Bill James' Projections and Andy LaRoche
If you're a long-time WHYGAVS reader, you know that I love off-season player projections. No projection model is perfect, but I think that canvassing the best of them can give you a fairly good idea of what's going to happen next year. Accordingly, I was excited to get an e-mail from the publisher of the Bill James Handbook telling me that they would send me their projections early if I was willing to review them and write about them on my blog.
I don't know how to "review" projections, exactly, but I suppose it's fair enough for me to say that James' projections are certainly not my favorite. That distinction goes to the system that I'm fairly certain is the best: BP's PECOTA. Still, I like the projection section of his book, even if the projections in it are a bit unadventurous (more on that later) and even if my favorite parts of the book are John Dewan's Fielding Bible Awards and Plus/Minus defensive ratings. Every year James takes a look at the previous year's projections and discuss the ones they got especially right or terribly, terribly wrong (this year: Andruw Jones). It makes for interesting discussion about what caused the breakout or the drop in production which is always interesting given that James is one of the better baseball writers out there, even if people only give him credit for being a stat geek. And I do enjoy the level of transparency that allows him to say, "And Andruw Jones ... man, that has to be the worst projection we've ever published."
James' projections did quite well with some of the Pirates this last year. He picked Jason Bay for .276/.374/.498, accurately calling his bounce-back (actual: .282/.375/.516). He came close on Adam LaRoche too, calling .275/.350/.494 (actual: .270/.341/.500). He even did a nice job nailing down Ryan Doumit's breakout (projected: .280/.354/.484 in 103 games), actual: .318/.357/.501 in 116 games. He did miss on some of the Pirates. His predictions were a bit optimistic for Jack Wilson, didn't see Freddy's drop-off at all (to be fair, it was more severe than even I expected and I did think he'd fall off a bit in 2008), whiffed incredibly on Steve Pearce (.314/.366/.514 was James' projection there), and missed pretty badly on Nate McLouth (.255/.332/.429). That's fine, projecting baseball stats is an inexact science and James himself would tell you that. The goal is to get a feel for what a player's going to do, not write his performance in the coming season in stone.
It is interesting to note that the players that James' system seemed to miss by the most on where the players that did something significantly different in 2008 than they had in the past. This fits James' described model of "We just predict that players will mostly continue to do what they've done in the past." I don't know exactly how he does his projections, but it makes sense that basing things more on the players' past record than that of similar players would make drop-offs and breakouts harder to foresee. Of course, that means it might be hard to get a read on the projections for the one guy that I'm going to be keeping an awfully close eye on this winter: Andy LaRoche.
What, exactly, does an awful season at the age of 24 do to a guy that the computers loved coming in to this year? Last year, James had LaRoche pegged for .265/.367/.458. That would've been a great rookie year. After his 2008 stinkbomb, James' projection for him drops to .252/.347/.401. The projection still loves his plate discipline (74 walks and 83 strikeouts) and sees some home run power (18 dingers), but it basically sees him as Jose Bautista. That would be a huge improvement over what he was this year, but not particularly great as they key return on Jason Bay. It's strange, because the peripherals on the LaRoche projection are quite good. If he hits 18 home runs next year and walks 74 times with 83 strikeouts, I feel like he's going to be putting up numbers we're all pretty happy with. Still, the rate stats that James has him projected for are pretty bad, mostly thanks to the low batting average and not many doubles.
Anyways, keep this all in mind for the future because I'm going to be revisiting the Andy LaRoche projections when PECOTA and ZiPS and CHONE and the Hardball Times all come out with their numbers for 2009 as I try to get a handle on what the Pirates enigmatic (a nice way of saying, "I can't figure out why he can't hit") new third baseman and what we might be able to expect from him next year.
If you're interested in James' book, you can pre-order it here.
I don't know how to "review" projections, exactly, but I suppose it's fair enough for me to say that James' projections are certainly not my favorite. That distinction goes to the system that I'm fairly certain is the best: BP's PECOTA. Still, I like the projection section of his book, even if the projections in it are a bit unadventurous (more on that later) and even if my favorite parts of the book are John Dewan's Fielding Bible Awards and Plus/Minus defensive ratings. Every year James takes a look at the previous year's projections and discuss the ones they got especially right or terribly, terribly wrong (this year: Andruw Jones). It makes for interesting discussion about what caused the breakout or the drop in production which is always interesting given that James is one of the better baseball writers out there, even if people only give him credit for being a stat geek. And I do enjoy the level of transparency that allows him to say, "And Andruw Jones ... man, that has to be the worst projection we've ever published."
James' projections did quite well with some of the Pirates this last year. He picked Jason Bay for .276/.374/.498, accurately calling his bounce-back (actual: .282/.375/.516). He came close on Adam LaRoche too, calling .275/.350/.494 (actual: .270/.341/.500). He even did a nice job nailing down Ryan Doumit's breakout (projected: .280/.354/.484 in 103 games), actual: .318/.357/.501 in 116 games. He did miss on some of the Pirates. His predictions were a bit optimistic for Jack Wilson, didn't see Freddy's drop-off at all (to be fair, it was more severe than even I expected and I did think he'd fall off a bit in 2008), whiffed incredibly on Steve Pearce (.314/.366/.514 was James' projection there), and missed pretty badly on Nate McLouth (.255/.332/.429). That's fine, projecting baseball stats is an inexact science and James himself would tell you that. The goal is to get a feel for what a player's going to do, not write his performance in the coming season in stone.
It is interesting to note that the players that James' system seemed to miss by the most on where the players that did something significantly different in 2008 than they had in the past. This fits James' described model of "We just predict that players will mostly continue to do what they've done in the past." I don't know exactly how he does his projections, but it makes sense that basing things more on the players' past record than that of similar players would make drop-offs and breakouts harder to foresee. Of course, that means it might be hard to get a read on the projections for the one guy that I'm going to be keeping an awfully close eye on this winter: Andy LaRoche.
What, exactly, does an awful season at the age of 24 do to a guy that the computers loved coming in to this year? Last year, James had LaRoche pegged for .265/.367/.458. That would've been a great rookie year. After his 2008 stinkbomb, James' projection for him drops to .252/.347/.401. The projection still loves his plate discipline (74 walks and 83 strikeouts) and sees some home run power (18 dingers), but it basically sees him as Jose Bautista. That would be a huge improvement over what he was this year, but not particularly great as they key return on Jason Bay. It's strange, because the peripherals on the LaRoche projection are quite good. If he hits 18 home runs next year and walks 74 times with 83 strikeouts, I feel like he's going to be putting up numbers we're all pretty happy with. Still, the rate stats that James has him projected for are pretty bad, mostly thanks to the low batting average and not many doubles.
Anyways, keep this all in mind for the future because I'm going to be revisiting the Andy LaRoche projections when PECOTA and ZiPS and CHONE and the Hardball Times all come out with their numbers for 2009 as I try to get a handle on what the Pirates enigmatic (a nice way of saying, "I can't figure out why he can't hit") new third baseman and what we might be able to expect from him next year.
If you're interested in James' book, you can pre-order it here.