For John Burnson's book, The Graphical Player, I've been working on some extensions of my projection system, MINER. The fundamental system remains the same--it's the Marcel system, but with some obvious additions, such as adjusted minor league stats, park-adjusted MLB stats, batted-ball data, and some slightly more advanced age and position adjustments.
Frankly, I don't know just how good the system is, but it tracks Marcel pretty closely, except for those changes, most of which seem like improvements.
My biggest project has been to do a retroactive monthly projection for every one of 900 players that will be featured in the book. In other words, for every month in a player's major league career, I took the available data up to that point and came up with a projection using only that information.
(Some of this might sound familiar. In-season Marcels have been a focus of Sal Baxamusa's at The Hardball Times, and I'm indebted to Sal both for his published work and a few pointers he's given me behind the scenes.)
Without further ado, I want to share some of the results with you. There are no doubt prettier ways of sharing this data, but I'm saving that for later.
The table below shows month-by-month projections for James Jerry Hardy. The left-hand column is the month -- 200504 is April 2005, and the projections for that month include stats for that month. So the projections in the Apr 05 column take into account Apr 05 and the three years previous. I left out the months he was injured in 2006--his rate-stat projections didn't change during that time.
Month | AVG | OBP | SLG | OPS |
200504 | 0.262 | 0.325 | 0.385 | 0.711 |
200505 | 0.262 | 0.325 | 0.383 | 0.709 |
200506 | 0.258 | 0.324 | 0.380 | 0.706 |
200507 | 0.260 | 0.331 | 0.385 | 0.717 |
200508 | 0.262 | 0.329 | 0.386 | 0.716 |
200509 | 0.269 | 0.334 | 0.405 | 0.741 |
200604 | 0.272 | 0.342 | 0.424 | 0.766 |
200605 | 0.269 | 0.339 | 0.420 | 0.760 |
200704 | 0.275 | 0.344 | 0.447 | 0.791 |
200705 | 0.286 | 0.350 | 0.477 | 0.827 |
200706 | 0.283 | 0.350 | 0.470 | 0.82 |
200707 | 0.282 | 0.347 | 0.455 | 0.802 |
200708 | 0.285 | 0.345 | 0.464 | 0.809 |
200709 | 0.286 | 0.344 | 0.463 | 0.807 |
200804 | 0.278 | 0.338 | 0.443 | 0.782 |
200805 | 0.276 | 0.338 | 0.434 | 0.773 |
200806 | 0.276 | 0.338 | 0.442 | 0.781 |
200807 | 0.283 | 0.342 | 0.461 | 0.804 |
200808 | 0.283 | 0.343 | 0.464 | 0.808 |
200809 | 0.281 | 0.341 | 0.463 | 0.805 |
My favorite part about running projections like these is that it takes the noise out of monthly performance without ignoring its import. For instance, JJ had an absolutely smoking May last year, with a .961 OPS in 28 games. You can see that reflected in his projections.
But unlike some wishful-thinking fans, MINER wouldn't have thought that JJ suddenly turned into Alex Rodriguez. Instead, the projection was never lower than .791 or higher than .827. Sure enough, JJ gave us an .821 performance in 2008.
While JJ's 2008 season had a much different trajectory than his 2007, it's interesting to see that the July/August/Sept projections are almost identical. If that means we'll get a similar season from him in 2009--at whatever position--I'll be a happy Brewers fan.