Covering the Front and Back Pages of the Newspaper
March 6, 2005
BASEBALL: Age and Established Win Shares, Revisited
Back in November, I took an initial look at how Established Win Shares Levels correlated to 2004 performance, grouping the results by age. But, the data I was using was the raw EWSL - i.e., actual three-year Win Shares totals at the major league level - not the adjusted figures I had used for making the team-by-team EWSL computations. To refresh your recollection, I had included in those calculations:
1. Adjustments for players who hadn't played any significant major league time in 2001 or earlier, calculating EWSL by ((2002 WS * 2)+(2003 WS * 3))/5;
2. Adjustments for players who hadn't any significant major league time in 2002 or earlier, just slotting in their 2003 WS; and
3. Arbitrary WS totals for 2004 rookies: 10 WS for players projected to have everyday jobs, 5 for projected rotation starters, 3 for projected bench players, and 2 for projected relief pitchers.
If I'm going to add an age adjustment to EWSL, I would presumably want to do it after including these adjustments for the limited data - and so, to evaluate the accuracy of last year's figures, I recalculated them based on the adjusted figures. Here's the new chart:
As you can see, the improvement by young players is considerably less dramatic if you adjust for the fact that many of them don't have major league track records (in an ideal world, I'd use MLEs of minor league Win Shares, but I don't think those exist anywhere). Also, I broke out the 20- and 21-year-olds, since the one 20-year-old I listed last year was Edwin Jackson, who I had mistakenly thought was ahead in the race for a rotation job, so his lack of playing time in 2004 doesn't really say much about the method itself.
Will I use these factors for age adjustments? I think this year I will - as well as publishing the raw and pre-age-adjusted EWSL figures - and just tweak the figures over time, but I'm going to think long and hard about the age-35 number. I don't think it's actually realistic to project players to lose half their value at 35, even if that's exactly what happened to a sample of 21 players last year.
What, specifically, about the arbitrary adjustments? There were 6 rookies who I had penciled in as regulars (average age: 25), and gave them an arbitrary 10 WS. In fact, led by Khalil Greene, the six averaged 12 WS (74 WS for 6 players), reflecting the high quality of player who gets handed a regular job as a rookie in spring training (the other five were Kaz Matsui, Bobby Crosby, Aaron Miles, Adam LaRoche, and Joe Mauer).
Then there's the rookie bench players, given 3 WS in last year's system. There were also six of them (average age: 27), and they averaged 5 WS in 2004 (28 for 6), mostly due to Termel Sledge (15) and Jose Castillo (8) snagging regular jobs by year's end. But the number is 7 (26/4) if you leave out the two 30-year-old bench players, Kit Pellow and Cody McKay. The emergence to regular jobs of a few guys is relatively representative of bench players, so for now I'll up the projection to 6 for bench players who are under 30, and leave it at 3 for 30-and-up minor league veterans.
There were just three projected rotation starters given a 5 WS write-in: Jackson, Matt Riley, and Tyler Yates (average age: 23). Riley and Yates were disasters, and the three compiled the grand total of 1 WS. That's a small sample size, but this year I will cut the projection to 4 out of a desire to avoid over-projecting young pitchers.
The relief pitchers, on the other hand, fared well - there were four of them (average age: 26), and they averaged 5 Win Shares (21 for 4 players), led by Akinori Otsuka. Not every season produces a 32-year-old Japanese reliever, so I'll just harmonize the rookie pitcher numbers by dishing out 4 WS for all rookie pitchers.
Anyway, EWSL will never be a true projection system, as opposed to just a systematic way of analyzing past performance. But I think tweaking the adjustments based on the first year's experience should make it a little more useful in evaluating where teams stand in terms of the available talent in 2005. With my look back at the 2004 results wrapping up, I should be ready to start running the 2005 numbers shortly.
UPDATE: Yes, there are double-counting issues with the arbitrary plug-ins and the age adjustments, so going forward I don't intend to apply age adjustments to rookies.