So far in this series, we’ve introduced the data we’re examining, then looked at overall performance by ranking and broken the results down between pitchers and hitters. Now, we’re going for something more granular: examining top prospect lists by player age.
Just as it is for major league players, age is critically
important in assessing a prospect’s future. If you have a league average hitter in AA,
that performance is extraordinary if the player is 18, promising if he’s 21, and
completely uninspiring if he’s 24. The younger the player, the more room he has
to develop.
Of course, prospect evaluators already know this. A teenager
who makes any noise at all in even the low minors can rocket up the rankings
very quickly (recent examples would include Wander Franco – ugh – and Ethan
Salas); by comparison, once you’re 25 or so, you’re likely not cracking the top
50 no matter how well you hit at AAA. The question is, are they handling the
age adjustment correctly?
Splitting our 2300-player sample by age is going to wreak havoc
on our sample sizes. There certainly won’t be enough 17- or 27-year-old players
to produce any kind of usable numbers. The age groups I was able to produce with
reasonable data sets up and down the rankings were: 19 and under, 20, 21, 22,
and 23 and over. (Since these are offseason rankings, I used the player’s age
as of January 1 rather than the oft-cited “baseball age” as of June 30; do with
that information what you will.) Here are the player counts in each ranking bucket for each age group:
|
Rank |
19U |
20 |
21 |
22 |
23O |
|
1-10 |
46 |
49 |
63 |
40 |
32 |
|
11-25 |
63 |
71 |
79 |
83 |
49 |
|
26-50 |
98 |
107 |
123 |
111 |
136 |
|
51-75 |
92 |
83 |
129 |
126 |
145 |
|
76-100 |
100 |
77 |
123 |
133 |
142 |
|
Total |
399 |
387 |
517 |
493 |
504 |
You may notice that we’re using fewer ranking buckets for this
check than we did for either the overall rankings or pitchers and hitters. If we
didn’t do that, the samples would just be too small; there were only 12 top-five prospects 23 and over in the sample (basically one every other year). As we go
through the results, it’s worth remembering that even though we’re not seeing more
granular breakdowns near the top of the rankings for these subsets, we have already explored the overall performance of #1 prospects, and they do tend to substantially
outperform #2-5, who in turn moderately outperform #6-10. I don’t see any
reason to think that result wouldn’t be expected to hold here.
Also, we can get some initial results just from this table.
The 23-and-over group has 26% more total top-100 prospects than 19-and-under,
but the teenagers outpace the relative veterans 109-81 in top 25 placements. 20-year-olds
are even more disproportionately represented in the top quartile, as they’ve
had another year to establish themselves as professionals and climb the ranks,
but are still young enough to dream of especially long, productive careers. As you move down,
the pendulum swings toward the older prospects, who likely aren’t going to make
many All-Star teams but may have already displayed a major-league-ready level of
production in the high minors.
On to the actual WAR numbers. Let’s do average first:
|
Rank |
19U |
20 |
21 |
22 |
23O |
|
1-10 |
32.4 |
21.6 |
20.5 |
20.9 |
21.6 |
|
11-25 |
17.5 |
12.3 |
15.5 |
13.8 |
15.1 |
|
26-50 |
12.7 |
13.0 |
10.1 |
11.2 |
10.2 |
|
51-75 |
8.8 |
8.7 |
8.2 |
7.5 |
8.3 |
|
76-100 |
10.2 |
6.6 |
6.8 |
6.3 |
5.9 |
Fairly decisive win for the teenagers here, especially in
the upper echelons; from age 20 and up, I don’t see much to distinguish between the
groups. But as we’ve mentioned before, the average is prone to being skewed by
a couple of outliers, so if (say) Alex Rodriguez spent two years in the top 10
as a teenager, well, it might throw things off a bit. Let’s go to the median:
|
Rank |
19U |
20 |
21 |
22 |
23O |
|
1-10 |
25.3 |
11.2 |
13.5 |
14.9 |
17.9 |
|
11-25 |
9.7 |
6.4 |
6.1 |
9.0 |
10.5 |
|
26-50 |
6.2 |
5.9 |
5.8 |
3.0 |
4.5 |
|
51-75 |
1.2 |
0.9 |
0.9 |
2.2 |
2.8 |
|
76-100 |
0.6 |
0.9 |
0.5 |
0.8 |
1.5 |
The upper-tier teenagers still fare well, but once you get
out of the top 50 you’ll get more reliability with the 24-year-old in AAA who
profiles as a #4 starter or a utility infielder. 23-and-over is actually
treated pretty favorably here, posting the highest median in three of the five ranking groups. This is probably
where I should point out that not all 23-and-overs are created equal, as this
group will include players coming over from places like Japan or Cuba who would
have been in MLB years earlier if afforded the opportunity. That said,
especially in the larger groups, the number of special cases won’t be enough to
skew the results much.
There’s an effect here that is likely expected, but still
nice to see borne out in the results. The teenagers ranked outside the top 50
have comparatively high average outcomes, but comparatively low medians. That is natural for a group that produces more extreme outcomes, both positive and negative. Let’s see if this carries further by looking at
ceilings and floors. Ceiling first – let’s look at the 80th-percentile
outcome for each subgroup:
|
Rank |
19U |
20 |
21 |
22 |
23O |
|
1-10 |
55.2 |
39.9 |
34.3 |
35.7 |
39.4 |
|
11-25 |
29.0 |
22.3 |
34.0 |
26.2 |
28.1 |
|
26-50 |
18.6 |
22.4 |
20.0 |
20.2 |
15.7 |
|
51-75 |
15.5 |
18.8 |
17.2 |
15.2 |
14.8 |
|
76-100 |
15.8 |
11.0 |
13.0 |
10.8 |
11.1 |
Top-10 teenage prospects inspire big dreams, and they should;
that upper-quartile outcome is a Hall of Fame candidate. That being said, as
you move down the list, the high-level results for the youngsters aren’t especially
out of step with their older counterparts. The 23-and-olders are the weakest
group by ceiling, but even that result isn’t crippling.
How about floor? We’ll use 30th percentile here –
I looked at 20th first, but almost no group of prospects has anything like a favorable 20th-percentile outcome. Going up a bit allows for a little daylight to creep
into the results:
|
Rank |
19U |
20 |
21 |
22 |
23O |
|
1-10 |
8.0 |
4.8 |
6.9 |
7.5 |
11.6 |
|
11-25 |
0.4 |
0.7 |
1.4 |
1.0 |
3.5 |
|
26-50 |
0.1 |
0 |
0.6 |
0.1 |
0.4 |
|
51-75 |
0 |
0 |
0 |
0 |
0 |
|
76-100 |
0 |
0 |
-0.2 |
0 |
0 |
Even looking at non-worst case results, the floor is less than inspiring, especially for players outside the top 25. But for the higher ranking groups,
you’ve got a better chance of at least getting some value out of a relative
bust if the player is in the older age bracket, and that’s about what you’d
expect.
That, I think, is a sufficient (arguably excessive) number
of tables for this topic. The results were less obviously decisive than what we
saw with hitters and pitchers, but I feel fairly comfortable with the following
conclusions: BA’s evaluators have generally done a good job adjusting for age
in prospect evaluation, as no single group stands out as consistently overrated
or underrated here. The conventional assessment of prospect age groups (young
ones have higher ceilings, old ones have higher floors) is basically accurate.
And if you want to roll the dice on really impressing a friend, pick a teenage
top-10 prospect and pencil them in for a Hall of Fame induction in 25 years;
there’s something like a 20% chance you’ll be right.
In fact, your odds in that regard are notably better if you
pick a position player, which makes sense based on the results we saw last time.
I’m not doing a full multi-axis breakdown of prospect performance here, but I
did go through manually and pull every top-10 teenage position player from the 23 years we're considering. There were 29 such entries, and 12 of them came from either Hall of
Famers (Adrian Beltre, Ivan Rodriguez, Chipper Jones, Joe Mauer), players who
are on the cusp of being voted in (Andruw Jones), players who are enormously
likely to get in someday (Mike Trout, Bryce Harper), or players who would
easily be in if not for steroid use (Alex Rodriguez). 12/29 is just over 40%.
Careful observers will note that the list above contained
eight distinct players, but 12 top-10 appearances. Four of the players had
multiple top-10 showings as teenagers and are counted twice (as did two
additional non-Hall candidates); if you want to filter out the duplicates, the
result is 8/23, still over a third. But this does bring up a salient point.
Players who spend multiple years in the top 100 have been playing a silent role in our results all along, but
we haven’t spent much effort considering them so far.
Next time, we’ll change that, and try to determine whether the classic old-timey advice applies in this context: Should a member of a top prospect list act like they’ve been here before?
