— Where BTF's Members Investigate the Grand Old Game
Tuesday, October 29, 2002
Adjusted Average Age
Paul examines the effects of team age.
I got the idea to analyze team performance in the context of team
First things first, I need to describe my methods. Now, not having the Win Shares numbers in a convenient fashion (and having some philosophical issues with James’ method), I decided to use a different formula to arrive at Team Age. It’s pretty simple and fairly easy to compute quickly (well, with the help of a computer, anyway). For each hitter, multiply his baseball age for that season by his plate appearances. This gives you BatterAgeContribution. Then divide the sum of these for all players who played with a team in that season by that teams’ total PA to get AverageHitterAge. Do the same for pitchers, except to substitute IP for PA. Then, to combine the two, count the number of different hitters and pitchers to appear. The final average for the team is then arrived at by multiplying AvgHitterAge by the number of hitters, and AvgPitcherAge by the number of pitchers, adding the two and dividing by the total number of players. Now, we have what I will refer to as Adjusted Average Age.
So, now with all these Adjusted Average Ages computed for all teams since 1903, I decide to just look around at the numbers. For the course of the 98 seasons (1903-2001 inclusive), the average team age was 29.32 years. The chart that follows is the average age by decade:
The average age has stayed pretty steady across baseball history. It dipped in the 1960s and 1970s, and was pretty low in the teens. I can’t think of any particular reason for this. Perhaps the 1960s and the 1970s were eras predicated around speed, a “young player’s skill”, as opposed to the 90’s and 00’s which are largely power and walks, “old player’s skills”. That doesn’t explain the 1980’s though. Here’s a chart that shows team age movement by year:
There’s a huge dip at the end of the first decade, and a rise in the postwar era that gave the 5 “oldest” seasons in history. I cannot, however, explain the huge drop-off last year. There WAS a influx of young talent playing last year, but I didn’t think it was substantial enough to be reflected in the average team age. There was also a steady climb starting in the late 1970s that pushed the average team age up over 29.5 for most of the 1980s and 1990s. So the next natural question to ask is, “What effect does this have on team performance? Are older teams better? Are younger teams worse?” Well, that’s three questions.
I looked first at the correlations between team age and winning percentage. For the period studied, there was an overall correlation of .213. There’s a slight upward trend to the data, but nothing that jumps out at you on first glance. Here’s another graph, the year to year correlation of team age to winning percentage:
As you can see, it’s a mess. There’s no clear patterns or relationships emerging. The data fluctuates from a .7 correlation (for two consecutive years in the late 1970s) to a -.4 correlation in the mid 1930s. I think it’s more a small sample size issue than anything else (one team can influence the numbers in a drastic fashion). So let’s look at the numbers, in table form, of correlations by decade:
Excepting the 1980’s, there’s a trend in the free agent era towards older teams being better (I should rephrase that, older teams being better than their historical counterparts). Of course, medical science may also have something to do with that phenomenon as well. So now I’m going to focus on extreme teams: the oldest and the youngest teams in history.
First, let’s look at the teams that fall into each category. The oldest and youngest:
As you can see, there’s a lot of expected teams on these lists. The recent Jeffrey Loria debacles rank as some of the youngest, while the “Geezerbacks” of last year rank as one of the oldest. These old teams as a whole posted a .528 winning percentage, while the young teams posted a .431 winning percentage. You can begin to see that having an old team is not a recipe for success, but having a young one is definitely inviting failure. This becomes more apparent when we examine division placement, grouped the young and old teams:
The younger teams routinely finished amongst the weaker teams, whereas the older teams finishes were more evenly distributed. The two “young” first place finishers were the 1970 Cincinnati Reds, who lost in the World Series to the Orioles, and were the 28th youngest team ever; and the 1914 Philadelphia A’s, who were swept out of the WS by the Boston Braves. The Orioles were a full three years “older” than the Reds, the Braves barely missed the youth cutoff themselves by about .05 of a year. The seven old pennant winners are the 1945 Chicago Cubs, who lost to one of the other “old” pennant winners, the 1945 Detroit Tigers in the WS; the 1982 California Angels, who lost to Harvey’s Wallbangers in the ALCS; the 1983 Philadelphia Phillies, the so-called “Wheeze Kids”, who lost to Cal Ripken’s Baltimore team; the 1992 Oakland A’s, the end of their late 80’s/early 90’s run, who fell to the Toronto Blue Jays in the ALCS; the 1997 Baltimore Orioles, also at the apex of a run, went as far as the ALCS; the 1999 New York Yankees, who went all the way to the top; the 2000 edition who also took home the crown; and finally, last year’s WS champion, the Arizona Diamondbacks. So this made me wonder, “Are World Series winners ?older’ than average?”
It turns out that they are. Of the course of all the World Series that have been played so far, the average World Series champion is 29.322 years old, and averages being 0.36 years older than that team’s average, a 1.21% increase. Here’s a table of the “oldest” WS winners, listed by Difference From League Average:
Notice a trend? Three of the top four champions are the last three WS winners, last year’s “Geezerbacks”, who were a whopping 4.1 years “older” than the league average, and the NY Yankees of 1999 and 2000, as all mentioned in the previous section. (For reference, last year’s Yankee team wouldn’t have placed on this list, only creeping in at +.03). Let’s look at the youngest teams to take the top prize:
Both NY Met championships are on here (hint, hint Steve Phillips). One that surprises me is the 1995 Atlanta Braves. That’s probably colored by my perception of them now, for back in 1995 the Three Aces were still in their 20s. The only entries on here in the last twenty years are that team, the surprise 1990 Cincinnati Reds team, the Buckner Mets, and 1982’s St. Louis team that relied on speed, speed, and speed (they hit less HRs as a team that year than Barry Bonds did last year). Since there’s obviously a range of WS winners, let me get back to the original questions: “Do old teams sink? Do young teams rise?”
Taking our young group, they averaged 0.006 improvement on their winning percentages in their next season. So it doesn’t seem like young teams necessarily have to improve. Players may not pan out, injuries, and inconsistent play may do these growing teams in. The older group averaged a loss of 0.025 percentage points in the following season. This is a lot more significant, but still not very significant. It may make the difference between winning your division and some good golfing in October, though. Part of the problem is the old teams already are up high in the standings for the most part, and thus are more likely to fall instead of improve, whereas the young teams tend to be distributed more towards the bottom making it easier to move up. The overall correlation between age and next year’s performance is -0.158, and here’s the numbers by decade:
Note I left 2000 off the chart. There’s only one season pair, so I decided to do without it. Bill James, in the section of Win Shares that I mentioned at the beginning, says he noticed this correlation during the period he was doing his Abstracts. Interestingly enough, the correlation was fairly low in the 1980s, only to swing back higher in the 1990s.
Some conclusions: having a veteran laden team isn’t a recipe for success, but completely lacking any veterans is dangerous for your team’s health, the game is getting older, and last year’s Diamondbacks were truly a “historical” team, in more than one sense of the word.
 James, Win Shares, pg. 233
 James’ methodology multiplies WS by Age. I feel that we shouldn’t be looking at performance as measure of team contribution in this sense; rather, I felt that measures of playing time were more appropriate.
 It turns out my method is similar to the method used by Sean Forman in calculating Team Ages on baseballreference.com. His method differs from mine, in that he keeps hitter and pitcher averages separate, and the formulas used, although the batting formula is very similar, to arrive at these team ages. I was unaware of this method when I started my work. His weighted averages can be found here.
 Yes, pitchers in non-DH leagues are double counted under this method. They are counted for their PAs and IPs. I originally took out their hitting contributions, but I feel more comfortable leaving them in.
 Spotty DOB information caused problems doing calculations for any season before this. Besides, it happily coincides with the first year the World Series was played.
 1900s=1903-1909, 2000s=2000 and 2001.
 Two standard deviations less than the all time average. For older teams it was the top 37 to match the fact that were 37 “young” teams.
 As noted by Clay Davenport in his essay, “Graying the Game”, printed in Baseball Prospectus 2002.