Minors to the Majors: Adding Plate Discipline to Hit Grade
I have been breaking down the prospect Hit grade over the past few weeks. Now it’s time to find what’s usable for projections. One aspect I continue to find is that a plate discipline component seems to not be incorporated into the Hit grades. For this reason, I created a minor league Plate Discipline grade and used it with existing data to hopefully better map out a hitter’s future potential.
To start with, I used Walk Rate (BB%-IBB%) minus Strikeout Rate (K%) as a measure of minor league plate discipline. I grouped the values by the low minors (any A ball) or the upper minors (AA and AAA). Then, I gave the players a 50 grade for each level’s average plate discipline and a 60 for one standard deviation from the median. Here are the results:
| Grade | Lower Miners | High Minors |
|---|---|---|
| 80 | 26.9% | 25.3% |
| 75 | 20.6% | 19.4% |
| 70 | 14.2% | 13.5% |
| 65 | 7.8% | 7.7% |
| 60 | 1.5% | 1.8% |
| 55 | -4.9% | -4.0% |
| 50 | -11.3% | -9.9% |
| 45 | -17.7% | -15.7% |
| 40 | -24.0% | -21.6% |
| 35 | -30.4% | -27.4% |
| 30 | -36.8% | -33.3% |
| 25 | -43.1% | -39.1% |
| 20 | -49.5% | -45.0% |
Next, I needed to find some comparable hitters. To do this, I used the newest prospect grades (min 600 PA) from the high minors (AA or AAA). Finally, I combined the hitter’s Plate Discipline grade along with his other components to see if I could predict future OBP. I found OBP to be the ultimate goal when evaluating prospects. Here are the various test results.
Mix: r-squared
30% Hit, 30% Power, 30% Plate Discipline, 10% Speed: 0.08
All four perfect mix (Hit was – value): .0974
Perfect mix w/o Hit Grade (64% PD, 13% Speed, 23% power): 0.109
Just Plate Discipline: 0.044
At least I ended up with a positive correlation this time. Barely. But it’s no real improvement over my previous tests.
On the “Perfect w/o Hit Grade” equation, the possible range of output grades is only between a 40 and a 60 grade. I wonder if this precision level is all that can be estimated. Maybe the Hit Grade should be treated like pitcher Control, assume a near 50 Grade and allow just slight adjustments from that point until more information is available.
I ran one final test. I compared the hitter’s high minor league on-base rate to his major league rate. The r-squared value ended up at 0.22. which is twice any other combination I analyzed.
I have one more test I’d like to eventually run but I don’t have the data. Until then, I feel using prospect Hit grades is still useless to determine future batting production.
So here’s a comparison of the 2017 grades from Baseball American top 100 prospects with their hitting grades.
| NAME | PA | OBP | Plate Discipline | Hit | Power | Speed |
|---|---|---|---|---|---|---|
| Aaron Judge | 950 | 51 | 46 | 40 | 70 | 50 |
| A.J. Reed | 533 | 58 | 49 | 55 | 60 | 30 |
| Albert Almora Jr. | 931 | 45 | 52 | 55 | 45 | 40 |
| Alex Verdugo | 529 | 49 | 54 | 55 | 55 | 45 |
| Amed Rosario | 247 | 57 | 45 | 60 | 45 | 60 |
| Andrew Benintendi | 263 | 53 | 56 | 70 | 60 | 55 |
| Austin Meadows | 363 | 50 | 49 | 60 | 60 | 60 |
| Bradley Zimmer | 771 | 52 | 44 | 45 | 55 | 55 |
| Carson Kelly | 362 | 50 | 49 | 40 | 50 | 20 |
| Casey Gillaspie | 560 | 58 | 52 | 45 | 60 | 30 |
| Chance Sisco | 581 | 59 | 53 | 60 | 45 | 30 |
| Clint Frazier | 520 | 49 | 46 | 50 | 60 | 55 |
| Cody Bellinger | 477 | 54 | 52 | 60 | 70 | 50 |
| Dansby Swanson | 377 | 50 | 50 | 60 | 50 | 60 |
| Dominic Smith | 542 | 55 | 54 | 60 | 50 | 50 |
| Franklin Barreto | 525 | 50 | 49 | 60 | 50 | 50 |
| Hunter Renfroe | 1372 | 47 | 45 | 45 | 70 | 50 |
| Ian Happ | 274 | 46 | 46 | 55 | 55 | 55 |
| J.P. Crawford | 956 | 52 | 58 | 60 | 45 | 50 |
| Jake Bauers | 866 | 53 | 55 | 60 | 50 | 50 |
| Jorge Alfaro | 741 | 47 | 41 | 45 | 60 | 45 |
| Josh Bell | 1157 | 57 | 56 | 55 | 55 | 40 |
| Kevin Newman | 268 | 54 | 59 | 60 | 40 | 55 |
| Lewis Brinson | 576 | 47 | 46 | 50 | 60 | 60 |
| Manuel Margot | 848 | 51 | 54 | 60 | 40 | 60 |
| Matt Chapman | 592 | 48 | 42 | 40 | 55 | 40 |
| Ozzie Albies | 618 | 53 | 52 | 70 | 40 | 70 |
| Raimel Tapia | 567 | 54 | 53 | 60 | 40 | 50 |
| Rowdy Tellez | 514 | 58 | 53 | 50 | 60 | 20 |
| Tyler O’Neill | 575 | 56 | 45 | 50 | 60 | 45 |
| Willie Calhoun | 560 | 46 | 55 | 50 | 60 | 30 |
| Willy Adames | 568 | 55 | 51 | 60 | 55 | 50 |
| Yoan Moncada | 207 | 57 | 43 | 60 | 60 | 70 |
The high 70 Hit grades for Benintendi and Albies are probably inflated consider past examples production. The one player who is getting little love is Casey Gillaspie, who has a 45 Hit grade but his high minors rates point to a 60-grade Hit tool. It will be interesting to see which values, if any, are predictive this season. We’ll find out soon. And happy prospecting.
Jeff, one of the authors of the fantasy baseball guide,The Process, writes for RotoGraphs, The Hardball Times, Rotowire, Baseball America, and BaseballHQ. He has been nominated for two SABR Analytics Research Award for Contemporary Analysis and won it in 2013 in tandem with Bill Petti. He has won four FSWA Awards including on for his Mining the News series. He's won Tout Wars three times, LABR twice, and got his first NFBC Main Event win in 2021. Follow him on Twitter @jeffwzimmerman.
Isn’t it interesting that here we are trying to create more granular statistics, when so much of the work in advanced metrics is an attempt to create compound metrics like WAR? To be clear, I am 100% on board, but I like to take any opportunity I can to point it out. Quality analysis consists of both ends of the spectrum although too many people don’t get it. IMO the granular stuff is far more valuable than the crude compound metrics. Compound metrics are great for those who can’t follow the more intricate analysis or for a quick sort of league leaders/laggards. Good work trying to create some true progress!
The macro statistics like WAR are great to show a narrative, but if you want to know how that narrative was achieved or you want to know if it can be repeated, you need the granular statistics. Both have a time and a place.
I think everyone who can follow the granular data is already aware of the conclusions of macro statistics. I feel like the macro stuff is great for those who don’t understand either statistical analysis or baseball – which is most people. I don’t think something like WAR is worthless, but its certainly not the tool that many people mistake it for. It is a very crude approximation of value, not THE measuring stick that many people think it is. The author’s premise here is to take a crude measure and find some more useful granular data, which I think sounds like a good idea for those seeking deeper understanding. Most people will always seek the simplified superficial analysis because it is easier.
This is true however I think the #1 goal here is simply to find something that correlates to real major league success. I think it does seem granular but the idea is to get something to compound the separate factors of the hit tool.
The attempt here is to create a compound metric, that’s precisely the problem. Just as WAR uses granular data (# of outs made, hits, extra bases, etc.) to arrive at a generalized “value”, the attempt here is to use granular data to determine a generalized “Hit Tool”.
The key difference is that WAR uses a much more simplistic framework, i.e. WAR assumes the goal is to achieve the best result, therefore a HR is always better than a 2B is always better than a BB is always better than a K. Therefore, we know all of the components, and for the most part how they stack up against each other.
This analysis is attempting to do something much more complex though, since other traits such as “Power” and “Speed” are rather arbitrarily (not by the author, but by the system currently in place) taken away, the goal is slightly amorphous. Who has the better “Hit Tool”…
A. the player who walks 40% of the time and Ks 60% of the time, or
B. the player who singles 40% of the time and grounds out 60% of the time?
A. The player with a .350 OBP and .550 SLG, or
B. The player with a .400 OBP and .500 SLG?
A. The player who goes .300/.400/.500 with 40 Home Runs (I don’t even know if this is possible, but you get the idea), or
B. The player who goes .300/.400/.500 with 0 Home Runs?
The issue here is that “Hit Tool” is attempting to judge 2 skills, while simultaneously excluding at least 2 others. Skill #1 is the swing/don’t swing decision. Take lots of balls, don’t take lots of strike 3s. Skill #2 is the ability to make optimal contact if you do swing. These skills don’t easily weigh against one another though.
Take for example a player who walks a ton. This could be do to an exceptional eye, discerning precisely whether or not the pitch will be a strike every time. Or it could be due to failing to identify hittable pitches, and therefore failing on the swing/don’t swing decision in a way that is extremely difficult to measure. Despite his high walk rates, there is the potential for him to be even more valuable by actually hitting the ball.
Conversely, there is the player who almost never swings and misses. This could be due to incredible bat control, or it could be due to overly conservative swings. This difference is also extremely difficult to measure (although exit velocity and launch angle might be key factors in isolating this). Despite his low K rates, he could be even more valuable by swinging harder.
In addition to this already immense weighting problem, Power and Speed are also taken out of the equation, meaning the author now has to figure out some way of isolating each of those from the swing/don’t swing decision, and contact rates/outcomes.
And finally, the coup de grace is that in another sense, this question is actually pretty easy… You wanna predict a MiLB player’s future OBP?, Batting Average?, Slugging?, BABiP?… We can already do that almost as well as predicting an MLB player’s future stats, we just don’t use visual scouting measurements to do it.
Jeff, this series is amazing, some of the best stuff I’ve ever read on this site. Thank you.