I’m certainly not claiming modern statistical analysis, but I also don’t think I’m using random numbers. I managed to combine scouting reports, with a set of comparisons and, yes, MLE’s, to help determine what Castro’s potential rates might be in K%. BB% and XBH% (and BABIP). The raw numbers were built from that, using round numbers when possible. I hardly contend this is the right way, but I just wanted to show a really rough draft for what I suggest in the beginning: that I think there represents an untapped area available to better value prospects.

To answer your first question, down the road my hope is that we can come up with accurate WAR intervals for a player’s mean performance over his team-controlled seasons.

]]>For my tastes, I’d like to see something along the lines of the first post, where there’s a distribution of outcomes projected for each controlled season, which would result in a distribution of controlled value (not just a single number and not just a range)… Of course, working out all the necessary probabilities is not trivial, but using another idea posted above (grouping prospects into a few categories/buckets) might allow for a large enough sample size to make some reasonable estimates…

]]>Regardless of our difference of opinion on Castro’s capabilities for this year, I do not think I understand this direction of this article. Want to branch out from MLEs and use scouting reports? And modern statistical analysis? And then you proceed to throw random numbers based essentially on your opinion and plug them into some formulas (which I can do with any forecasting system). As for scouting reports, you take a couple of comments and come up with unscientific numbers. This is not modern statistical analysis.

Want to use expect knowledge (scouts) and statistics? There’s a branch of modern statistics literally used for exactly this: Bayesian statistics. If you’re not familiar, their popularity is fairly recent, and you need a computer to actually do interesting Bayesian analysis (and not Excel).

]]>It might have some value with players that were just drafted, but I’d hate to lump everybody in the same group. Some guys drop due to reasons outside of their control, and the draft is never in order of talent. It’s fine to do a large study — it’s going to work out with a big sample size — but to evaluate at the micro level with that work would be irresponsible.

]]>Is that always true, though? As a Jays fan, I know that the plethora of pitching that was used last season (with a handful of kids jumping directly from AA) was not intentional. Injuries can really throw a wrench in the “prospect development” plan…

Granted, I will concede that in typical cases, injury call-ups tend to be marginal/middling/aging “prospects” already on the 40-man and not the potential studs, but I’m not convinced that all call-ups are always pre-planned.

]]>I haven’t thought about what you thought about though, in regards to maximizing a players value to a team by when you call him up. I’d assume the major change would be in older prospects with same median/ceiling WAR would be valued more than younger prospects, since you’d be controlling the older prospect’s prime years.

Even if a younger, high ceiling prospect has the same “cost controlled median WAR” as an older, lower ceiling prospect, you still go with the younger one because he might go Miguel Cabrera/Hanley Ramirez. And in cases of something like keeping down Heyward, there’s the possibility that that then stalls his development.

]]>