I get asked how our rankings work all the time, enough that I decided it was time to publish a glance into how we do things around here.
We take a very different approach than most of the prospect rankers in the industry, many of whom rely on scouting reports and gut-level instincs.
While I recognize that good scouting reports contain valuable information that numbers could never begin to recognize, I firmly believe that well-designed statistical analysis programs can also offer helpful evaluations when paired with the proper data.
Stastical projection systems are commonly used in Major League Baseball due to the abundance of data a 162-game season offers. And that's where my ranking program days began – drawing upon multi-year samples of MLB statistics to rank players for my fantasy leagues.
Last summer, I attempted to apply something that was pretty similar to my MLB ranking program to minor leaguers. The results were some of the ugliest prospect rankings I'd ever seen.
Abandoned by my roommates and stuck in my college town for the summer, I spent hundreds of hours trying to improve my program. I crunched numbers for typically ten hours a day, pausing to eat and give my computer time to cool down – my apartment would get so hot that my computer would frequently overheat to the point where it could barely handle browsing the internet.
Months of research turned into rankings that I felt comfortable enough to share with my peers, debate, and publish with many amendments in the form of our Top 100 Prospect List. But this list left me only partially satisfied. I wanted a way to capture a prospect's production-based value in a form that I felt comfortable publishing, unamended.
Upon spending countless more hours this winter sifting through the feedback I received from my peers, co-workers, members of the baseball industry, and emails from readers like you, I crafted a new rankings program.
My weekly Top 25 Rankings are the unamended product of my newest ranking program.
Below is some insight into how the program I created works:
The program is Microsoft Excel-based and uses uses two-season data samples. When less than two seasons are available, I either weight the available data twice as heavily or use data from winter leauges or college (ex. Joba Chamberlain's Hawaii Winter Baseball stats and Ross Detwiler's NCAA stats).
The program that I developed last year used deviations from a "baseline prospect" to determine a rank for each minor leaguer. This resulted in positive scores for players above the baseline prospect and negative scores for prospects below. My current program offers up a score based on how well a prospect does in select metrics. Upper-end scores range from 90-100. The prospects who make my weekly Top 25 Rankings typically score above 85.
My newer program incorporates blends of the following stats:
For pitchers...
Strikeout percentage: (K/TBF)
K/BB
WHIP: (BB+H)/IP
For hitters...
XBH%: (XBH/H)
BB/K
BABIP (Batting average on balls in play; for hard contact and speed, not luck)
I'm considering adding isolated power on the hitters side so that guys who have high XBH percentages because they hit a lot of doubles (ex. James Loney and Dustin Pedroia) don't get an extra edge. Some of my readers have suggested using line drive percentages over extra-base hit percentages, but the line drive data that is currently available to the public doesn't give me all the information I need.
I also employ a factor that I primarly use to make adjustments for age vs. level and draft position, but I'll sometimes use it for extreme league and park factors, too. This is by far the most subjective part of my system. However, for guys with similar backgrounds who are in similar leagues, the playing field is set as even as possible. For example, 2006 1st round high school hitters who are in the South Atlantic League and are approximately the same age would have essentially the same score in this field.
The only advanced mathematics classes I took in college were some calculus classes for non-engineering majors, so I'll never claim to be a mathematics ace. But I think proper indicator stats in a solid blend and sufficient samples can tell a lot about a prospect. And that philosophy is the backbone of all the rankings you'll see here at Project Prospect.
If you have questions that I have not answered above or would like to offer suggestions, feel free to email me at adamf@projectprospect.com.