A Statistical Approach to Identifying And Classifying Future Major League Starting Pitchers
Martin Perez - Photo courtesy of Scott LucasBaseball America defines a top-of-the-rotation pitcher (TORP) as a starter with two plus pitches, one average pitch, and plus command. Middle-of-the-rotation and back-of-the-rotation pitchers are defined as those with three average to above-average pitches and average command.
If you have the time to attend a lot of minor league baseball games, the money for seats behind home plate, a radar gun, and a trained eye, then this grading system is likely to be very effective in allowing you to predict the Rangers' starting rotations in the next few years. For those of you who lack one or more of these qualifications, I offer you an alternative approach.
Minor League Statistical Metrics That Distinguish Future Major Leaguers
To address whether the minor league statistical profiles of current major league starting pitchers are distinct from those of the minor league population, I calculated the average age, ERA, WHIP, H/9, HR/9, BB/9, and SO/9 at each minor league stop for a group of major league starting pitchers and compared the results to the averages of the players pitching in the various minor league levels in 2009. The major league starters used for the study are listed in the second table. The calculated averages are displayed in the first table, with the number before the colons representing the minor league averages of the 34 major league pitchers and the number after the colons being that of the 2009 minor league pitchers:
As expected, the minor league statistical profiles of the major league starting pitchers are significantly better than the general minor league population at every level. The greatest variance is observed in the ages, ERAs, and strikeout rates of the future major leaguers, suggesting that these are good indicators to use when seeking to identify prospects. In theory, one could use the averages of the major league pitchers in each row as rough cutoffs to identify interesting prospects at each level of the minor leagues.
Minor League Statistical Metrics That Classify Future Major Leaguers
In my opinion, a more interesting question is whether it is possible to distinguish minor league pitchers who are destined to front major league rotations from those who are going to slot as No. 3, No. 4 or No. 5 pitchers. To develop sample sets to accommodate such a study, I used the WAR calculations provided by FanGraphs to rank the performances of starting pitchers from 2007-2009 and then created a group of elite pitchers, a group of above-average pitchers, and a group of below average pitchers. As noted in the table below, the three groups represent TORPs (Top of the Rotation Pitchers), MORPs (Middle Of the Rotation Pitchers), and BORPs (Back Of the Rotation Pitchers).
Worth noting is that Tim Lincecum and Johan Santana were originally in the TORP group, but their minor league careers were so unusual (Lincecum because he completely dominated everyone in 23 minor league games and Santana because his time in the minor leagues was cut short by the Rule 5 draft) that I eliminated them from the study.
The career minor league stats and ages that the pitchers made their major league debuts were assembled and used to calculate averages and standard deviations for each of the three classes of pitcher. Shown in the table below are a few key metrics that are often used to describe minor league pitching prospects. For each cell, the first number is the average for all pitchers in the group and the number in parentheses is the standard deviation. The fourth row in each table is the calculated p-value that expresses whether the separation in the groups of pitchers is statistically significant. Highlighted in gray are stats with p-values less than 0.05. All stats are career minor league numbers except for ground-to-fly ball rates, which derive from major league numbers since I was unable to find a database of minor league groundball rates that predates 2005:
TORPs vs. MORPs
The average TORP pitches fewer minor league games and makes his major league debut almost two years before the average MORP. The average ERA, WHIP, HR/9, and BAA are all lower for TORPs, though only ERA is significantly different. Along with ERA, the best performance stat for distinguishing a TORP from a MORP is strikeout rate as future elite pitchers strikeout nearly two more hitters per nine innings than do mid-rotation pitchers. Though it is not statistically significant, it is interesting to note that the average MORP exhibits slightly better control than the average TORP (2.7 vs 2.9 BB/9). This might be due to MORPs relying on control to overcome less dominant stuff or it might simply reflect that hitters are more likely to hit pitches from MORPs before working a walk.
TORPs vs. BORPs
After the comparison of TORPs and MORPs produced results that one would predict, I was gratifyingly shocked by the results of the comparison of the top and back of the rotation pitchers. My original analysis did not include ground-to-fly ball rates and so I was dumbfounded to discover that there was not a single statistical metric that clearly separated TORPs from BORPs. In their minor league careers, the TORPs that I used were only marginally better than BORPs in each of the commonly used metrics.
A fairly broad search of other statistics revealed ground-to-fly ball ratio as the one area where TORPs clearly exceeded the performances of the BORPs. Among the 11 pitchers who were classified as TORPs, only two (Greinke and Hamels) have career ground-to-fly ball rates below 1.25 and even they induce more ground balls than fly balls. In contrast, Garland and Cabrera are the only BORPs with ground-to-fly ball ratios of at least 1.25 and 5 of the 11 BORPs have yielded more fly balls than ground balls in their major league careers. These results suggest that identifying future TORPs begins with a search for young for their league pitchers with high strikeout and solid walk totals and ends with a review of GB/FB ratios.
The Developmental Profiles of Future Major League Pitchers
To get a sense of the statistical profiles of future major leaguers as they work their ways through the minor leagues, I averaged the career numbers for the TORPs, MORPs, and BORPs for each minor league level. The table below displays the results. Worth noting is that the averages are for the players who pitched more than five games at a given level.
TORPs
The remarkable thing about the table above is how little the statistical profiles of TORPs change as they move through the minor leagues. These numbers suggest that if a pitching prospect struggles following a move to a higher level, he is probably not a TORP. In addition, it doesn't require very many innings for a TORP to graduate from one level to the next.
MORPs
MORPs can be distinguished from TORPs based upon strike out rates as soon as they reach the full-season leagues. On average, MORPs maintain solid walk rates and mid-3 ERAs throughout their minor league careers. Their strikeout rates step down as they advance, beginning around 9.0 K/9 in the short-season leagues and ending at approximately 7.0 K/9 in AAA-ball.
BORPs
In addition to the fly ball rates, BORPs can be distinguished from TORPs due to their dip in strikeout rates in AAA-ball. Among the BORPs used for this study, only Hill and Snell had AAA strikeout rates of at least 9.0 K/9. In contrast, only Halladay and Wainwright produced fewer than 9.0 K/9 among TORPs who pitched at least 50 innings in AAA-ball.
Classifying Rangers Pitching Prospects
The purpose of this exercise was to develop a sense of how current and future Rangers prospects stack up relative to current major league starting pitchers. To standardize the analysis as much as possible, I developed a simple formula that estimates whether a pitching prospect is producing TORP-like numbers. The formula incorporates ERA, BB/9, SO/9, GB/FB rate, and predicted age of ML debut. Each metric for each pitcher is compared to the average of the major league TORPs used in the study and normalized to 100. To avoid problems with over-weighting any given category, I capped the maximal weighted score for each statistic at 110.
The average of the five normalized numbers is the TORP rating. A prospect with minor league numbers that are identical to the TORP averages would have a TORP rating of 100 and would presumably be on track to develop into a major league top of the rotation pitcher. Due to sample size and competition level, TORP ratings for pitchers at the lower minor league levels will be less reliable than those for pitchers in the upper minors.
Below are the career statistical profiles of the Rangers' most interesting starting pitching prospects listed in order of their TORP rating. To facilitate comparisons, the minor league career averages for the TORPs, MORPs, and BORPs used in the study are provided in the bottom three rows. The average GB/FB rates displayed in the bottom three rows were calculated using the career major league statistics of the pitchers. All other values are from the minor leagues.
[* in the ML Debut column indicates the projected age when a pitcher will make his major league debut.]
Not surprisingly, Perez, Holland, and Feliz rank among the top four in TORP rating and appear to be on track to develop into top of the rotation starters. Though their ratings are based upon very small sample sizes, Robbie Ross and Edwin Escobar appear to be the only two other pitchers in the Rangers' system with the combination of strikeout, walk, and groundball rates that are common among TORPs.
Boscan, Wieland, Hunter, Beavan, Gomez, and Poveda all profile as MORPs. Boscan, Wieland, Hunter, and Beavan all appear capable of overcoming pedestrian strikeout rates with outstanding control and solid ground ball rates. Gomez gets overlooked a lot, but he has an outstanding ground-to-fly ball ratio (1.99) and very solid strikeout (8.3 K/9) and walk (3.3 BB/9) rates. Don't be surprised to see him at least match what Kiker and Poveda have done in 2009 when he gets his turn in Frisco next year. I did not include Braden Tullis or Matt Thompson in this study, but I suspect that they will fit nicely into the MORP category in future lists.
Main, Kiker, Moscoso, and Font currently profile as back of the rotation starters or relievers. The first three have displayed the high strikeout/high fly ball rates that define BORPs. Font's solid ground-to-fly ball rate belies the 0.87 GB/FB that he has posted this season and suggests that he rightfully belongs in the BORP group. Of these four pitchers, Main and Font appear to be the most likely to develop the ability to induce ground balls at an acceptable rate. If not, then look for them to join Kiker and Moscoso in struggling to become league average major league starting pitchers.
[Just for fun, I calculated TORP ratings for a few highly regarded pitching prospects: Bumgarner - 96.8, Tillman – 82.2, Matusz – 96.3, Poreda – 95.5, Hanson – 89.8, Parker – 96, Davis - 80, Bowden - 83.8, Niese - 85.7, Cecil - 97.6, Chacin - 100.7.]
30 Comments | in Analysis, Prospects, The Season
Reader Comments (30)
Thanks for the analysis. Great read. How do you have time to do all this?
Very good work. Thanks. I wonder how the Rangers system would have looked 4 years ago? I am sure the normal Ranger fan doesn't really appreciate the Rangers' farm system. Thanks JD.
Fantastic piece, David. I've got great respect for your research skills and patience, heh.
Of all our exciting new pitching prospects, I did not expect to see Ross' name at the top. Outstanding stuff, David.
Won't Perez and Ross's (and all of the young guys who haven't been to upper levels yet) numbers still be pretty unreliable at this point? I feel like Feliz, Holland, even the guys in AA that have seen 3-5 levels could be projectable, the younger guys have so far to go, they're practically crapshoots.
Not sure Ross should be up there with only 10 games under his belt. There should be a min. number of IP to qualify... with that said, it is what it is and I get your point.
I'm amazed at the amount of stats you guys come up with. Are you employed by JD and crew? You should be...
I'm still holding to my prediction that Kiker will be a #2 or #3 SP in another 2-3 yrs. I know his "stuff" and his numbers/stats say otherwise but the kid is fiesty and competitive...something that can't be measured. I will humbly proclaim I'm wrong if indeed that is the case... but I want my name in 3" headlines if I'm right. Do we have a deal?
I'm actually not that surprised about Ross being so high on the list. Imagine what it would have looked like if you didn't cap each score at 110. Sure his GB/FB will normalize as he gets another 20 starts under his belt, but I believe it might stay above 2 until he reaches AA/AAA. The opposite could be said for Main. He still has plenty of time to erase his terrible 2009 (especially if he's pitching without some unidentified viral infection).
As for another idea for statistical analysis, could you take a look at consistency? I guess standard deviation on per start averages would work. Think of the difference this season between Halladay and Holland. Halladay usually guarantees you 7 innings pitched and about 2 runs whereas Holland could give you anywhere from 4-9 innings and 0-5 runs allowed. I would probably eliminate Holland from the study and focus on pitchers with 30 or more major league starts. Don't feel obligated to look into this; it just popped into my head, and I thought I'd feed your demons.
WOW! I'm impressed. After reading it and getting the insights, I'm definitely going to reread it later with more time on my hand. It is so detailed and powerful. Fantastic work.
BTW, this is my new go-to-site for all things Texas Rangers because of you, Joey, and the great analysis!
Martin Perez!!!!!!!! Can't wait to see him!
The Rangers 2011 rotation could look like this:
Holland
Feliz
Perez
Feldman
Hunter
That may just be playoff worthy... and think of all the guys I'm leaving out... SCARY!
I always love the detailed statistical analysis. Wonderful stuff. How would Feldman and Harrison fit into this analysis?
Can we get an update on signing status of the draft picks?
I know it's a fairly small sample size for Mr Ross, but you've gotta be kidding me that he has a 4.13 GB/FB ratio. Does he hurl a cannon ball or shot put?
Great stuff, as usual, David. I wish you had included Tullis, since he has only 8 fewer innings than Ross, though I agree he'd probably rank in the MORP area.
The system has an unbelievable number of guys who could be BORPs and MORPs in addition to the top, sexiest handful. Even with the usual attrition rates it seems the Rangers will only be signing free agent starters as a matter of choice rather than desperate need as in so many years past.
I'm not sure if there is anyone who enjoys your work more than me. Excellent work David as always. However, I do have a bone to pick with you on this study. You stated that MORP's have better walk rates per 9 innings. That's fine, but you claim that they have better control also. I disagree. Since TORP's have a better K/9 rate, this implies that their walk rates may be higher also. Think of Nolan Ryan as an example. He had both high strikeout and walk ratios. In other words, batters had a hard time putting the ball in play against Nolan. If a batter can't put the ball in play, then this means that they can only muster a strikeout or walk. So when a pitcher misses more bats, his walk rates will necessarily go up in tandem.
Thanks for the comments today and sorry for the delay in responding - a half day meeting became a full day meeting and kept me from checking in.
Andrew - Joey does the heavy lifting around here with daily commentary. I get to come in with my once per week fluff pieces.
Rob - The Rangers' Top 5 pitching prospects and their TORP rankings in 2005 were Danks (borderline TORP), Diamond (BORP), Volquez (TORP), Hurley (BORP), and Galarraga (MORP).
Dan and Pablo - The data for Ross (and Escobar) should be viewed with great caution. Sample size and level are issues fro both. They are on the list because their early numbers are consistent with what major league TORPs posted in the short season leagues. Because the algorithm that I used relied upon career numbers, it is far more likely that a pitcher will post TORP-like numbers in his first ten games than it is that he will do it in his first fifty. With that said, those guys are definitely on my watch list given their combination of K's, BB's, and groundballs.
Regarding Martin Perez, he is probably close to half way through his minor league career and seemingly improving with each outing. Unless he stumbles at the upper levels, his career minor league numbers look very TORP-like.
Pablo - I wish only the best for Kasey, but his high flyball rate and iffy control suggest that he is going to have a tough time developing into anything more than a back of the rotation pitcher.
Dave - Thanks for the idea on pitching consistency. I would hypothesize that there is a strong correspondence between consistency and experience. Perhaps I will give it a look.
jmhard - Harrison and Feldman both have the groundball and walk rates of mid-rotation pitchers, Their strikeout rates are below average which would leave them as questionable major league starters according to this analysis. Obviously, Feldman is doing well despite his low strikeout totals. Hopefully, Harrison will be able to do the same. In Harrison' case, he absolutely has to limit walks or he will not be able to stick in the rotation.
JP - I am really excited about Ross. Even if he doesn't sustain his strikeout rate, it seems possible that he could follow the path of Jhoulys Chacin who, in my opinion, is one of the more interesting pitching prospects going.
t ball - Your comment regarding the benefits of having a system full of MORPs and BORPs is spot on. Assuming the Rangers continue to emphasize defense, it is reasonable to expect that the team could roll out a starting rotation with a sub-4.50 ERA for less than $15M in salary from 2010-2015. Assuming a reasonable number of the mid-rotation prospects pan out, the Rangers should have a surplus of league average starters that they can trade to teams that are desparate for cheap, effective pitchers. Those trades can either be used to supplement the major league roster or strengthen the minor league system. And therein lies the secret to a sustainable model for contending with a limited budget.
Rod - Thanks for the compliment and the critique. You are absolutely correct about the relationship between strikeouts, walks, and control. I will eliminate that sentence from the article. Thanks for pointing it out.
David: I really wasn't serious about you going back 4 yrs, but. thanks.
Rob - I thought it was an interesting question. The funny thing about the 2005 pitching prospects is that it was the best group of minor league pitchers that the Rangers had had in a very long time. It's hard to get excited about a system with two TORPs, a MORP, and two BORPs in 2009 when you can see that collection of prospects pitching on one night in the Rangers farm system. Good times, my friend!
David, thanks for one of the best stat analyses I have read. Not surprisingly, the data indicates little separation between the two constructs "MORP" and "BORP."
It will be interesting to see if and how Michael Main's projected performance changes when he pitches healthy next year and hopefully outside of the launching pad called Bakersfield.
Appreciate your work...........
great work david. timely and interesting studies -- that's what this site is all about. great job again.
One of the most interesting things I saw... was focusing in on the standard deviations of the TORPs, MORPs & BORPs in the various statistical categories. There was a lot more volatility in the TORPs and BORPs, and there was significantly less volatility in the MORPs.
As far as the tables that show how the stats breakdown by minor league level, for each category... I'd be interested in seeing the delta between the averages at each minor league level and the overall averages. What that would tell us is at what levels they over or under-performed compared to their average performance. This could be eyeballed (with ERA for example) by looking at the column and for the values that are lower than the others you know that means they performed better than average in those levels.... But if you did this across all stat categories, it would be interesting to see where the +/- are across the entire cross-section.
I find the classification of Wilmer Wingding as a BORP or a reliever a little disappointing. I understand that his Walk Rate is WAY too high, which influences his WHIP. You get improvement on the Walk Rate and he shoots up to TORP status (All of his other numbers besides GB:FB mirror Escobar's almost to a T). So if the walk rate is corrected or at least cut in half, does this trigger a massive movement up the ladder? Or does the GB:FB metric weight the analysis to where he'd still be a BORP and maybe a MORP until fixed?
To me, there isn't a lot separating Font from the other TORPs on the list. The If is a big one, as Walk rates are always a tough thing to minimize, but I feel he'll be walking less once he grows into his body and learns to control his pitches a bit better. I think Font will move up the ladder very quickly once he gets another full season (and off-season) under his belt.
David, is there enough data to do a comparative analysis of the different minor leagues? Perhaps you could prove (or disprove) the Cal league being extremely hitter friendly.
Troy - The larger standard deviations for the TORPs were driven by the surprisingly mediocre career minor league numbers of Roy Halladay (638 IP, 3.84 ERA, 5.9 K/9) the relatively variant numbers of Sabathia, Beckett, Hamels, and Greinke who all had relatively short minor league careers and very little time in AAA. I'm not sure what you are suggesting in your second paragraph, though it sounds like it is interesting.
HypoLuxa - If Font can't get his walks under control, then he is probably not going to make it to the major leagues. Assuming he does (and I think that is reasonable given teh strides he has made this season), then he really needs to develop an ability to pitch low in the strike zone and induce groundballs. Fortunately for Font, improving control and command is easier than improving velocity so he has a better chance of developing into a bonafide major leaguer than someone without his natural ability.
Dave - The Cal League inflates slugging but doesn't appear to do much else relative to the upper levels. Below is a rundown of the 2009 league average hitting numbers for each of the leagues in which the Rangers participate:
Arizona - 270/345/391
NWL - 266/348/386
SAL - 254/323/369
CAL - 270/340/415
TEX - 271/345/396
PCL - 272/341/416
Thanks for the quick analysis. I like the vertical study, but I also wonder how each league compares to others at its level. How does Cal compare to Carolina and Florida State?
Sorry, Dave, I misunderstood your question.
The CAL League is the undisputed hitter's paradise of high A. The other two High A leagues are in the southeast and they are both considered to be pitcher-friendly.
Carolina League hitters are currently posting a 257/331/384 line and pitchers have an average ERA of 4.00.
Florida State League hitters are currently posting a 252/321/363 line and pitchers have an average ERA of 3.55.
Hitting is down this year in the Cal League and yet hitters are currently posting a 270/340/415 line and pitchers have an average ERA of 4.49.
Holland, Feliz, Hunter, Poveda, and Moscoso will all have to be on the 40 man to be protected in the off season. Is there anyone else on this list that will be eligible for the Rule 5 draft this year? Do you think they'll keep Poveda protected again?
You didn't misunderstand my question, I just wondered about another angle, and you also answered that well.
jmhard - Michael Kirkman failed to rate highly enough on the TORP list to be included in the study, but he is only Rule 5 eligible pitcher of any consequence that you did not list. I think that there is enough wiggle in the Rangers' 40 man roster to accommodate Poveda for another year. He needs to really pick things up next year though or he will lose his spot to someone in the 2010 class of Rule 5 eligibles (Beltre, Boscan, Font, Gomez, Kiker, Lemon, Moreland, Pimental, Smith).