The 'Retrospective' will be a regular posting where we will attempt to archive work that has previosuly been done, poke fun at things we have said in the past, and try to help the reader understand where we have been so that they can get a better understanding of where we are going. This week we will take a look at how we got involved in analyzing prospects.
I have been ‘following’ Minor League baseball for over 30 years now, since my next door neighbor growing up was drafted by the Expos, put together a solid Minor league career, but never made it out of AAA. About 12 years ago, I took a position that required me to travel 3-4 days per week throughout the country. I was sitting in my hotel room one Spring evening, getting ready for my fantasy draft, when I came across an article written by Tony Blengino (now serving as special assistant to the General Manager with the Mariners after a successful stint as assistant director of Minor League scouting with the Brewers, but at that time he was working with Ron Shandler—doing most of his Minor League evaluations and writing Future Stars). The article was about using Standard Deviations to League Averages to evaluate AAA/AA Minor League Players and uncover positive outliers to predict positive future performance.
Now remember in 1997, the Internet was in its relative infancy and Minor League Data in electronic format was virtually non-existent. However, I was able to poke around on the 'net' and get my hands on AAA data from the previous season in a very crude text format and was able to convert it into an Excel Spreadsheet and began to play around with it. I worked late into the night (early into the morning) and was amazed at how, by using the methods Blengino described, I was able to quickly distill a league’s worth of data into easily comparative numbers. Now understand at the time I was the Vice President of Field Operations for an IT Consulting Company and my background was in IT and Finance with a specialty skill set in developing data modeling systems for a wide variety of real-life, mostly financial, systems. This was a natural union between, both, my interest (at the time I was getting to see about 60 or so Minor League games each year while I traveled around the country for work) and my skill set.
The biggest problem was how to ‘measure’ the effectiveness of the concepts, as there was very little test data available in electronic format. So I dug out my copy of the 1992 Minor League Digest and began inputting, by hand over the next several months, all of the 1991 Eastern League data. I then assigned numerical ratings, as to the 1997 perceived value, of those players that I had data on from 1991. From that, I was able to draw some initial correlations as to what statistics appeared to be relevant in predicting future performance and how relevant were they. Beginning in 1998, I started applying those concepts and began developing my first ‘prospect’ lists. I began reading, lots of reading, anything I could get my hands on about scouting and baseball statistical analysis, that took me in various other directions. In 2001, I did a significant study on the effects of aging on professional players and I figured out a way to use those results, along with Major League Equivalencies, to develop what I called ‘Peak Peformance Values’…essentially what was the expected future ‘ceiling’ of young Minor League players. As electronic data became more readily available, I was able to do more detailed analysis, revise the correlating effects of various statistics to more accurate levels, and finally to go back to my original data sets, that were now further removed from the players’ minor league performance, and get better results.
Around this time I began writing for a site (http://www.topprospectalert.com/) as their senior writer. This got me a bit of notoriety, a few regular appearances on a couple of sports talk radio shows, and a rather time consuming hobby. Like so often happens, ‘real life’ got in the way of my ‘hobby’ and in late 2003, I put away my hobby for a couple of years. When I finally got back to it, I had a different job and I only traveled to two cities with any regularity so my actual first hand observations of the players dropped from 60+ games per year to a mere dozen or so, primarily in the Midwest League. The positive though was that now historical minor league data was more readily available and I was able to amass forty years or so worth of data that changed the dynamic of what I was doing. The other major change was that I moved from a mindset that Player X will eventually perform at Y Level to one now where Player X has the probability Y to perform at Z Level. In other words, I realized that I can’t predict with any certainty what level of performance Jaff Decker will play at in five years, but I can tell you extremely accurately how many players, of say 10 players like Jaff Decker, will perform at what various levels. This was a tremendous breakthrough, akin to going to the horse races and knowing that, while I may not know the exact winners in each race, if I systematically distribute $2000 worth of bets over 10 races that I will walk home with approximately $2,240 plus or minus $60.
Of course this 'brand' of prospect analysis draws contempt from the diehard scouting community. Trust me, this is their fight not ours, and in another article soon I will explain Diamond Future's thoughts on the subject, but suffice it to say I find first-hand observation extremely useful, BUT, even watching 60+ games per year as I used to do, I was only able to see, firsthand, about 100 or so prospects play each year. Sure I developed contacts that I trusted what they were seeing, but without a significant staff/budget you just can’t see all of the players you would need to see--noone can. If I had that staff scouring the country, trained the same way, looking for the same things, compiling the same formatted reports, I could do tremendous things with that kind of dataset of information. Reality doesn’t allow for that, so what I have done instead is find things that allow me to approximate the information that that team of scouts would provide me--things like how much money was the player offered to originally sign or what round a player was drafted, projectability based on what age a player was signed, how tall are they, and how is their body-mass distributed, etc. I can take these inputs and ‘systemize’ them in a way that produces measurable results. Maybe more importantly, I found that when I actually did see players, there was a tendency to personalize certain players, both good and bad, in a way that I am sure didn’t add to my accuracy. This way I don’t have to guard against this personalization nearly so much. Maybe I don’t have a radar gun reading on each pitcher, or truly know that his fastball is straight as an arrow with no late life--both which would be helpful if I could consistently get that kind of information, but I do have other information that tells me other things that I need to know, that I know not only that it actually works, but how well. I still have a few contacts, both inside the game and out, that I can ask certain questions of or alert me to things, but the overwhelming majority of what I do is based on the numbers, and things that I can accurately measure. Hopefully, over time, you will come to appreciate the value of what this provides.
Now remember in 1997, the Internet was in its relative infancy and Minor League Data in electronic format was virtually non-existent. However, I was able to poke around on the 'net' and get my hands on AAA data from the previous season in a very crude text format and was able to convert it into an Excel Spreadsheet and began to play around with it. I worked late into the night (early into the morning) and was amazed at how, by using the methods Blengino described, I was able to quickly distill a league’s worth of data into easily comparative numbers. Now understand at the time I was the Vice President of Field Operations for an IT Consulting Company and my background was in IT and Finance with a specialty skill set in developing data modeling systems for a wide variety of real-life, mostly financial, systems. This was a natural union between, both, my interest (at the time I was getting to see about 60 or so Minor League games each year while I traveled around the country for work) and my skill set.
The biggest problem was how to ‘measure’ the effectiveness of the concepts, as there was very little test data available in electronic format. So I dug out my copy of the 1992 Minor League Digest and began inputting, by hand over the next several months, all of the 1991 Eastern League data. I then assigned numerical ratings, as to the 1997 perceived value, of those players that I had data on from 1991. From that, I was able to draw some initial correlations as to what statistics appeared to be relevant in predicting future performance and how relevant were they. Beginning in 1998, I started applying those concepts and began developing my first ‘prospect’ lists. I began reading, lots of reading, anything I could get my hands on about scouting and baseball statistical analysis, that took me in various other directions. In 2001, I did a significant study on the effects of aging on professional players and I figured out a way to use those results, along with Major League Equivalencies, to develop what I called ‘Peak Peformance Values’…essentially what was the expected future ‘ceiling’ of young Minor League players. As electronic data became more readily available, I was able to do more detailed analysis, revise the correlating effects of various statistics to more accurate levels, and finally to go back to my original data sets, that were now further removed from the players’ minor league performance, and get better results.
Around this time I began writing for a site (http://www.topprospectalert.com/) as their senior writer. This got me a bit of notoriety, a few regular appearances on a couple of sports talk radio shows, and a rather time consuming hobby. Like so often happens, ‘real life’ got in the way of my ‘hobby’ and in late 2003, I put away my hobby for a couple of years. When I finally got back to it, I had a different job and I only traveled to two cities with any regularity so my actual first hand observations of the players dropped from 60+ games per year to a mere dozen or so, primarily in the Midwest League. The positive though was that now historical minor league data was more readily available and I was able to amass forty years or so worth of data that changed the dynamic of what I was doing. The other major change was that I moved from a mindset that Player X will eventually perform at Y Level to one now where Player X has the probability Y to perform at Z Level. In other words, I realized that I can’t predict with any certainty what level of performance Jaff Decker will play at in five years, but I can tell you extremely accurately how many players, of say 10 players like Jaff Decker, will perform at what various levels. This was a tremendous breakthrough, akin to going to the horse races and knowing that, while I may not know the exact winners in each race, if I systematically distribute $2000 worth of bets over 10 races that I will walk home with approximately $2,240 plus or minus $60.
Of course this 'brand' of prospect analysis draws contempt from the diehard scouting community. Trust me, this is their fight not ours, and in another article soon I will explain Diamond Future's thoughts on the subject, but suffice it to say I find first-hand observation extremely useful, BUT, even watching 60+ games per year as I used to do, I was only able to see, firsthand, about 100 or so prospects play each year. Sure I developed contacts that I trusted what they were seeing, but without a significant staff/budget you just can’t see all of the players you would need to see--noone can. If I had that staff scouring the country, trained the same way, looking for the same things, compiling the same formatted reports, I could do tremendous things with that kind of dataset of information. Reality doesn’t allow for that, so what I have done instead is find things that allow me to approximate the information that that team of scouts would provide me--things like how much money was the player offered to originally sign or what round a player was drafted, projectability based on what age a player was signed, how tall are they, and how is their body-mass distributed, etc. I can take these inputs and ‘systemize’ them in a way that produces measurable results. Maybe more importantly, I found that when I actually did see players, there was a tendency to personalize certain players, both good and bad, in a way that I am sure didn’t add to my accuracy. This way I don’t have to guard against this personalization nearly so much. Maybe I don’t have a radar gun reading on each pitcher, or truly know that his fastball is straight as an arrow with no late life--both which would be helpful if I could consistently get that kind of information, but I do have other information that tells me other things that I need to know, that I know not only that it actually works, but how well. I still have a few contacts, both inside the game and out, that I can ask certain questions of or alert me to things, but the overwhelming majority of what I do is based on the numbers, and things that I can accurately measure. Hopefully, over time, you will come to appreciate the value of what this provides.
Posts
No comments:
Post a Comment