A transfer doesn't become a success with the first goal or the first victory. When a team acquires a new player, the player typically receives a contract of at least two-and-a-half years -- usually more. Therefore, the choice to make the acquisition doesn't merely involve the question of what the player can do right now, but what he will be able to do in the future. To project performance over several seasons, you need a model of aging.
The age curve is one of the most powerful tools in analytics. It can both generally identify which players a team should target and, more specifically, provides guidance in which types of skills a player may be more or less likely to improve as he gets older.
It is one of the first issues that analysts must study in any sport but successfully modeling aging presents problems. The main one is the problem of the "missing old guys." If you take all of the 26-year-old strikers in the big four leagues (Premier League, Bundesliga, La Liga, Serie A) since 2010-2011, they total about 0.39 non-penalty goals per 90 minutes. If you test the aging curve by comparing that number to all the 34-year-old strikers, they also score 0.39 NPG per 90.
Does this mean that players don't get worse at scoring goals as they age? Of course not. The problem is that when a player loses the ability to play at the highest level, he ceases to get any minutes at the highest level. He drops out of the data set. The only older forwards remaining are the freaks and the greats.
For example, Zlatan Ibrahimovic is still scoring goals in the Premier League but Andy Johnson, who was born in 1981 just like Ibrahimovic, is now retired and has taken a position as a club ambassador for Crystal Palace. If he were still playing in the Premier League, Johnson's numbers would surely be quite bad and drag down his peers' averages. But he isn't playing and so the older players' numbers are falsely inflated.
There's a simple way to fix the "missing old guys" problem and several more complex ways. The simple method provides the most straightforward way of mapping age: You just add up all the minutes played. There are many, many 26-year-olds in the Premier League and very few 34-year-olds because players reach their peak performance levels around age 26.
As this chart shows, players generally peak between the ages of 25-28, but there are differences by position.
Wingers tend to peak and decline the earliest. Wide attackers under the age of 23 play more minutes than U-23 players at other positions, but wide attackers over 30 are much rarer than strikers or center-backs of an equally advanced age. While most of the curves are well on the downslope by 30, the curve for center-backs by contrast doesn't really begin to decline until 31 or 32.
This suggests a general rule. Players tend to get slower and generally lose athleticism as they age, but they also gain skills and know-how to balance that out. Positions that require the most athleticism are a young person's game, whereas older players will more often be found at positions that most prize guile.
(Ranked from the most to the least "age-sensitive" positions: Wide attacking midfielder, central attacking midfielder, full-back, central midfielder, striker, center-back, goalkeeper.)
It's extremely difficult to play midfield in the modern game without peak athletic skills. At striker or center-back, a player who adds skill and guile may hang on for much long as his athleticism declines. One way players can thus buck their aging trends is by moving positions, just as Javier Mascherano extended his career by moving from central midfield to center-back; likewise, Cristiano Ronaldo subtly moved to a "true" striker role after years as a hybrid winger.
Players can do this by developing other skills to make up for what they lose. Different skills, then, have different aging curves, which can help to predict more specifically how player production will evolve over the course of a career.
To identify player skills, we are using a new method that compares a player's performance in one season to his performance in the next. This method avoids most of the problems of the "missing old players" because if a player disappears from the data set in the next season, his previous season is also not included.
One thing this age curve can help with is roster planning for the future. Right now, Chelsea have the Premier League's top scoring striker in Diego Costa, as well as highly-touted prospect Michy Batshuayi in reserve. Batshuayi, now 23, averaged nearly four shot attempts per 90 minutes last season in Ligue 1, while the 28-year-old Costa has averaged a little over three Sh/90 with Chelsea. Costa, however, has been scoring more goals.
Such a pattern of high shot rates with lower conversion at a young age is common with forwards. The aging curve for attacking players shows a tendency for players to take fewer shots as they age, but to increasingly get on the end of better chances and therefore make up the difference. A player's skill at getting quality shots peaks around 26-27 years old.
With good management, Chelsea should be able to avoid the error made by Manchester United in 2013 when they failed to sign a younger replacement for Robin van Persie, then 29 and coming off a title-winning golden boot season.
Van Persie's production declined precipitously over the next two years and United were unable to replace his production until this season with Ibrahimovic. Like Van Persie, Costa may be hitting the end of his peak while Batshuayi should be starting his.
Analyzing the aging patterns of midfielders is more complicated due to the wider variety of roles available and skills necessary in the center of the park.
Liverpool recently signed 28-year-old Adam Lallana to a hefty contract extension that will see him earn £110,000 ($140,000) per week. Lallana has played well in multiple roles for but, over the next few years, his positional flexibility is likely to be limited. He has played an all-action role at times, pressing and harrying opposition ball carriers to possession, but such a role is unlikely to suit Lallana as he moves forward in his career.
Analysis shows that players decline in tackle rate and tackle success rate over their careers. As Colin Trainor has shown, dribbling ability likewise peaks very early and declines over the course of a career. Open field individual contests, whether attacking or defensive, are the moment in the game where a young player's greater athleticism is most useful.
It's expected that midfielders will add other, more subtle skills over their careers to balance out the loss of the speed that allows them to win one-on-one contests in the open field. This is where Lallana's career may be headed, to a more creative, deep-lying role.
It should be noted, however, that just as goal-scoring rates peak in the mid-to-late 20s, so likewise do assist rates. The following chart looks at two measures of creativity, expected assists and the rate of progressive passing, defined as "passes that drive an attacking move forward into space."
While the rate of expected assists falls off quickly after 28, creative players may still be productive members of their teams. A player's rate of progressive, attacking passes increases until 28 years old; even in a player's early 30s, his numbers remain competitive.
Perhaps the best example of a creative player who maintained his performance level by moving to a deeper position and running the show from central midfield is Santi Cazorla. However, the number of paths for an aging midfielder is limited and Liverpool need a high degree of confidence that Lallana can make such a transition if they are to offer him big money through his age-31 season.
The age curve is a powerful tool for analysis because it can not only give a broad sense of when a player may decline but it also can also provide insight into which skills might be least affected by usual physical decline. While shot-taking, tackling and dribbling will likely fall off, more refined skills like creative passing and finding space for big chances can keep a player going past the usual peak of 26 or 27 years old.