In part one of The Business Case For Player Development series, we calculated prospect valuations for top tier minor leaguers and validated those estimates by looking at several trades throughout MLB. Building on that work, we use those valuations in this post to model the net-present value of improving prospects by providing them with an individualized training program. Through this analysis, we quantify how much money organizations leave on the table by not getting their pitchers fully equipped with a custom training plan based on a thorough assessment.
The State of Player Development in Professional Baseball
With front offices expanding and marketplaces becoming more efficient, professional baseball has become a copycat industry, of sorts. Successful teams have seen their employees poached and ideas copied, all in an attempt by competitors to nullify a team’s perceived advantage as quickly as it is capitalized on.
With this in mind, it is not surprising that more teams have begun heavily investing in player development after witnessing the Houston Astros’ rise from laughing stock to perennial contender. Taboo only a few years ago, Rapsodo units and Edgertronic cameras are now commonplace in one of baseball’s last frontiers where low-hanging fruit still exists—the backfields.
While the inclusion of flashy technology to develop pitches is certainly a welcome change in a stubborn industry, it has the potential to cause one to overlook the single most important attribute to a pitcher’s success: velocity.
For example, despite teams investing more money into technology, it is still common practice for them to prescribe cookie-cutter throwing programs that cater to an unrelenting minor-league schedule before the specific needs of the individual.
By failing to provide players with an individually designed throwing program based on personalized assessment, teams inevitably fall short in designing an optimal development plan for their pitchers.
Using in-gym data from Driveline and several MLB metrics to isolate the relationship between changing velocities and future production, we quantify how much this suboptimal behavior costs teams in terms of Net Present Value (NPV).
Can We Develop Velocity in Professional Athletes?
To set the baseline for our analysis, we first need to look at whether we can significantly improve the velocity of professional pitchers using a data-driven assess-train-reassess model.
To address this, we gathered data on 58 high-level athletes who went through our velocity development program within the past 18 months. All players averaged at least 88 mph on their fastballs during their initial mound velocity, were 19 years or older at the time of their assessments, and trained at Driveline for at least three weeks. The cumulative averages of the athletes who met this criteria are provided below:
Now, developing velocity is not a linear process and not every athlete is guaranteed to see results. To gain a better feel for what the distribution of the changes in velocity were for each pitcher in our sample, we binned outcomes in .5 mph increments and smoothed the buckets by frequencies:
As shown above, 41.95% of all athletes in our subset gained at least 1 mph from entry to exit mound velocity, whereas only 18.39% of athletes lost over 1 mph.
Thus, despite arguments from days gone by, it’s clear that velocity levels can increase over time by instituting proper training methods in a pitcher’s routine. With this finding, we can show that even small increases in velocity have positive effects in production and overall value.
Translating Changes in Velocity to Production
While improving velocity is a net positive, it’s important to acknowledge that velocity alone does not get batters out. So, if we are interested in quantifying how an uptick or downtick in velocity impacts a player’s projected net value to an organization, we must consider a pitcher’s entire projection in the process.
To do that, we turned to the publicly available Steamer Projections system, which has been providing yearly MLB pre-season projections for fastball velocity (FBv) and RA9 (or at least runs and innings pitched) since 2013.
The idea in this exercise is simple; we controlled for changing run environments on a year-to-year basis and found the difference between projected FBv and actual FBv (min 50 IP) as well as projected RA9 and actual RA9 for each individual pitcher at the MLB level. We expected that a pitcher who threw his fastball slower than expected in a given season would have worse outcomes than previously projected, and vice versa.
This relationship is exactly what we found over a sample of 1,838 athletes. More specifically, a pitcher who out performed his fastball velocity projection by +1 mph produced an RA9 that was .2443 points lower than expected, on average.
(This chart shows that as velocity increases above expectations, ERA, RA9, and FIP all decrease while K% increases.)
To scale this finding to WAR, we took a hypothetical player, held his IP constant at 75 (a hedge between a starter and reliever), and calculated his WAR before and after a .2443 point reduction in RA9. We found that adding 1 mph to a player’s FBv increased the expected production of said player by ~.25-.35 WAR per season, depending on whether he was a starter or reliever.
Translating Updated Talent Levels to Prospect Grades
With estimates that relate changes in velocity to changes in performance, Kiley McDaniel’s scouting scale provides us with a link between the expected production level of a prospect at maturation (WAR) and his overall prospect grade (FV).
For most noteworthy prospects, McDaniels estimated that a .5 WAR change in projected talent level coincided with a half-grade increase or decrease in FV. Thus, we can deduce that a +2 mph increase in FB velocity causes a player’s talent level to increase by ~.5 wins and 5 points in FV.
The process here is generally straightforward. Say, for example, we have a 50 FV prospect that averaged 92 mph on his FB last season and gained 2 mph during the most recent off-season. Given the findings above, we would expect that this prospect’s projected RA9 would drop by ~.5 points due to the increase in velocity. Over the course of the season, we estimate this to be worth ~.5 WAR of production, which would boost his yearly projection from 2 WAR/year to 2.5 WAR/year. Using the chart above, this change in production would be the equivalent of moving his grade from 50 FV to 55 FV.
Simulating Value That Teams Leave on the Table
Since we now have a link between throwing harder and production from an individual standpoint, we next set out to apply this finding on an organizational level. This allows us to determine how much a league-average farm system could improve their overall production level if they provided their top prospects with a customized throwing plan.
To gain a proxy of a league-average farm system, we obtained both the number and quality of pitching prospects graded 40 FV and above for systems ranked 12th – 18th from 2016 to 2018 by Fangraphs.
The selected farm systems contained an average of approximately 13 pitchers with a grade of 40 or higher. The distribution of grades for these prospects is shown in the table below:
With these frequencies at our disposal, we created a multinomial simulation that generated 13,000 pseudo pitching prospects across 1,000 “league-average” farm systems. These prospects were designed to be representative of a prototypical pitching prospect that had only been exposed to traditional throwing programs in their past.
(The graph above shows the average grade of a corresponding prospect rank within each farm system.)
To apply the impact that Driveline’s throwing program could have on these generated prospects, we created an additional multinomial simulation by using the results from our velocity development program to find expected changes in their velocity via our program. By linking changes in velocity with a coinciding change in future value, we obtained a before and after snapshot of future value for each prospect after having gone through a Driveline program.
The Initial Results
We converted these FV grades to NPV values and then averaged the change in surplus value for each farm system before and after the Driveline throwing program. From our experiment, it was found that an organization increased the NPV of their farm system by approximately $38 million after having their prospects go through a velocity development program. This is roughly the equivalent of adding a 55-FV prospect to a farm system or having two 45-FV prospects jump to a 50 FV.
(Above is a hypothetical farm system both before and after the Driveline throwing program.)
Scaling Back Our Estimates
While a $38 million estimate might be a little eye opening initially, we believe that this estimate is likely inflated by a few shortcomings in our initial methods that need to be addressed.
First, our initial methods do not consider that some prospects would have likely seen changes in their velocity had they just continued with their traditional throwing program. In other words, hypothetical prospects went through a Driveline throwing program and either gained or lost velocity based on the expected results of our program. The control group, however, had their velocity stay the same despite the fact that they would have been training as well. Thus, this needs to be accounted for.
Second, although the sample of players used to build our Driveline throwing program simulator was comprised of many professional athletes, the average player within our sample was not representative of a 45-FV prospect or higher. As a result, our initial estimate is likely biased upwards because higher-level prospects are more likely to be closer to their velocity ceiling than those in our sample. This makes marginal velocity gains for top prospects more difficult to obtain than our simulation led to believe.
(Graphic above is taken from our Summer 2018 Pitching Analysis)
To adjust for both of these shortcomings, we regressed our initial results by 25% and threw out any prospects in our sample graded 45 FV or higher. While both somewhat arbitrary concessions, we think these additional steps tackle both limitations head on.
For example, in regressing our changes in NPV by 25%, we acknowledge that pitchers can either develop or lose velocity by sticking with a more traditional program. That said, we also recognize that changes in velocity are likely to be less volatile for players using traditional training methods than for players using unfamiliar training modalities.
In removing any prospects graded 45 or higher, we remove a subset of pitchers that predominantly already throw hard and have had large amounts of success training with traditional methods. Furthermore, this subset of prospects would likely have more jurisdiction over their own program, and thus they might be less likely to follow through on a new training stimulus that is foreign to them.
With only 40-grade pitching prospects remaining within our hypothetical sample of athletes, we repeat the same methodology as described above and regress the changes in NPV for each farm system by 25%. With these adjustments, the average change in NPV for each farm system drops down to ~$11 million with a wide distribution that is skewed leftward.
This conservative $11 million estimate is still a large sum of money that most teams have not cashed in on, and it does not consider other elements of value that individualized throwing programs provide teams, which we will address in future pieces.
For example, the methods above do not give individualized throwing programs to prospects graded 35 FV and below, who constitute 80-85% of all Minor League Baseball players. This subset of players should be able leverage their individualized programs more so than top prospects, given that they benefit the most from increasing outcome volatility. The same can be said for AAAA players who are heading into the final stages of their careers and high profile prospects labeled as projectable, whose value is predicated on improving velocity.
The estimate above also does not consider the value of programming daily warm-up and recovery modalities in preventing injuries, which keeps players on the field longer and mitigates significant rehab costs for organizations in the long run.
A Better Framework for Development
Given the influx of dollars now being spent on player development, it is clear that giving pitchers a development plan that addresses their individual deficiencies and helps develop velocity has a massive ROI for organizations. The reason for this is straightforward: the upside of increasing the volatility of outcomes for 40-FV prospects and lower will always outweigh the risk.
Rather than having these players continue to accrue 25 starts season after season, perhaps it is time teams re-evaluate and provide their athletes with a plan that is going to make them more likely to reach their goal of making it to the big leagues.
After all, it has been about 10 years since we learned that when you gain a tick, you lose one, so how long will it be before teams decide to put this research into practice?
Written by sabermetircs analyst Dan Aucoin.