How important is age relative to league? A quick disagreement over semantics with RZNJ in a different thread made me want to quantify age-advanced prospects. This post is my first attempt to start to get a handle on things.
Frobby, in another thread, provided average ages for the various minor league:
Sally League 21.7
Carolina League 22.5
Eastern League 24.6
International League 27.2
and we might consider a baseline progression for a prospect that would lead to a 25-year-old rookie (very few productive players make their debut after age 25):
21 in Sally (Delmarva)
22 in Carolina (Frederick)
23 in Eastern (Bowie)
24 in Norfolk (International)
25 as a rookie (MLB)
Being as far ahead of this curve as possible is best, and putting up good results at the same time is ideal. That's why Bundy and Machado are such good prospects: Bundy is 19 and spent most of the season doing well at Frederick (3 yrs ahead of baseline); Machado is 19 and spent most of the year at Bowie (4 yrs ahead).
Just to get started on this topic, I went and looked at every position player in the Sally League who was 18 or 19 and got at least 150 plate appearances in a season between 1997 and 2005. 197 players fit this criteria. I recorded their SAL OPS, whether they made the majors, and their career fWAR.
The young Sally Leaguers did better than I was expecting. Overall, 53.8% of them made the majors, and 28.9% provided some value (>1 career fWAR). Almost as many (24.4%) had career fWARs >5, and 12.7% were stars (career fWAR >10). Here's a table, including a breakdown by their age-18 or -19 Sally League OPS:
Unfortunately, the sample sizes for some of the OPS boxes are a bit small, but still large enough to draw some inferences. As we would expect, better performance in the Sally League is correlated perfectly with a higher chance of being a successful major leaguer as well as being a star. I did two quick best-fits that predict the chance of having, respectively, >5 and >10 fWAR based on SAL OPS:
All >.800 .700-.800 .600-.700 <.600
Total# 197 25 62 65 45
Majors% 53.8 84.0 71.0 44.6 26.7
>1 fWAR% 28.9 52.0 43.6 23.1 4.4
>5 fWAR% 24.4 36.0 27.4 13.8 0.0
>10 fWAR% 12.7 32.0 21.0 6.2 0.0
CHANCE(>5) = 121.5*OPS - 65.7
CHANCE(>10) = 110.8*OPS - 62.8
There was little difference in the results when using rWAR instead of fWAR, though most of the WAR values did decrease.
I'm quite surprised at just how good these turned out. Keep in mind that position, other years in the minors, scouting reports, etc., were not taken into account. Just (1) that they were 19 years old or younger, with 150 PAs in the Sally, and (2) their SAL OPS. Nevertheless, simply having an OPS over .700 gave prospects about a 3 in 10 chance of successful major leaguers.
Part of the advantage of this approach is its simplicity, but it's worth noting that breaking things down by position may also be useful. There were four players with >10 OPS in the .600-.700 bracket, and relied of position/defense for value: Jhonny Peralta, Cristian Guzman, Brandon Phillips, and Joe Crede. Perhaps a positional breakdown could help separate these players from .650 OPS 1Bmen.
It's important to remember that the generated odds of success/stardom are empirical, which in this case means that it's just reflective of how other (in this case, similar) players have done in the past. As they say with stocks, past performance is no guarantee of future results, and it's important to temper any given number with additional data.
The success/failure rates of BA Top 100 prospects make a good standard of comparison for any study such as this. As written up at Royals Review, (http://www.royalsreview.com/2011/2/1...-mlb-prospects), BA Top 100 position prospects have a bust rate of about 63%, and are "superior" 22% of the time. It's also worth mentioning that the bust rate is much lower for position players ranked in the top 20 (39%), around 64% for players ranked 21-40, and around 70% for the rest. Their study use per-year fWAR to determine whether a prospect was a bust or success and only looked at cost-controlled years; I used the simpler and quicker method of looking at career WAR. As a result, the percentages are not directly comparable - mine are somewhat more lenient. Nevertheless, for this sort of rough ballpark work, I think we can get some impression of just how strong an indicator age relative to league is.
In the last two years, the Orioles' SAL team (Delmarva Shorebirds) has had six 19 year old position players: Manny Machado (actually, he was 18), Jonathan Schoop, Dudley Leonora, Gabe Lino, Nick Delmonico, and Roderick Bernadina. Leonora only played 20 games and didn't do well. The other five are all interesting.
Machado has been discussed at length, and I'll just note that my predictor gives him (based on his .859 OPS) a 39% chance of success in the majors and a 32% chance of stardom. Most likely, his odds are even higher, since the sample is composed almost entirely of 19 year olds (there were not enough 18 year old Sally Leaguers to separate them out).
Schoop (.890 OPS) also grades out well: 42% chance of success, 36% chance of stardom.
Bernadina is very interesting and extremely promising. His OPS sits at .838, but the sample size is very small, since he spent the first of the season at Aberdeen, and only has 86 PAs. Unfortunately, since the research above is empirical, no exceptions can be made to inclusion or exclusion, since you can't know how many additional cases should have been included in the sample to get a guy with (say) 135 PAs. If Bernadina makes it to 150 PAs by the end of the year (I don't think the Shorebirds have enough games left) with a similar OPS, this empirical method will think he has a fantastic chance to be a good MLB player.
Nick Delmonico has a .762 OPS in 393 PAs, making him excellent to be evaluated with this system. The equations give him a 27% chance of success, with a 22% chance of stardom.
Gabe Lino has a combined .651 OPS in 351 PAs. Success chance: 13%. Star chance: 9%.
I'm really happy with what I've got here. I plan to extend the method to A+ and AA; ideally, the resulting odds could then be combined into a single odds-of-success metric. Though it's very obviously not the be-all, end-all of prospecting, I love the simplicity and quickness of use (once the equations are derived). In the end, I think it has the chance to provide a sort of baseline that can be modified by positional/defense considerations, strikeout or contact issues, and scouting reports.