This is the latest iteration of the method that I have been using to grade performance of minor league arms. I've transitioned recently to using run expectancies rather than relying simply on rates of walks, strikeouts, flyballs plus linedrives, etc. There are 3 basic steps that are explained in more detail here. First, a run expectancy matrix is constructed for the league under analysis (in the vein of the one put forth in The Book) via review of play-by-play data; this basically defines the typical number of runs that would score in the relevant league and season given each of the 24 possible permutations of outs and baserunners at the time of a plate appearance. That matrix can be used to define how each plate appearance of that league's season impacted run expectancy by subtracting the run expectancy just before the final pitch of the plate appearance (defined by the baserunners-outs state) from the run expectancy after it (defined by the baserunners-outs state) and adding however many runs scored on the pitch. And in so doing one can filter out all events of certain general types (such as line drive to the center-third of the field) from the league's seasonal play-by-play data and compute the average effect of that event type on run expectancy and that allows for construction of a second table along the lines of what is shown below (so a strikeout, on average, decreased the expected number of runs in a 2013 New York Penn League inning by 0.20 runs while an outfield flyball to the batter's pull-third of the field increased the expected number of runs in the inning by 0.30 runs, on average).
The third step is to quantify the pitcher's performance by multiplying the number of occurrences of each event type by the average effect that the corresponding event type had on run expectancy (per the table), summing those products, and then dividing the sum by the total number of plate appearances of those 12 event types to determine the average effect on run expectancy per plate appearance versus said pitcher (I do make some corrective adjustments for observer bias on event type classifications by ballpark, such as line drive versus outfield flyball). Computing this quantity for a group of pitchers using the average expected effect on run expectancy of each event type (rather than the actual effect on run expectancy of each plate appearance) allows the hurlers to be compared independent of ballpark, defensive, context (did the homers happen to occur with runners on?), luck and other largely-out-of-the-players'-control factors that impact run allowance. Here the more negative the average effect on run expectancy the better the pitcher performed. And reviewing the table above, one can see that the ideal pitcher in terms of performance per this scheme would walk few, strike out many, and largely avoid allowing line drives or pull-third outfield flyballs when the batter sends the ball into fair territory.
There are two groups under review independent of one another: 1) the 88 pitchers who faced at least 150 batters in the 2013 New-York Penn League (NYPL) season and averaged at least 10 batters per game, and 2) the 58 pitchers who faced at least 150 batters in the 2013 Appalachian League season and averaged at least 10 batters per game.
Data Under Review
All plate appearances against said pitchers in the corresponding league in 2013, excluding bunts and foul outs.
As described earlier, the average effect on run expectancy of all plate appearances versus each study group pitcher in their league in 2013 was determined assuming that each of the 12 event types listed in the earlier table always had the same effect on run expectancy (as specified in the table above for the NYPLers, or as was specified in the table published in the prior Appalachian League post for the Appy Leaguers). After determining the mean and standard deviation (SD) of those 58 or 88 values, each pitcher is assigned a Performance Score where 50 denotes league-average performance (equal or better than 50% of peers), 60 denotes 1 SD better than league-average (>= 83% of peers), 70 denotes 2 SD better than league-average (>= 97% of peers), 40 denotes 1 standard deviation (SD) worse than league-average (>= 17% of peers), 30 denotes 2 SD worse than league-average (>= 3% of peers), and so on with 10 points essentially being 1 SD. That Performance Score can be resolved further down into a Control Subscore (based on walk plus hit-by-pitch rate per non-foulout, non-bunt plate appearance), a Strikeout Subscore (based on strikeout rate per non-foulout, non-bunt plate appearance), and a Batted Ball Subscore (based on summing the assumed average effect on run expectancy of each batted ball per its general event type and dividing that sum by the number of batted balls); each subscore is also expressed on a 20-to-80 scale with the ones above 50 being better than league-average. Finally, an Age Score is computed which assesses how young the pitcher is relative to the league study group on a 20-to-80 scale (a score above 50 indicates younger than average).
New York-Penn League: Tri-City ValleyCats
In this and future tables, green text denotes a number that beat league-average by at least 1 SD and red denotes one that trailed league-average by at least 1 SD.
Michael Feliz graded out as the 2nd best pitcher in the league per his Performance Score, trailing the leader (the much older Cole Sulser of the Indians' affiliate) by a few tenths of a point. Note that Feliz excelled at strikeouts (almost 2 SD above average) without sacrificing anything in the batted ball department where he also happens to thrive - that is a very rare combination to see in a pitcher and especially for one of his age relative to competition level (91st percentile on youth). Kyle Westwood ranked 7th overall and fits the profile of strike thrower plus groundballer extraordinaire; it would be premature to dock the 2013 draft 13th-rounder's performance on account of age given that he competed at the level where an Astros' four-year-college draftee would usually begin their pro career. Evan Grills put himself back on the organizational prospect ranking charts with a strong 2013 short-season campaign; despite taking a bit of step backwards from relieving in 2012 full-season A ball he managed to rate relatively young in the 2013 NYPL season by league starter standards. 2012 Appy Leaguers Kevin Comer and Adrian Houser graded out as nearly average across the board in 2013, which is not necessarily a bad thing as both still have some degree of youth on their side per their respective Age Scores. The debut of 2013 second-rounder Andrew Thurman was a bit of a mixed bag. The good news was that he racked up strikeouts thanks to his excellent change-up; the bad news was that his fastball got hit hard at times and especially so by righties who frequently turned on the pitch and drove it in the air. Thurman would seem to be in a similar boat to the one that the system's most well-known changeup artist, Nick Tropeano, was in following his short-season debut; a major-league caliber change-up is a nice asset to sport, but Thurman will have to improve the quality and placement of his fastball (if not also do the same with a second secondary pitch) if he is to succeed against more advanced hitters and the righthanded-batting ones in particular. Tanner Bushue retired just prior to the end of the season. 2013 29th-rounder Randall Fant beat just 5 of the 88 pitchers in Performance Score. Squint below to espy a lot more detail on each hurler's campaign.
Appalachian League: Greeneville Astros
Jordan Mills and Chris Lee ranked as the 9th- and 10th-best Performance Scorers of the 58 arms evaluated and as the 2nd- and 3rd-best southpaws of the dozen in the study group. The combination of Mills' age and that he debuted at Greeneville rather than Tri-City makes it harder to get a read on a starting pitcher like him who was just drafted (28th round) out of a smaller-time Division I program (a bit like Daniel Minor from the 2012 draft class). Lee missed nearly all of the 2012 Greeneville campaign, so his age relative to level doesn't detract quite so much from his very good 2013 rebound performance. 2012 Greeneville holdover Frederick Tiburcio followed immediately on the heels of Mills and Lee and graded out similarly to an older righthanded version of them with batted balls again being the strong suit. Edison Frias was very good at control but poor at batted balls, leading to a slightly below-average Performance Score. On the plus side, Jandel Gustave can reach the upper 90s with his fastball and he was able to parlay that into a well-above-average strikeout rate. On the down side, Gustave remains a bit wild and his batted ball profile was a bit too rich in line drives and pull-third outfield flyballs and those two shortcomings dragged his Performance Score down below league-average. Enderson Franco is far too similar of a name to Edison Frias, and he'll need to perform better in 2014 and beyond to perpetuate the organizational confusion.
Faced At Least 100 Batters, And Here's How He'd Score If He Had Faced 150+ With Equivalent Results
Morton (who also threw another 18+ innings for Quad Cities that aren't accounted for in the table), Cotton, and Chrismon were chosen in the 32nd, 14th, and 26th rounds of the 2013 draft, respectively. Scribner, the brother of occasional major league reliever Evan Scribner, was signed as an undrafted free agent this summer and broke in with the Gulf Coast League Astros shortly thereafter (those numbers aren't included above). Frias' weighted score for the 2 leagues has been thrown in for good measure.