Jeff Niemann (31) - not enough in 2013
Ricky Nolasco (31) - mid rotation - Consistent Improvement
Sean O'Sullivan (26)- not enough in 2013
Roy Oswalt (35)- not enough in 2013
Mike Pelfrey (30) - end of rotation - Consistent Improvement
Greg Reynolds (28) - not enough in 2013
Clayton Richard (30) - not enough in 2013
Ervin Santana (31) - mid rotation - Consistent Decline
Johan Santana (34) - not enough in 2013
Joe Saunders (33) - replacement level - Consistent Improvement
Kevin Slowey (30)- not enough in 2013
Masahiro Tanaka (25) - not enough in 2013
Jason Vargas (31) - mid rotation - Consistent Improvement
Ryan Vogelsong (36) - not enough in 2013
Edinson Volquez (30) - mid rotation - Consistent Improvement
Tsuyoshi Wada (33) - not enough in 2013
P.J. Walters (29) - not enough in 2013
Jake Westbrook (36) - not enough in 2013
Chien-Ming Wang (34) - not enough in 2013
Suk-Min Yoon (27) - not enough in 2013
Barry Zito (36) - replacement level - Consistent Improvement
I took a look at potentially creating at a tool that was a bit more complicated that ERA/FIP or xFIP differential that can tell a couple of things at the same time: a. how good has someone been and b. how good is someone going to be in the near future. I did not want to predict W-L, ERA, IP and such. That is silly in my book. So I run the thing through 2009 to 2010, 2010 to 2011, 2011 to 2012 season differentials of individual pitchers to look for accuracy as far as improvement and decline went, and I got about 80% accuracy for pitchers who started a baseline of 100 innings the previous season.
Not that bad, but the algorithm still needs refinement for the lower inning pitchers; ideally I would like it to work at 50, so one could be able to potentially extrapolate September call up performances for the next season. Also, there is one thing that math cannot do, and that is take into account whether a young pitcher improves a particular pitch or learns another before the next season. At this point, I would say that it is not really great to predict young pitchers' performance. So it is not ready for release. One thing that I feel pretty confident about is that it is pretty good to predict mid-late career pitchers' performance. Free agents do fall into this category, so with the winter meetings coming up, I felt that I could present the predictions about starting pitcher performance in 2014. This includes Ricky Nolasco and Phil Hughes, the two Twins' free agent signees.
Mainly a bookmark, so I can check again after the season to see how it did vs. actual performance, but I thought that it might be fun to share.
Here is the list (with a lot of incompletes, as I indicated)
I am indicating the Twins' signees in bold and potentially good targets yet unsigned in italics and underlined. According this crystal ball the Twins did pretty well...
Alfredo Aceves (31) - not enough in 2013
Bronson Arroyo (37) - mid rotation - Consistent Decline
Scott Baker (32) - not enough in 2013
Erik Bedard (35) - end of rotation - Consistent Improvement
Travis Blackley (31) - not enough in 2013
A.J. Burnett (37) - top rotation - Consistent Improvement
Chris Capuano (35) - mid rotation - Consistent Improvement
Chris Carpenter (39) - not enough in 2013
Bruce Chen (37) - end of rotation - Consistent Decline
Bartolo Colon (41) - mid rotation - Consistent Decline
Scott Feldman (30) - end of rotation - Consistent Decline
Gavin Floyd (31) - not enough in 2013
Jeff Francis (33) - not enough in 2013
Freddy Garcia (37) - not enough in 2013
Jon Garland (34) - not enough in 2013
Matt Garza (30) - mid rotation - Consistent stay the same
Chad Gaudin (31) - not enough in 2013
Roy Halladay (37) - not enough in 2013
Jason Hammel (31) - end of rotation - conflict: same or improvement
Aaron Harang (36) - mid rotation - conflict: decline or improvement
Dan Haren (33) - top rotation - Consistent Improvement
Roberto Hernandez (33) - mid rotation - Consistent Improvement
Phil Hughes (28) - mid rotation - Consistent Improvement
Ubaldo Jimenez (30) - mid rotation - conflict: decline or improvement
Josh Johnson (30) - not enough in 2013
Jair Jurrjens (28) - not enough in 2013
Jeff Karstens (31) - not enough in 2013
Scott Kazmir (30) - top rotation - Consistent Improvement
Hiroki Kuroda (39) - mid rotation - Consistent Decline
John Lannan (29) - not enough in 2013
Colby Lewis (34) - not enough in 2013
Ted Lilly (38) - not enough in 2013
Paul Maholm (32) - end of rotation - Consistent Improvement
Shaun Marcum (32)- not enough in 2013
Jason Marquis (35) - replacement level - Consistent Decline
Daisuke Matsuzaka (33) - not enough in 2013
James McDonald (29) - not enough in 2013
Jeff Niemann (31) - not enough in 2013
Ricky Nolasco (31) - mid rotation - Consistent Improvement
Sean O'Sullivan (26)- not enough in 2013
Roy Oswalt (35)- not enough in 2013
Mike Pelfrey (30) - end of rotation - Consistent Improvement
Greg Reynolds (28) - not enough in 2013
Clayton Richard (30) - not enough in 2013
Ervin Santana (31) - mid rotation - Consistent Decline
Johan Santana (34) - not enough in 2013
Joe Saunders (33) - replacement level - Consistent Improvement
Kevin Slowey (30)- not enough in 2013
Masahiro Tanaka (25) - not enough in 2013
Jason Vargas (31) - mid rotation - Consistent Improvement
Ryan Vogelsong (36) - not enough in 2013
Edinson Volquez (30) - mid rotation - Consistent Improvement
Tsuyoshi Wada (33) - not enough in 2013
P.J. Walters (29) - not enough in 2013
Jake Westbrook (36) - not enough in 2013
Chien-Ming Wang (34) - not enough in 2013
Suk-Min Yoon (27) - not enough in 2013
Barry Zito (36) - replacement level - Consistent Improvement
3 comments:
May I ask what stats were used to create this tool? Were you looking at other defense-independent numbers (tERA, SIERA)? I'm specifically interested in learning what points to Érik Bédard's consistent improvement prediction. Thanks!
The performance part of the tool is defense independent but it is not based on complex defense independent measurements like tRA & SIERA. And, unlike FIP, I totally ignore HRs and 3B, which I think are park factor and batter (not pitcher) dependent. (a) Components are hits, non-intentional bases on ball, strikeouts, hpb and such. I also (b) use BABIP at some point to normalize. (a) and (b) goes into the "expected performance" (top, mid, bottom, replacement) part, and it is relative to xFIP and SIERA and I bet there is a strong correlation esp. with SIERA, since there is with xPE (but it is not xPE or PE or aPE) and xPE correlates really strong with SIERA. Then I look at age and age/performance for individual pitchers and extrapolate where they are relative to an achieved or imaginary (depending on age) career peak (C). That all (a) and (b) and (c) goes into the "decline" or "improvement" determination.
Right now is a determination and not a measurement; which I am not sure I like, but on the other hand it is hard to express a bunch of things in one number...
Just a little math exercise.
Hope I did not confuse you much .
Oh, you didn't confuse me much, just some. * * It's hard for me to visualize it without the numbers, but I think I get the concept. Thanks for the reply!
Post a Comment