How Old Is Too Old?
I’ve recently been contemplating a trade opportunity that would net me Nelson Cruz. I don’t need to do it, I’m well behind the first place team who seems poised to take his fifth first place finish in six years. I seem to be locked in a two-man battle for second. Without Cruz, I’m confident that my team will eventually pull away from my competitor. At the same point, not only is Cruz’s production this year of interest to me, he shows no real signs of slowing down and might produce a few more years of good power production for my team. This piqued my interest to determine how projectable decline is for players in their mid-30s.
I had previously compiled stats for all batting average qualifiers from 2008 through 2015 so I simply made a few alterations to measure increase or decrease in power production – home runs and RBI – and ran a few regressions.
My first effort had a sample size of almost 1500, that is, I had 1500 player-seasons to compare with the season before it. There appears to be a very slight negative correlation, however the ridiculously small r-square suggests that this attempt to explain power decline failed.
I tried again, but only considered players age-30 and up.
Once again, I found a very slight negative correlation but no evidence that age really had anything to do with power production. So I reset my starting point to age-34 and up.
This time there was a bit more of a pronounced decline as the players aged from their mid-30’s. However, once again, the linear equation was a poor fit for the data. Despite this, I could visually see that range of each age group appeared to be decreasing along with the median, although to be fair, it’s hard to visually hundreds of points on top of each other.
My next step was to average each each age group, to counter the weight that many more players were exerting on my trendline.
Finally, here’s a result that makes sense. According to this regression equation, each additional year of aging reduces power between .02 and .03 standard deviations. That’s not particularly enlightening, as it’s such a small amount but the fact that the model explains about 40% of the variability is reassuring. Lastly, I decided to do a statistical “no-no” and remove the one data point that seems to be an extreme outlier. There’s no good reason to accept that 39-year olds will generally recover a large portion of their early career power numbers. It’s more likely that there are many fewer 39-year olds playing and receiving a qualifying number of at-bats. I’m guessing it was David Ortiz who skewed my chart but I didn’t check for sure.
The results finally match what I had expected. There is a slight but pronounced downward trend in power production that seems to really begin around 30-years old. This equation explains 64% of the variation around the mean which is not bad at all. The slope is of far more interest than the entire equation as it begins at age 24 rather than 0, which is what an equation actually portrays.
In summary, on average, a player begins to lose value at around age-30. The decline becomes more pronounced at age-35. That’s pretty much what most people would have assumed. Going back to 2011, here is a list of age-30 players who hit over 20 home runs: Jose Bautista, Curtis Granderson, Josh Hamilton, Nick Swisher, Nelson Cruz. Two of those players are still elite, one is usable, and two are finished.
Here is a list of current age-30 players who provide value in power and their projections for 2020.
|2015 Home Run Value
|Proj 2020 Home Run Value
It’s not hard to imagine Josh Donaldson, Todd Frazier, and Yoenis Cespedes providing value late into their 30’s while others, such as Mark Trumbo and Chris Davis, seem like they might not age as well. The final point is this, if you’re in the process of acquiring talent in a dynasty league, don’t be afraid of age-30 players. It’s a psychological cutoff, much like $.99 compared to the next whole dollar amount. You can reasonably expect 5 or so more productive years out of them and that’s long enough in any format. And to answer the initial question, I decided not to acquire Nelson Cruz at the cost of Rafael Devers. There are no signs that he’s about to statistically fall of a cliff, but he’s at that age where it could happen quickly and with little warning.