Slices of the Blame Pie

The Needler

lurker
Dec 7, 2016
1,651
It would be amazing to do similar age curves with statcast data. I'm not sure if 4-5 seasons is enough to do this kind of analysis, but in a few seasons they'll start to have entire cohorts that have entered and left their prime completely within the statcast era.
Five seasons should be plenty. You'd have hundreds of data points for each two-year set.

For fun, I looked at the top 20 of the StatCast sprint leaders from 2017, to see how those with elite speed fared the following two years (this is just a fun exercise; not intended to be representative, and understanding it is a tiny sample):



The overall average speed dropped in both subsequent years. Every player but one (Trea Turner) was unable to maintain his first year speed into his last, and only one player recorded a higher average speed in either year over his first (Harrison Bader).

In terms of speed lost by age, we're obviously working with a super small sample. But here they are, anyway. We have six two-year periods with at least 3 players in the group, and those results are:

Age 21-22 (3 players): -.43 ft/sec (I wouldn't put any stock in this sample, which is skewed by Barreto, who lost a full foot for whatever reason.)

Age 23-24 (4 players): -.175 ft/sec

Age 24-25 (5 players): -.32 ft/sec

Age 25-26 (5 players): -.2 ft/sec

Age 27-28 (3 players): -.233 ft/sec

Age 28-29 (3 players): -.434 ft/sec

While this certainly doesn't prove anything, the mid-20s numbers are right in line with the general observation that on average, a player will lose about .2 ft/second per year at that age, and suggests that the heuristic applies even to elite speedsters.
 

Pitt the Elder

Member
SoSH Member
Sep 7, 2013
600
It turns out you can actually download the CSV files of sprint speed for all players for the last 5 years. This isn't perfect, but if you average sprint speed by age for all players from 2015-2019, here is the breakdown:

AgeSprint SpeedDeltaCount
2028.46-5
2128.230.2315
2227.950.2842
2327.820.1383
2427.800.02164
2527.710.09214
2627.550.16234
2727.340.21239
2827.160.18204
2927.010.15192
3026.720.29153
3126.610.11129
3226.510.10128
3326.250.26106
3426.150.1090
3526.000.1668
3625.630.3743
3725.600.0320
3825.530.0714
3925.020.519
4025.030.004
4123.751.282
4327.80-4.051
4427.100.701
4527.30-0.201

If you discount what is almost certainly Ichiro completely skewing the results for the 43-45 age range (and wow, he really maintained decent speed well into his 40s), it looks like a pretty linear relationship, just like The Needler suggested. It would be more useful to see what the average age curve is for individual players, but you have to imagine it looks pretty similar to this.

Edit: I updated the table

Edit 2: I think it's important to note that this data shows that the *population* of MLB gets slower as age increases. It's not clear that players get slower at the same rate as the population as a whole. A couple of observations and thoughts:
  • I think it's fair to explore how speed/defense as matures as a skill relative to hitting, especially power. A speed/defense-first player is probably at his peak value at a younger age than a hitting/power-first player, which means that they likely enter the league at a younger age. It also means that they probably leave the league at a younger age once that skill begins to erode and they get more expensive. Easier to just replace you with the next speed/defense-first player.
  • The number of players per age cohort peaks in the age 26-27 seasons, which makes sense given that players tend to peak at that age. The question, though, is which type of player survives in the league beyond this point?
  • Without looking at player-specific age curves, it's hard to tease out how much of this is age-related decline and how much of this is survivorship bias.
 
Last edited:

Pitt the Elder

Member
SoSH Member
Sep 7, 2013
600
This isn't a perfect analysis but here is Betts and a few other comps:

Betts:
Age Sprint SpeedHP to 1stBoltsAge Rank% Rank
2328.715590.4
2427.9101975.4
2527.94.2162673.7
2628.14.1862676.4
2727.94.312772.6


McCutchen:
Age Sprint SpeedHP to 1stBoltsAge Rank% Rank
2928.57587.5
3029.19294.6
3128.84.319290.2
3228.74.3213287.9
3328.84.323290.6


Trout:
Age Sprint SpeedHP to 1stBoltsAge Rank% Rank
2429.674298.1
2528.917693.8
2629.34.2522396.3
2729.24.2537394.2
2829.34.2823395


Springer:
Age Sprint SpeedHP to 1stBoltsAge Rank% Rank
2628.816692.2
2728.4151185.7
28284.3141675
2927.74.3141966.2
3028.24.30479.5


First, apart from his rookie season, Mookie was never an *elite* speed guy, though he's always had good speed (aside: my God, is there anything Trout isn't amazing at?). Second, despite the data I included in my previous posts, I don't think it's inevitable that every player gets significantly slower with each passing years. Clearly, some guys can maintain near-peak speed well into their late 20s and early 30s. McCutchen is still among the league's fastest players at age 33, despite his overall game degrading. Also, there is year-over-year variability that is somewhat independent from longer-term trends. For example, Springer had a steady 4-year decline from age 26 through 29, only to bounce back with more apparent speed at age 30. Who knows why this happened (was hurt for part of the season, so maybe he's more rested) but I think we should be careful not to see one-year changes from 2018-2019 for Mookie as a macro age-related decline.
 

The Needler

lurker
Dec 7, 2016
1,651
Also, as I mentioned before, statcast sprint speed tells only one part of the speed story—one second peak speed. It does not take into account how long it took to get there, or how quickly the player slowed down, i.e., what the player’s two second peak was.

So while the sprint speed numbers might tell you McCutchen hasn’t lost much speed, his home to first numbers are below league average, and his outfield burst numbers are poor, and declining at almost .2 ft/second each year. Unfortunately, Mookie’s first and second 1.5 seconds (reaction and burst) in the OF have declined at an even greater rate in each of the last two years.
 

Plympton91

bubble burster
SoSH Member
Oct 19, 2008
11,863
what Is the mechanical precision of StatCast in measuring things like foot speed and jumps?

How is something determined to be a “burst”?

Who controls the StatCast data, and what are their financial incentives?
 

The Needler

lurker
Dec 7, 2016
1,651
Burst measures the feet traveled in the second 1.5 seconds after a ball is hit.

I’ll leave others to engage you in your tinfoil hat suggestions if they choose.
 

Plympton91

bubble burster
SoSH Member
Oct 19, 2008
11,863
Burst measures the feet traveled in the second 1.5 seconds after a ball is hit.

I’ll leave others to engage you in your tinfoil hat suggestions if they choose.
When people are discussing differences of 2 one-hundredths of a second in an exercise requiring the identification of two discrete events to get to that difference requires a “tin foil hat” to ask whether that difference is even within the machine’s tolerance.

Keep in mind, the same people pushing the miracle of this technology simultaneously argue that it is not accurate enough to call balls and strikes.

Those seem incongruous to me, unless there are considerations other than the equipment’s reliability driving the discussions.
 

Max Power

thai good. you like shirt?
SoSH Member
Jul 20, 2005
3,577
Boston, MA
I think we can assign 100% of the blame to Adrian's Dome. He said anything less than 100 wins was unthinkable. Maybe he has some inside information about the team getting ready to pull off a 33-3 run.
 

bosockboy

Member
SoSH Member
Jul 15, 2005
10,611
St. Louis, MO
I think we can assign 100% of the blame to Adrian's Dome. He said anything less than 100 wins was unthinkable. Maybe he has some inside information about the team getting ready to pull off a 33-3 run.
I think we can assign 100% of the blame to Adrian's Dome. He said anything less than 100 wins was unthinkable. Maybe he has some inside information about the team getting ready to pull off a 33-3 run.
Yes. I was strung up for suggesting 95 wins was our baseline, which we would gladly take now.