POLL(S): The Projections and You
Last Friday, Dave wrote about MGL writing about the significance of in-season projections. While we often find ourselves trying to find the value in current-season statistics, what MGL demonstrated is that, overall, projections worked better than putting too much weight on recent events. What MGL demonstrated is that, overall, the projections were outstanding, even when dealing with potential outliers. A short, hypothetical example: if you had a guy projected to hit .300, and for a little while he hit .400, and the projection was increased to .305, then that .305 would be the smartest bet the rest of the way. Players, simply, don’t often dramatically change their levels of true talent.
But of course, every rule has exceptions. Every projection system has players who disobey it and come out of nowhere to excel or suck a lot. This is where we spend a lot of our time — trying to identify players who are in the process of meaningfully changing. Players who, say, add pitches, or players who change their swing patterns. We’re always looking for guys for whom the projections might miss the mark. We know those guys exist — we just have to find them.
So in this post, you’re going to see 20 players and 20 polls. That’s a lot of polls, but I promise they’ll go by fast. There are five hitters out-performing their updated projections*, five hitters under-performing their updated projections, five pitchers out-performing their updated projections, and five pitchers under-performing their updated projections. For every player, one could make the argument that something has changed, and the projections just haven’t caught up yet. I want to get a feel for who you think has actually changed, and who might just be riding a streak. And while the poll answers are kind of subjective, they should do well enough — use your best judgment. From this exercise, what we’ll eventually learn is almost nothing. But it’ll be kind of fun and kind of informative, to look at tomorrow and to look at in October. Let’s just get this over with so you can do what you’re actually supposed to be doing.
(* – ZiPS and Steamer, blended, per usual around here)
Position Players
Lonnie Chisenhall
- Season: .438 wOBA
- Projection: .342
- What’s up: former good prospect hitting for more power, with reduced strikeouts!
Nelson Cruz
- Season: .419 wOBA
- Projection: .355
- What’s up: dingers, all over the place!
Victor Martinez
- Season: .419 wOBA
- Projection: .356
- What’s up: already more homers than last season and the season before!
Michael Brantley
- Season: .399 wOBA
- Projection: .334
- What’s up: for the first time, Brantley has hit like a power hitter!
Brian Dozier
- Season: .359 wOBA
- Projection: .317
- What’s up: Dozier supposedly made changes last year to stop being bad!
David Wright
- Season: .301 wOBA
- Projection: .347
- What’s up: if only every single peripheral weren’t trending in the wrong direction!
Mike Moustakas
- Season: .266 wOBA
- Projection: .312
- What’s up: Moustakas has already been demoted once after making awful contact!
Domonic Brown
- Season: .257 wOBA
- Projection: .331
- What’s up: Brown’s ISO is last year’s ISO minus half of it!
Brad Miller
- Season: .240 wOBA
- Projection: .308
- What’s up: the Mariners have been reduced to occasionally starting Willie Bloomquist!
Jedd Gyorko
- Season: .215 wOBA
- Projection: .304
- What’s up: not his BABIP and not his ISO!
Pitchers
Dallas Keuchel
- Season: 2.81 FIP, 2.79 xFIP
- Projection: 3.83
- What’s up: with a new slider, Keuchel has pitched like an ace for a team that isn’t bad anymore!
Jake Arrieta
- Season: 2.32 FIP, 2.89 xFIP
- Projection: 4.12
- What’s up: from last year, his K% – BB% has tripled!
Collin McHugh
- Season: 3.00 FIP, 3.37 xFIP
- Projection: 4.36
- What’s up: the supposed-to-be spot starter has struck out ten per nine innings!
Garrett Richards
- Season: 2.56 FIP, 3.28 xFIP
- Projection: 3.75
- What’s up: Richards is finally getting the strikeouts people have expected for years!
Jake Odorizzi
- Season: 3.24 FIP, 3.50 xFIP
- Projection: 4.28
- What’s up: Odorizzi has the same strikeout rate as Felix Hernandez!
Matt Cain
- Season: 4.79 FIP, 4.21 xFIP
- Projection: 3.83
- What’s up: we don’t know if Cain was really a DIPS-beater after all, given what he’s been lately!
Shelby Miller
- Season: 4.63 FIP, 4.56 xFIP
- Projection: 3.95
- What’s up: very early on, Dave declared that Miller was broken, and he hasn’t pitched nearly as well as his ERA!
Justin Verlander
- Season: 3.96 FIP, 4.72 xFIP
- Projection: 3.62
- What’s up: Verlander hasn’t struck out more than eight in a game all season, and lately he’s had barely more strikeouts than walks!
Clay Buchholz
- Season: 4.84 FIP, 4.79 xFIP
- Projection: 4.12
- What’s up: Buchholz was sufficiently bad that people figured his DL stint was for a phantom injury!
Felix Doubront
- Season: 5.03 FIP, 4.86 xFIP
- Projection: 4.28
- What’s up: Doubront’s strikeout rate has dropped from 24% to 20% to 15% since 2012!
Jeff made Lookout Landing a thing, but he does not still write there about the Mariners. He does write here, sometimes about the Mariners, but usually not.
The tricky thing is with a guy like Jake Arrieta. A 4.12 projection might be too high, but a 3.90 smells right. And is 3.90 statistically different than 4.12?
Are those pitcher projections for ERA or FIP?
They look like FIP David.
Ps. Nice exercise Jeff. Now, let’s re-visit after the year.
We have more faith in the projections now after getting burned with a bet on Chris Johnson.
I got this.
*Steps up to the podium. Taps mike*
“CHRIS JOHNSON WAS WHO WE THOUGHT HE WAS!”
Thank you.
Confused here. Didn’t MGL say he NEVER looks at in season stats. It made it sound like he was just using the beginning of the season projections and not incorporating the in season stuff to adjust the projections up or down.
“I have a database of my own proprietary projections on a month-by-month basis for 2007-2013. So, for example, 2 months into the 2013 season, I have a season-to-date projection for all players. It incorporates their 2009-2012 performance, including AA and AAA, as well as their 2-month performance (again, including the minor leagues) so far in 2013. These projections are park and context neutral.”
(etc)
Of course I incorporate the in-season stats into the projection. When I say I never “look at them” I mean that they have to significance to me whatsoever after they are incorporated into the projection, regardless of how much they differ from the projection.
It is funny how virtually EVERYONE quotes seasonal stats when talking about or implying future performance, especially 3 or more months into the season, yet, when they are asked to “put their money where their mouth is (IOW, be accountable),” 2/3 of them say, “The projection looks about right.” These are the people complaining about Wright, Moustakis and Butler, etc.
I’ll note for people, of course, this is based on ZiPS and Steamer, and not on MGL’s projections, but they shouldn’t differ very much, at all, since they’re based largely on the same data.
I’d like to see a comparison of *pre-season* projections and in-season performance. Comparing *updated* projections to in-season performance seems like it only proves the truism that projections project. The real dispute seems to be over how much weight to give this year’s partial-year performance versus previous years’ performance. Updated projections obscure that question.
Thanks for the reply. Like others have said the question becomes how heavy of a weight for recent vs old data and what kind of method is best for testing and declaring a winner. I think it depends on what you are using the projections for. A simple RMSE may be good for one person but a test that punishes big misses more will be more important to another persons application of projections. And the weighting will determine which does better.
People’s choices may be heavily influenced by the recent spate of articles on the subject. I wonder what this would have looked like a few days ago? Or if the poll had offered a choice in between 2014 actual and the projection?
“no significance”
Right now, big believers in Keuchel and Dozier, big disbelievers in Brown.
The one thing, I’ll bring up is projections are bad a recognizing a change in true talent level, especially in pitching.
Keuchel greatly improved his slider.
McHugh’s projection was based off a 50IP sample size and he started using breaking stuff more.
Arrieta started using his breaking stuff more.
Odorizzi added a splitter.
I’m not saying he should be using using current season numbers. I am saying when we see a large difference between projections and current season numbers, we should be digging deeper. Look at some more stats or a bit of basic scouting. I mean look at Puig. He has clearly made an adjustment in his approach and therefore changed his “true talent level” and projections don’t really see this.
Can I be “that guy” that just blindly chooses the projection for all of them because it would give me the bast chance of being right?
Sorry, GilaMonster, that wasn’t meant to be a reply to your comment.
Humans are still needed to analyze data and make sense of it. The human brain, biases and all, remains more powerful than these projections.
Brian Dozier over the last calendar year: 250/340/463. Second highest 2B WAR only below Matt Carpenter. His 30 HR’s are the most by 8 (Neil Walker). 22 SB’s as well. He does just about everything but hit for a high average.
Why not show some guys where the projection is obviously wrong? My pet pick is Betances