How does the Prospect League make hitters better?
By Owen Crandall
Introduction
How well does the Prospect League develop hitters? In other words, is there a discernible difference between the average performances of hitters before and after they play in the league? The Prospect League’s primary goal from a player standpoint is to provide a place for college athletes to stay in shape throughout the summer, continue to play baseball, and further develop their games. Therefore, it would be assumed that most hitters who spend their summers in the league should show a statistical improvement in their college performance.
Process
For the purposes of this paper, the Prospect League will often be abbreviated as PL, and Baseball Reference will often be abbreviated as BR.
To answer this question, I needed a list of all the hitters that have played in the PL during a certain time frame, as well as their NCAA statistics from the seasons before and after their summer in the league. The PL statistical archives contain the player lists I am looking for, as well as their PL stats. However, they obviously do not contain NCAA statistics, which is what I am looking for. Luckily, Baseball Reference (baseball-reference.com) does contain this information. It also contains PL statistics, although slightly less than what can be found in the PL archives. I could’ve pulled the player list from the PL archives and used that list to pull the NCAA stats from BR. However, the player naming convention differs between the two sites (PL uses F Last, while BR uses First Last). This would’ve made it difficult to pull the correct data. It made more sense to pull the player list from baseball reference and then pull their stats from the same site.
BR has PL stats going back to 2013, so I decided to focus on the 2013-2023 seasons, giving us a sample of 11 years, more than enough to make conclusions. BR only allows you to view the top 100 PL players according to a certain statistic for each season, so I pulled the top 100 hitters by plate appearances for each season, yielding a sample of 1000 players. This also worked out nicely by giving us a list of hitters who played a significant amount in the PL, allowing us to make more significant conclusions on the differences in their performance. There were some repeats in the data consisting of hitters who spent multiple summers in the PL, but that shouldn’t disqualify them from this study, as we are still able to examine their respective seasons before and after their multiple PL stints. The goal was to take this list of hitters, use the R package rvest to access their respective BR pages, pull their college statistics from the seasons before and after their PL season, and combine these into “before” and “after” datasets that analysis can be performed upon.
This required looping through the hitter list, accessing the page for each hitter, and pulling their data. Player data was only added to the datasets if the player had recorded statistics at the NCAA level for the seasons before and after their PL season, ensuring that only hitters who fit the criteria of this study would be included. Given the URL structure on BR’s site, this also required looping through several potential URLs for each hitter until the correct one was found. (The hitter specific part of the URL usually follows “las###fir”, with the middle part being a numeric code that distinguishes between hitters with similar names, i.e. cra000owe, cra001owe, cra002owe, etc.)
After working through some issues with creating the loop, there was still a major problem. Baseball Reference’s parent company, Sports Reference, really doesn’t like it when you attempt to scrape too much data from their sites. You can read more about it here, but according to their terms of service, you are not allowed to make more than 10 requests a minute from their main sites. Violating this rule can block your session for up to a day. As I was trying to access the data for up to 1000 hitters in a short amount of time (and often making multiple requests per hitter to try to find the right URL), this obviously posed a large problem. This presented me with two main choices, rotating between proxy sessions and implementing a time delay in my program.
BR’s main reasoning behind limiting scraping is to not overwhelm their servers. Therefore, using a rotating proxy system to get the data as quickly as possible felt like the wrong thing to do. So, I implemented a time delay into my code that would put my scraping in compliance with the site’s rules. The downside to this was it made the process take a really long time. A delay of slightly over six seconds between each request for ~1800 requests resulted in a roughly three-hour data collection process. This was exacerbated by me accidentally erasing the data once it was finally done and having to run it again, as well as stopping and starting the program to fix encountered errors. For example, one of the hitters had to be removed from the data, as he shared the same name as a former Major League manager, and it made more sense to remove him than slowing down the code by checking that each hitter was not a former big-league skipper.
Results
Out of the sample of 1000 hitters, 524 had the necessary data for inclusion in the study. This means that they had NCAA data available on BR from the seasons immediately before and after their summer in the Prospect League. I feel that, although it is only about half of the original data set, this group of hitters is better for answering the questions being asked than hitters with partial data. For example, a hitter who did not play college baseball after the PL because they went pro would not be a good data point, as their level of competition would have been drastically different in their next season. They would’ve also likely been much more developed in their respective college seasons than the average PL hitter.
To further the significance of the data, I limited the before and after data sets to only include hitters who recorded at least 50 plate appearances in their respective NCAA seasons. This yielded 430 hitters who recorded significant statistics.
For analysis on the data, I decided to focus primarily on the triple slash, consisting of batting average (AVG), on-base percentage (OBP), and slugging percentage (SLG). With more advanced metrics not being available for this data, I felt these numbers would give a simple yet highly informative insight into these hitters. The average numbers from the before and after groups can be seen in the table below.
| AVG | OBP | SLG | |
| Before | .281 | .369 | .410 |
| After | .290 | .383 | .440 |
These differences can also be observed in the graph below.
As you can see, there is certainly a positive difference between the performances of these hitters before and after their time in the Prospect League. To confirm that these results are significant, two-sample t-tests were performed on each statistic. Given that there was a notable increase with all three numbers, I doubt there is any chance they aren’t statistically significant, but it is always important to check and confirm.
Performing these tests resulted in p-values of 0.025 for AVG, 0.0006 for OBP, and 0.0005 for SLG. This tells us that there is significant evidence that there are differences between these numbers in the before and after datasets, and the differences are not just due to random variation.
We now know that these increases are significant. However, it must be kept in mind that correlation is not always causation. These differences could be attributed to any number of things, namely hitters tending to play better in each college season, as amateur hitters usually tend to improve as they age.
Based on these results, the average PL hitter tends to improve his NCAA batting average by 9 points, OBP by 14 points, and slugging percentage by 30 points. These may seem like small differences, but in a game as exact and competitive as baseball, they could be the difference in going pro, transferring to a bigger school, or earning a starting spot in the lineup.
Overall, I believe there is a significant amount of evidence to conclude that hitters who spend their summer playing in the Prospect League tend to perform better in their next college season. Whether this is due to natural improvement with age, spending additional time practicing and playing baseball, or something specific with the Prospect League is hard to tell. However, it is reasonable to assume that continuing to play baseball through the summer will allow you to maintain and improve your performance, and the Prospect League is certainly an excellent facilitator of that.
As a quick aside, I thought it would be interesting to find the hitters who had the biggest jumps in performance between their two college seasons. In the table below, you can see the NCAA and PL stats of six hitters who experienced the largest increases in their triple slash numbers.
| Name | Team | Season | PA | AVG | OBP | SLG |
|---|---|---|---|---|---|---|
| Brennan Orf* | SIUE | 2021 | 59 | 0.137 | 0.254 | 0.255 |
| Normal | 2021 (PL) | 145 | 0.361 | 0.441 | 0.533 | |
| SIUE | 2022 | 185 | 0.354 | 0.492 | 0.708 | |
| Andrew Chuba | Seton Hill (D2) | 2018 | 53 | 0.167 | 0.308 | 0.262 |
| Butler | 2018 (PL) | 186 | 0.282 | 0.371 | 0.344 | |
| Seton Hill (D2) | 2019 | 186 | 0.373 | 0.484 | 0.587 | |
| Itchy Burts | Texas A&M CC | 2017 | 149 | 0.248 | 0.271 | 0.271 |
| Danville | 2017 (PL) | 133 | 0.276 | 0.318 | 0.301 | |
| Texas A&M CC | 2018 | 157 | 0.394 | 0.462 | 0.46 | |
| Ryan Uhl* | Indiana PA (D2) | 2014 | 142 | 0.267 | 0.383 | 0.388 |
| Butler | 2014 (PL) | 163 | 0.308 | 0.374 | 0.427 | |
| Indiana PA (D2) | 2015 | 181 | 0.416 | 0.536 | 1.085 | |
| Nick Sell* | Seton Hill (D2) | 2014 | 219 | 0.334 | 0.401 | 0.524 |
| Butler | 2014 (PL) | 146 | 0.256 | 0.297 | 0.364 | |
| Seton Hill (D2) | 2015 | 231 | 0.444 | 0.52 | 1.032 | |
| Doug Votolato* | Central Ark. | 2013 | 80 | 0.164 | 0.325 | 0.312 |
| West Virginia | 2013 (PL) | 172 | 0.277 | 0.396 | 0.376 | |
| Central Ark. | 2014 | 263 | 0.379 | 0.456 | 0.457 |
*= Signed Professionally
It is interesting to see how most of these hitters performed well in the Prospect League after subpar college seasons and continued that success into their next college season. It’s plausible that playing in the PL allowed them to improve skills or gain confidence that helped them succeed. Or maybe their coaches saw how well they performed and gave them more opportunities. Regardless, it’s clear that the Prospect League had at least some part in the improvement of these hitters.
Further Work
This would take a lot of data to answer, but I’d like to see if there’s any difference in the average development between college hitters who compete in summer leagues and those who don’t. I would especially like to see if there’s a difference between PL hitters and hitters who played in other summer leagues. Pursuing either of these ideas could give further credibility to this study and help to answer the question of how well the Prospect League develops hitters.
Thank you for taking the time to read this piece and please reach out with any questions, comments, or suggestions.
The R script used to acquire the data and perform analysis on it can be found here.