/cdn.vox-cdn.com/uploads/chorus_image/image/43218978/505664978.0.jpg)
Introduction
After working with hitting prospects for so many years, it is finally time to turn my attention to pitching prospects. Pitchers are known to be more volatile than hitters, as evidenced by the popularity of the phrase TINSTAAPP – there is no such thing as a pitching prospect. However, just because it is hard to project pitchers, does not mean we should not try. Is there some signal in the noise?
Method
The data for pitchers in the minor leagues are extremely limited. As such, the only statistics I use to compare players are walk rate per inning, strikeout rate per inning and percent of games started. I also factor height and handedness into the equation.
All the thank yous go out to Sean Lahman and Harry Pavlidis for helping me get the necessary biographical data for thousands of seemingly-insignificant minor league players. This allowed me to find additional comparisons using height and handedness. Without their assistance, I would not have been able to include these factors into the analysis.
The only other available statistics dating to 1979 are home runs and hits. Both of these statistics are very dependent on environment, and with the incredible variations between minor league locations, I would rather not use them.
Regression and age-adjustment
I regressed walk rate by 14 innings and strikeout rate by 8 innings, based on the following logic. Russell Carleton says that walk and strikeout rate stabilize at 120 and 240 PA, respectively, for hitters; this leads to regressing by 65 and 35 PA. Assuming these statistics stabilize at the same rate for pitchers*, I would regress by the same amount. Unfortunately, I do not have data for total batters faced for minor league pitchers, so I make another assumption that each pitcher faces 4.5 batters per inning. The 65 and 35 PA numbers then become approximately 14 and 8 innings pitched.
*As it turns out, they actually stabilize earlier for pitchers (170 and 70 TBF, respectively).
Once the numbers are regressed, I adjust them by age in the same manner as with hitters. I calculate the ratio of their age to the league average age and multiply their regressed zK and zBB numbers by that.
Comparisons
Once the zK and zBB numbers are regressed, age-adjusted, and accumulated for the player's entire minor league career prior to losing rookie eligibility in the majors, the comparison system is ready to go. I find ranges around each pitcher's zK, zBB, and height that allow for approximately 100-200 comparisons. The similar players must also have the same handedness. Since there are fewer left-handed pitchers than right-handed, the ranges for southpaws are larger than those for righties.
Once these similar players are found, JAVIER calculates their average career fWAR. I used fWAR as opposed to WARP as there seems to be more support behind Fangraphs’ calculations of pitching statistics than Baseball Prospectus’s.
JAVIER score
The concept of a JAVIER score as the final output for the system arose from the desire to combine hitters and pitchers in one leaderboard. Since everything else is done with z-scores, why not this too? I found the average and standard deviation of the raw average VORP or fWAR for hitters and pitchers and created the z-score for each from this.
After creating the z-score, I adjust for how far the player has progressed in the minor leagues. I multiply the JAVIER score by a percentage based on his highest level reached. These numbers are based on what percentage of players bust at that level.
Highest Level | Factor |
---|---|
Rk | 0.2 |
FRk | 0.2 |
A- | 0.2 |
A | 0.3 |
A+ | 0.3 |
AA | 0.5 |
AAA | 0.8 |
MLB | 1.0 |
For instance, if a player has a raw JAVIER score of four, but has only reached single-A, his adjusted JAVIER is 1.2. If he has reached AAA, it is 3.2. This way players further along in the development process find their way higher up the list.
JAVIER Categories
In order to compartmentalize the results even more, I created categories based on the raw JAVIER Score (not league factor adjusted) calculated from the percentage of historically busted players with that score.
Category | Highest JAVIER Score |
---|---|
Terrible | 0 |
Bad | 0.75 |
Average | 2 |
Good | 3.75 |
Elite | N/A |
Any player with a JAVIER score below zero has terrible minor league statistics, while any player with a score above 3.75 is elite.
Of the 17,000 eligible minor league players, 161 are labeled elite. 74% of those 161 are hitters, which I like, since pitchers have a greater bust rate. We are less confident that they will succeed, so there should be fewer elite pitchers. The ratio might be a little off, but I do believe there should be more hitters.
Results
Here are the JAVIER scores for all pitchers younger than 25.
How does career fWAR trend with pitching JAVIER score for qualifying pitchers?
Here are the minor league pitchers who were ranked as elite based on their minor league statistics and were at least 28 years old in 2014.
NAME | Comps | JAVIER | MLB Career fWAR | IP | Height | Throws | GS% | zBB | zK |
---|---|---|---|---|---|---|---|---|---|
Jered Weaver | 103 | 7.37 | 31.7 | 164.1 | 79 | R | 97% | -1.01 | 2.51 |
Phil Hughes | 101 | 6.15 | 17.1 | 306.1 | 77 | R | 95% | -1.09 | 2.37 |
Josh Beckett | 120 | 5.78 | 39 | 212.3 | 77 | R | 95% | -1.02 | 3.31 |
Rick Ankiel | 103 | 5.70 | 3.4 | 390.6 | 73 | L | 100% | -0.82 | 4.61 |
Erik Hanson | 122 | 5.43 | 32 | 387.6 | 78 | R | 96% | -0.82 | 2.38 |
Kerry Wood | 118 | 5.33 | 22 | 275.3 | 77 | R | 100% | 1.36 | 3.77 |
Brandon McCarthy | 160 | 5.23 | 14.4 | 392.3 | 79 | R | 97% | -1.37 | 2.54 |
Brad Penny | 158 | 5.17 | 25.2 | 428 | 76 | R | 100% | -1.03 | 2.32 |
Roger Clemens | 157 | 5.02 | 139.5 | 127.7 | 76 | R | 94% | -1.68 | 2.81 |
Anthony Reyes | 186 | 4.98 | 1 | 356.4 | 74 | R | 100% | -1.29 | 2.18 |
Ismael Valdez | 213 | 4.94 | 18.3 | 111.3 | 75 | R | 95% | -1.31 | 2.22 |
Paul Wilson | 167 | 4.93 | 8.2 | 243.9 | 77 | R | 100% | -0.91 | 1.92 |
Juan Pena | 166 | 4.76 | 0.7 | 577.3 | 77 | R | 97% | -1.10 | 1.78 |
Yovani Gallardo | 136 | 4.75 | 18.8 | 376.1 | 74 | R | 89% | -0.53 | 2.55 |
Ruben Quevedo | 141 | 4.74 | -1.4 | 579.4 | 73 | R | 89% | -0.78 | 2.36 |
Barry Zito | 105 | 4.62 | 30.6 | 170 | 74 | L | 100% | -0.10 | 2.36 |
Yusmeiro Petit | 171 | 4.61 | 2.9 | 595.9 | 73 | R | 100% | -1.29 | 2.05 |
Carl Pavano | 166 | 4.57 | 23.9 | 526.7 | 77 | R | 99% | -1.07 | 1.45 |
Andy Benes | 120 | 4.57 | 33.7 | 135 | 78 | R | 100% | -0.45 | 2.20 |
Rob Bell | 181 | 4.56 | 1.6 | 546 | 77 | R | 99% | -0.91 | 1.31 |
Francisco Liriano | 136 | 4.54 | 20 | 414.1 | 74 | L | 100% | -0.58 | 2.82 |
Bruce Chen | 133 | 4.48 | 9.3 | 499.6 | 74 | L | 95% | -0.75 | 2.39 |
Joel Zumaya | 103 | 4.45 | 2.9 | 358.3 | 75 | R | 96% | 0.00 | 3.58 |
John Stephens | 188 | 4.39 | 0.3 | 838.4 | 73 | R | 98% | -1.26 | 1.97 |
Dan Haren | 180 | 4.36 | 39.7 | 474.7 | 77 | R | 95% | -1.39 | 1.41 |
Clay Buchholz | 242 | 4.27 | 14.1 | 443.5 | 75 | R | 98% | -0.81 | 2.08 |
Jake Peavy | 170 | 4.18 | 38.7 | 358.3 | 73 | R | 98% | -0.61 | 3.01 |
Jeff Suppan | 244 | 4.15 | 20.9 | 591.7 | 74 | R | 99% | -1.23 | 1.99 |
Chin-hui Tsao | 249 | 4.14 | -0.2 | 371.6 | 74 | R | 99% | -0.91 | 1.99 |
Randy Wolf | 142 | 4.12 | 24.7 | 290.3 | 72 | L | 98% | -0.96 | 1.92 |
Jesse Foppert | 135 | 4.11 | 0.1 | 242.3 | 78 | R | 100% | -0.48 | 3.09 |
Mike Mussina | 264 | 4.05 | 82.5 | 177.9 | 74 | R | 100% | -1.20 | 1.88 |
Scott Olsen | 129 | 4.02 | 3.9 | 351 | 76 | L | 98% | -0.17 | 2.14 |
Mike Pelfrey | 187 | 4.00 | 9.7 | 176.5 | 79 | R | 100% | -0.50 | 1.26 |
Jeff D'Amico | 194 | 3.97 | 7.8 | 231 | 79 | R | 97% | -1.46 | 1.11 |
Bud Smith | 136 | 3.87 | 1.2 | 547.9 | 72 | L | 99% | -0.98 | 1.21 |
Jose Rosado | 134 | 3.80 | 12.1 | 247.7 | 72 | L | 100% | -0.90 | 1.21 |
Marty Bystrom | 159 | 3.79 | 6.1 | 267 | 77 | R | 100% | -0.97 | 1.05 |
Ramon Martinez | 182 | 3.79 | 19.6 | 543 | 76 | R | 100% | -0.35 | 2.10 |
Rich Harden | 120 | 3.79 | 16.9 | 334 | 73 | R | 91% | -0.13 | 2.82 |
Ryan Rupe | 166 | 3.77 | 3.8 | 115.2 | 78 | R | 100% | -0.90 | 0.72 |
Matt Cain | 172 | 3.77 | 28.4 | 378.4 | 75 | R | 100% | -0.45 | 3.26 |
While this list is not as impressive as the elite minor-league hitter one, it's still a grouping of some of the best pitchers in recent memory. Greg Maddux, Randy Johnson, and Pedro Martinez are notable omissions, but they do rank favorably in JAVIER, just not among the elite. Kevin Brown is the best pitcher to rank poorly according to JAVIER. His strikeout rate was well below the league average in the minors, but even so was ranked by Baseball America as the top Rangers' prospect in 1987 and 1989.
Finally, the moment we've all been waiting for: Combined JAVIER. How well does this system perform for both hitters and pitchers?
Below is the JAVIER file for every player contained in Baseball Prospectus's minor league data. You can download the data and look at it for yourself. Any thoughts on the system? Anything missing? Any improvements?
. . .
Statistics courtesy of Baseball Prospectus and Fangraphs.
Chris St. John is a writer at Beyond The Box Score. You can follow him on Twitter at @stealofhome.