In a previous article, Ian York looked at three players on the Red Sox whose debut ages made them outliers. Mookie Betts and Xander Bogaerts became full-time major-leaguers at 21 years old. Daniel Nava made the big leagues at 27 years old but still managed to put together three (and counting) solid seasons. Ian looks to see if there is anything that can be predicted about their career trajectories from their unusual beginnings. In particular, he looks to predict a MLB player’s offensive peak and decline based upon previous players through baseball history.
Let’s start with the simplest possible look. (Warning: We will have to look harder than this to see strong trends.) We can calculate OPS and its components (SLG and OBP) for players of different ages, considering every 200-plate appearance season since 1950 (the ridiculous outliers in the age 36 through 39 years are all, of course, Barry Bonds):
To get the technical notes out of the way: The data were all generated using Lahman’s baseball database. We used the player’s age as of June 15 of each season as his “age”. These boxplots show the mean value (the line in the middle of each box); the quartile range is shown by the box (that is, the box contains the values from the 25th percentile to the 75th percentile); the whiskers indicate the 5th to 95th percentiles; and the fliers (the little diamonds) indicate outliers that are outside those ranges.
In any case, this plot does not really show much of a trend: The averages improve a little for players between 19 and 22, but then they look pretty much constant for all ages up to about 36. This makes sense because OPS numbers do not keep going down; below a certain point, the player is simply out of baseball. A marginal player may get two or three seasons to improve, but not much more. That means that most of the players who leave baseball are in the bottom part of the averages, balancing out the declining performance of the aging stars and superstars, so that the overall numbers stay fairly flat for many years.
What we are really interested in is when players reach their own personal peak offensive value, whether that is a Hank Aaron-like peak OPS of 1.079 or a Mario Mendoza-like .596. When we express players’ offensive numbers as a percent of their peak, trends for age become obvious. In this chart, I only included players who met the 200 PA cutoff for at least five major-league seasons. Because we are looking at percent of peak performance, a player who only lasts for one season will by definition reach his peak performance in that year. This actually does not make much difference to the age vs performance plot, but would dramatically skew the next plot we will look at, performance vs. seasons in the big leagues, so for consistency we will use the same dataset here. This time, a trend is obvious (the two players with over 200 plate appearances at the age of 18 are Robin Yount and Ed Kranepool):
Player offense peaks at age 26, driven both by slugging and on-base percentages. SLG ramps up and drops off faster than the “old man’s skill” of OBP: An average 36-year-old player, ten years past his (average) peak, retains about 87.5% of his peak OBP, but just 80% of his peak SLG. Players who reach the big leagues at less than 22 years old almost always improve their offense, often significantly. If Betts and Bogaerts are completely average 21-year-olds, they are hitting about 81.5% and 86.8% of their future peak SLG and OBP, respectively. (Again, this particular chart only shows player who lasted 5 years or more, but the same trend for very young players is still obvious even when we include players who only lasted a single season.)
Note that a low percent of peak does not necessarily mean a poor performance, though it usually is. For example, because Bonds’ peak was so late, and so spectacular, his age 23 year is only about 60% of his peak, even though he hit a perfectly satisfactory .368/.491/.859 that season.
This chart tells us something else important about the predictive power of age. There is a wide variation at every point, meaning that while there are clear trends for the overall population, individual players could plausibly end up all over the place. Several players peaked when they were 19 or 20 years old; several were only at 70% of their peak at the age of 26. We can say what Betts would do if he were an average 21-year-old, but there are many other outcomes that are almost equally probable.
Is the improvement we see due to a player getting older, or is it because of his major-league experience? Obviously, the two are closely, but not perfectly, linked. If we ask about performance vs number of seasons in the majors, instead of age, we still see the trend toward improvement over time, but it is not as dramatic as if we only look at age:
Players tend to peak about their third or fourth season in the majors, followed by a brief plateau, and a gradual decline from their seventh season on. In other words, since the average debut is in their 23rd or 24th year of age, they peak around their 26th or 27th year, as we have already seen. The fact that the trend is stronger when plotted against age than seasons, suggests that the more important effect actually is age rather than major-league experience. (There are more rigorous ways to pull apart the effect of age and experience, but these methods are not necessary for the questions we have here.)
Now we can look at our two outlier groups. Here is the chart for players who debuted at 21 years or less:
Players who reach the majors at 21 years or younger tend to play their first season at 75% to 80% of their eventual peak level. Again, we see these young players improving over their first four seasons and then plateauing, but these players seem to hold close to their peak levels for a significantly longer time: Their decline does not start in earnest until after their tenth season. That’s roughly the same age (about 31) as for the overall player population, but since these players both start younger and reach peak levels somewhat younger than average, they tend to contribute at high levels for longer than the average player.
Players who, like Nava, debut at 27 or more tended to be one-and-done players, even more so than the average. However, some of them did last long enough to make the cutoff of 5 seasons with 200 or more PA:
These players show a very different pattern from those who debuted at 21. Many of these older players (who include a number of Japanese players like Ichiro Suzuki and Hideki Matsui with significant experience in the Japanese major leagues) debuted at or near their offensive peak, with their decline beginning after their fourth season at, once more, approximately 31 years old. Only nine of these players since 1950 have had significant playing time of ten or more seasons after their debut in the majors. It may be realistic to hope for Nava to maintain his offensive effectiveness for a few more seasons, but a long career seems unlikely.
Finally, are the overall patterns we see constant, or has the effect of age on peak performance changed since 1950? To look at this, we split the players since 1950 into four groups: Players who made their debut in the majors from 1950-1965, 1965-1980, 1980-1995, and 1995-2000. (Why is the last “era” shortened? Since we are asking about durability, we need to give time for careers to be completed so that we do not skew the analysis by looking at players who have not yet reached their peak, so we stopped with players who debuted in 2000.)
There was little or no effect on OBP; players from each of the three eras showed a very similar pattern of OBP improvement and decline, so I will not show OBP charts here. However, SLG in these eras showed different patterns. I will show them as simple line plots here to make the trends easier to see, but remember that each of these points has significant variability that is not being shown:
Players who made their debut between 1950 and 1965 (and whose careers therefore continued on into the 1980s in some cases) peaked in SLG at about 25 years of age, and showed the fastest decline in SLG. Those who reached the majors between 1965 and 1980 showed a slightly later peak, at 26 years, and a more gradual decline. Players who started their career between 1980 and 1995, however, showed a much longer peak with very little change from about 26 through 31. Of course, many of these players would have been playing in the heart of the steroid era, from the early 1990s through 2005, giving at least some of them the opportunity to artificially prolong their careers.
The cause of the more gradual decline in peak performance in players reaching the big leagues between 1965 and 1980, compared to the earlier cohort, is less clear to me. Of the 341 players in this group, only 20 lasted into the steroid era, and few if any of that group (which includes names like Rickey Henderson, Tim Raines, and Carlton Fisk) have been linked to PED. The most likely explanation is that players in this period saw the dawn of free agency in 1975 and the subsequent explosion in baseball salaries, meaning that these players had both the opportunity and the incentive to be health-conscious, with weight training and off-season fitness programs contributing to player longevity.
Some of the players who started their careers between 1995 and 2000 also would have had the chance of artificially enhancing their careers. However, hopefully, after about 2005 (when these players had played between 5 and 10 seasons) the introduction of PED testing would render this less of an effect. These players seem to show two peaks of offensive value, one at about 26 and one at 28 years of age. Though this is probably just sample noise, it is possible that it reflects the residual effects of the steroid era followed by the ten years or so in which PED use has been reduced, and career arcs are now reverting back to a more normal pattern.