Projections Hate Top Hitting Prospects by Jeff Zimmerman May 10, 2018 A week ago, I examined how prospect rankings could add more context to hitter projections. It’s time to take the research a step further by dividing up the prospect list to see if projections can be refined. And they can be. Initially, I shied away from dividing up the prospect lists because the sample size quickly gets into single digits. I started dissecting the data hoping to keep reasonable sample sizes. I sort of achieved my goal. I used the same parameters in the last article. I compared a hitter’s Steamer projected OPS (on-base percentage plus slugging percentage) from 2010 to 2017 to the actual results in their debut season. To designate prospects, I used Baseball America’s top-100 which has been compiled since 1990. I collected the average and median change in OPS. The median value helps to smooth out any major outliers. For the first step, I divided the players into four ranking groups, 1 to 10, 11 to 25, 25 to 50, 51 to 100. From previous studies, the talent drop is not linear with the top players being significantly better. Also, I found these four groups to have a similar number of players since the top prospects are usually closer to the majors. Here the results. Difference in Actual & Projected OPS for Hitting Prospects BA Prospect Ranking Average Median 1 to 10 .032 .026 11 to 25 .030 .058 26 to 50 -.045 -.067 51 to 100 .032 -.015 While the numbers jump around, a break can be seen after the top 25 with the median values. Here the value of just two groups. Difference in Actual & Projected OPS for Hitting Prospects Rank Average OPS Median 1 to 25 .031 .038 26 to 100 .002 -.023 The next step is to incorporate the information from the previous study where I found a player’s previous minor league level had a significant bearing on how far the projections were off. I divided the hitters into six groups. First, I had the hitters who, in the previous season, made it to Double-A and Triple-A. I had to cut out the Single-A group because of a lack of samples. Then I divided these two groups into Ranked 1-25, 26 to 100, and unranked. Here are the results from these six groups. Difference in Actual & Projected OPS for Hitting Prospects In Triple-A Average Median 1 to 25 .013 .027 26 to 100 -.010 -.028 Unranked -.046 -.036 Spread .059 .063 In Double-A Average Median 1 to 25 .108 .139 26 to 100 .053 .112 Unranked .022 .034 Spread .086 .104 While just a small amount of information, some useful nuggets stick out. Concentrating on just the median values, the top-rated prospects outperform their projection by 25 to 50 points of OPS. The next prospect tier outperforms those unranked by 10 to 80 OPS points. The next piece of information is the smaller spread from those who had some Triple-A experience and those who just made it to Double-A. Additional higher level data helps to refine the projections. Finally, all the hitters, ranked or unranked, from Double-A outperformed their projections. Almost certainly these hitters exceeded expectations and push themselves quickly through Triple-A to the majors. If a hitter makes a major jump, owners need give the hitter’s preseason projection quite a boost. The results were what I expected with prospects beating their projections. Now, I just know to what extent. I think I’m done with hitters for now. Overall, I’m trying to think of ways to simply refine projections and I found touted hitting prospects could use a boost. I’m off to find the next area to exploit.