Loaiza vs others
DMZ · January 28, 2006 at 12:20 am · Filed Under Mariners
One more:
Loiaza: 196 IP, 202 H, 51 B, 131 K, 23 HR, 4.42 park-adjusted ERA. 3y, $21m
Rogers: 165 IP, 198 H, 49 BB, 80 K, 24 HR, 5.59 park-adjusted ERA. 2y, $16m
annnnd then back to Washburn.
Loiaza may be the best free agent pitching acquisition any team made this off-season.
I’m not sure about “the best” acquisition, but you are the first person I’ve seen suggest publicly that the Loiaza signing makes sense. Good to know I’m not alone.
(per Bill James handbook)
2005 MLB % Pitches in strike-zone
C. Silva: 65.2
B. Colon: 60.8
P. Byrd: 59.8
E. Loaiza: 59.4
D. Wells: 59.2
J. Towers/M. Buehrle: 58.8
G. Maddux: 58.7
A. Harang: 58.5
P. Martinez: 58.3
There’s no “rule,” but I don’t think it’s a coincidence either that all the leaders in this category had pretty good seasons. Pitchers who are effective while keeping the ball in the strike zone are a pretty nice and reliable commodity. If a pitcher can strike guys out without leaving the k-zone, like Loaiza or Buehrle, are particularly interesting. (I left out the Pedro’s and Sheets’s of the world because they are a totally different breed.)
Dave called it “a very good signing” back at the time.
I’m going to go out on a limb and say that Loaiza doesn’t perform all that well. I don’t disagree it’s a good signing from the risk / reward point of view, but with Loaiza’s inconsitency in his career, I’m amazed PECOTA came up with such favorable numbers.
I have a feeling PECOTA blows this one. It’s not perfect, and one of out every so often projections is wrong. I have a feeling it’s this one.
Or Kenji Jojima’s.
Well, of course it misses: like all projection systems, to some degree it sucks at predicting the future.
There’s a difference between being wrong on a projection and missing on a projection. If I have a fair coin and I predict that it will land heads 50% of the time, and then over 100 tosses it lands heads 40% of the time, then my projection was correct, but it missed. Give me 10,000 tosses, though, and my projection probably won’t miss by much.
If we really thought the Loaiza projection was wrong, and we wanted to show that it’s wrong, we need to:
A) Identify a reason why the projection is wrong
B) Find a group of pitchers whose projections are wrong for the same reason
C) See how that group of pitchers does relative to their projection
Yeah, and yet also I want to recognize that no system’s a good fit for every player, and that they’re all going to do really bad forecasts of some players for reasons that are outside the system, but the “why” of that is a huge article in intself.
I bet Nate Silver has to replace his hard drive every time PECOTA takes on Ichiro. Or any historically unique player.
The following quote (from a free BP chat) out to inflame your passions a bit, Jim. To Wit:
“ChuckR (Addison, IL): Looks like Pecota is pretty bearish on Ichiro – VORP less than 20 and collapse at practically 50%. Is this one of those outliers, Ichiro being a fairly unique player in MLB history? If so are there any other modern day players who you see as ‘unique’ and thus not subject to the deadly accuracy of the system?
Nate Silver: It’s easier to say this after last year, but I’ve started buying into PECOTA’s pessimism on Ichiro. One of the improvements we made this year was to look at groundball-flyball numbers for hitters. That does help his batting average a bit. But it has the opposite effect on his isolated power, and doesn’t lead to a more optimistic projection overall.”
man, I can not spell tonight. That should read “OUGHT to inflame” and “to WHIT”
“Or any historically unique player.”
It might be worth mentioning that PECOTA doesn’t especially think that Ichiro is historically unique. It doesn’t look like the similarity scores are out for 2006 projections yet, but it’s a safe bet they are pretty close to the 2005 similarity scores. In 2005, Ichiro had a similarity score of 21, which is on the low side, but not anything I’d call historically unique. Trying some names that came to mind, I found that Randy Johnson and Barry Bonds each had similarity scores of 0, and Roger Clemens had a similarity score of 6. On the other side of things, with guys who are clearly not all that unique, Sexson gets a fairly high similarity score of 49 and Beltre had a similarity score of 53.
And I’m not really sure how close PECOTA was on Ichiro before, but last year PECOTA had a weighted mean EQA of .282, a midpoint EQA of .277, and Ichiro actually posted a .279 EQA. Ichiro’s 2005 50th percentile projection was .306/.349/.410 and his actual 2005 was .303/.350/.436. I’d say that’s an awfully accurate projection, and that Silver is justified in his increased confidence in PECOTA’s Ichiro projection.
And in an attempt to stay on topic, Loaiza’s 2005 projection didn’t look a whole lot different from his 2006 projection:
2005 PECOTA: 158.7 IP, 160 H, 51 BB, 112 K, 18 HR, 4.39 park-adjusted ERA
If anything, it looks like PECOTA thinks Loaiza should see more playing time, see a slight decrease in his K, BB, and HR rates across the board, but wind up with about the same ERA just the same. He had a similarity score of 56, so I don’t see any reason to doubt PECOTA all that much here.
Yeah, PECOTA nailed Ichiro in 05, but it missed him completely nearly every year before that.
Who does PECTOA sees as Ichiro’s historic comps? I bet they’re not very comparable.
His Baseball Reference comps have low comparability scores … and, frankly, are pretty ridiculous. Benny Kauff? Bake McBride? Yeah, right.
I’ll stick with “historically unique” in regard to Ichiro.
I confess that PECOTA knows a lot more players than I do, and the only players I really recognize on his list are Tony Gwynn and Rod Carew. The Gwynn comp clocks in at a similarity score of 22, again, an alright comp, but not a great one, and Carew is down at 9, so he’s not an especially close ocmp, and likely isn’t weighted very heavily in the projection. Gwynn seems like a perfectly reasonable comp to me, though. If anything, Gwynn walked a bit more and slugged slightly more, but both were/are great contact hitters who don’t strike out much.
The top 4 most similar, by PECOTA, are Matty Alou, Lance Johnson, Cesar Tovar, and Mickey Rivers. (Gwynn was fifth.) Glancing through their stats, they all look like they have the same basic skill set as Ichiro–low K-rate, high avg, low ISO. And at any rate, I think the baseball reference comps are a joke compared to PECOTA comparables, and I have seen barely any correlation between the two systems.
Take Matty Alou, for instance. His career translated line is .322/.363/.420. Ichiro’s career translated line is .347/.396/.468. Ichiro’s is better, but if you add in whatever his age 22-26 seasons would have been and then some age 32-35 decline, is it that unlikely his line would be ~.020/.036/.048 or so worse? I’m not seeing a ton of evidence for Ichiro being historically unique. (And yes, I still think Ichiro is awesome.)
“Yeah, PECOTA nailed Ichiro in 05, but it missed him completely nearly every year before that.”
PECOTA also had a lot more data on Ichiro for the ’05 projection. The ’04 season was basically unpredictable (that is, it’s not prudent to project that someone’s going to break the hits record), and for ’03 and prior, PECOTA had two or fewer seasons of major league data to work with on Ichiro, whereas now it has five full seasons.
And PECOTA has 11 seasons of data for its Loaiza projection.
For 2005, PECOTA tagged Matty Alou as the best comp for Ichiro – Alou’s similarity score was 71. There was a big drop after that to Lance Johnson (39), and Cesár Továr (36). Behind Továr, there was another drop to Mickey Rivers (25), Tony Gwynn (22), and Luis Polonia (21).
Luis Polonia. Sheesh.
Ichiro should sue Baseball Prospectus for libel.
Polonia has a career translated line of .309/.360/.418, and at age 35, in a half season with the Tigers, he managed a .321/.351/.533 translated line. I don’t think Ichiro, or anyone for that matter, should consider that comparison libelous, especially when the attached similarity score of 21 admits that it’s not a perfect, but an alright comparison.
Except Polonia had the worst throwing arm of any regular in the majors, and Ichiro’s is among the best … if not the best.
Mat,
BP lists a sim score of 20 as the boundary of ‘historically unusual,’ so ichiro is clearly atypical. 50 or so seems to be avergage, so perhaps part of the disagreement here is what that sim score means. To me, a sim score of 21 means ‘don’t bet on it.’ Of course, Alou is in the 70s, though he’s the only one.
As for Loaiza, his comp lists are a real mixed bag – there are a few guys, like Dave Stewart and Gaylord Perry that had successful seasons after age 33. But there are just as many guys, like Steve Trachsel, Charles Nagy or Pedro Astacio, who burn out soon after that age. Separating those two categories of players, both on his top ten list, is more art than science. The tricky thing for PECOTA is that Loaiza is one of the most inconsistent pitchers I can think of. Forget ERA or W/L, his K rate vacillates between 5 and 7+ all the time. That’s huge – who knows where it’ll be next year. PECOTA splits the difference, and has him at 6k/9ip. That’s possible, but I think the overall picture is skewed by his great 03 and 05 seasons, and not by his awful…um…pretty much every other year.
I’ve said it before and I’ll say it again. He’s slightly worse than Jeff Weaver and he’s a few years older. It’s possible that he’s a good value, but no free agent signing has a better chance of posting negative VORP than Loaiza. Overall, it’s probably a decent risk, but I’m not losing any sleep over the M’s missing out on the guy.
Esteban Loaiza was at the top of my acquisition list on price+value, yeah. I’d far, far rather have him than *gulp* Washburn. (I’d rather have Vicente Padilla than Washburn, too, for that matter.) That Oakland was all over Loaiza at a reasonable price while the Ms didn’t bother to get in the game when they could have been top bidder easy says everything, to me, regarding the relative success of the two organizations. Even if the Ms decided to hang onto their dough while making eyes at Millwood and Burnett, only to get priced out and into Plan B by which time Loaiza was long gone, that’s still a lousy approach.
I’m with you, Derek, on ‘fit to player’ of a given system as the key issue. Baseball skills are so composite in nature that _no_ system models players at the level of outcome of coin flips where there are really only three outocomes—heads, tails, and on the edge—only two of which are statistically significant; in that sense, the analogy is a poor one. OTOH, the more seasons we have for a player, the more projections can depend on self-similarity in the data set, and the better the outcome. That is the crux in the improved modeling of Ichiro, as Mat says above, and the principal reason for high confidence in the projection of Loaiza.
I wouldn’t call Ichiro historically unique at all; his skillset was in high demand in the dead ball era, for example. It is highly unusual over the last generation to see a player who makes so much contact with so little isolated power, however; that approach simply is not one that either players or organizations have chosen to cultivate or promote. That Ichiro actually manages to generate significant offensive positive results from it is . . . highly unusual, re: the discussion (last week?) of the offensive value of batted balls in play. Guys who hit the ball on the ground routinely are generally offensive zeros—and then there’s Ichiro!
Given the conditioning Ichiro puts in, I would anticipate more of a bounce back than a decline in ’06. But bear this in mind: the fact that Ichiro refuses to walk, regularly expands the strike zone away, and has a swing geared to hitting the ball on the ground gives him _very little_ margin for error. His bat speed, and even more his leg speed, are everything. If and when either is impaired of declines, his results WILL plunge. I don’t think it’s going to be this year, or next, but I’m not surprised to see that the projection system sees him as a high risk: his skillset does not age well at all historically.
Okay, so Ichiro is merely unique to his era. Since PECOTA is based solely on postwar players, that doesn’t help Nate’s job.
WHAT Ichiro does isn’t unique. HOW WELL he does it is, and any projections based on what he does are going to look weird.
The top FOUR players on Ichiro’s PECOTA comp lists never had a single season with a .300 EQA season. Ichiro’s had 2 in his first four years. Outside of Tony Gwynn, there are two TOTAL .300 EQA seasons on his entire top 10 comp list. Looking at K/BB and gb/fb, which is entirely reasonable for 99% of players, just doesn’t really work with Ichiro. It’s not really a matter of getting more seasons of data either; he’s still going to slap a lot of groundballs and have a low ISO with a very good EQA. He shouldn’t, but there it is. PECOTA will see that as confirmation that he really is Micky Rivers mark II, but he’ll still beat his projection.
As a bp loyalist still poring over my PECOTA spreadsheet, this shouldn’t be taken as a slam on PECOTA, of course. Ichiro’s just weird.
Ichiro’s definitely slightly weird. I just think we should draw a line between the Ichiro weirdness and the Bonds/Clemens/Johnson otherworldliness that I consider to be historical uniqueness. Seriously, though, Ichiro rocks and I’m glad to have him on the M’s side.
Ichiro is just proof that statistical analysis doesn’t always work.
I think Ichiro and Felix over the course of their careers will prove to be outliers all along the way. Just rare specimens.
Re: 24
I don’t really understand your comment. Not only did PECOTA nail Ichiro’s performance last year but also stat. analysis provides fairly precise means with which to define Ichiro’s contribution to the team. Also, people love Felix because he both has great stuff AND great stat. indicators which foreshadow future success. Neither player defies statistics in any appreciable way.
Player A: .338 Batting average .388 OBP, .459 SLG
Player B: .332 Batting average, .377 OBP, .442 SLG
Both players play RF, both players had good arms and Gold Gloves while playing RF, both players have a career high in stolen bases at 56 that they haven’t approached again, both players have multiple batting titles.
Ladies and gentlemen, the obvious comp to Ichiro is Tony Gwynn, who isn’t exactly a pre-WWII player. Gwynn probably rates higher on power and walks, but it’s not a blowout.
Matty Alou isn’t a HORRIBLE comp, but the dude had 31 lifetime home runs, and Ichiro is totally wailing him on things like Black Ink.
Also: Ichiro was a pretty good player very early on in his career in Orix- he hit .385 at age 21 in the J leagues. I suspect if he had been in the majors, he could have easily won batting titles early like Gwynn did.
“Not only did PECOTA nail Ichiro’s performance last year”
And missed it every other year, which would fit perfectly with the statement that
“statistical analysis doesn’t always work”
PECOTA misses for a lot of players every year (if it did not, Silver would be an extremely wealthy man). He does not argue that every predictive system has an inherent fallibility. Rather he claims that Ichiro somehow defies stat. analysis