That Great Young Pitcher

              A week ago I tweeted out my list of the top 10 starting pitchers in baseball.   There are two reactions that I always get from people when I send out the list.   One set of people, who we could loosely describe as gamblers, disagrees with the list (usually rudely, because it is twitter) on the basis of expectations for the next start.    The gamblers believe that since Michael Fulmer or Jacob DeGrom may have a lower expectation for runs allowed in his next start than Justin Verlander, that Fulmer or DeGrom should be on the top 10 list rather than Verlander.   But these people misunderstand what the list is about.   It isn’t based on expectations for the next start; it is based on the expected value of the pitcher, over time.    It isn’t an answer to the question "Who is most likely to pitch well in his next start?"   It is an answer to the question "Who is likely to have the most value to his team, over the next season or the next two seasons?"

              If I was a gambler, I’d care about the same things that the gamblers care about, and I’d probably get the same list that they have.   The other reaction, from fans of different teams in different divisions, is that Aaron Nola should be on the list, or James Paxton, or, a month ago, Robbie Ray.    The guys like Michael Fulmer and Jacob DeGrom aren’t on the list, but they are CLOSE to being on the list; DeGrom is 12^th on the list and will probably soon pass Jon Lester to move into 11^th.   The guys like Aaron Nola and James Paxton aren’t on the list, aren’t anywhere NEAR being on the top 10 list, and should not be on the top list.  

              Why?

              Because they get hurt.    Most young pitchers get hurt.   Most young pitchers at the start of their careers don’t know their limitations and don’t respect their limitations, so they dominate for half a season, try to do something they should not be trying to do, and then they get hurt.   The most valuable guys are the guys who have proven that they can do the job over a period of time.  I don’t understand why people don’t get that.   It seems really obvious to me. 

              I’ve been sending out periodic tweets about the best starting pitchers in baseball for a year or so now and getting these responses, and this process has hardened my stance on this issue.   At the start of the year, when I sent out my list, THE GUY that people thought should be on the list was Noah Syndergaard.   Noah got hurt, of course, so then it was Dylan Bundy for a while, but then he got tagged a few times, then it was Robbie Ray, and then it was James Paxton.  Now it is Aaron Nola.   I figure Nola will probably get hurt some time in the next six weeks, but don’t quote me.  

              Look, I am not saying that young pitchers NEVER stay healthy or that young, inexperienced pitchers are never going to be the best bets moving forward.   What I am saying is that experience should teach you to be careful about them.  

              So I got to thinking about this. . . what percentage of young pitchers DO take a big step backward after their first good season?   Study idea. . . what percentage of young pitchers have a significant step backward after their first good season?  

              It’s 57 or 58%.    First of all you have to decide what is a "good season" for a starting pitcher, which means that you have to draw an arbitrary line somewhere.    Second, you have to mark some season as the pitcher’s FIRST good season, which is a process that is surprisingly tricky although it sounds easy.    And third, you have to decide what represents a significant step backward.   You make up rules for that, and sometimes those rules work, and sometimes they don’t exactly work the way you wanted them to.   I’ll deal with all of those issues at the end of the article, in the part of the article called "process".   

              But jumping toward the work product, let’s take the 1961 season.   In 1961, my first year of being a baseball fan, there were 14 starting pitchers who had their first good seasons in the major leagues:   Steve Barber, Don Cardwell, Joe Gibbon, Bob Gibson, Mudcat Grant, Joey Jay, Sandy Koufax, Jack Kralick, Jim O’Toole, Juan Pizarro, Ray Sadecki, Don Schwall, Bill Stafford and Ralph Terry.  

              There is already a problem with that list.   A pitcher named Ken McBride, with the expansion Los Angeles Angels, made 36 starts, went 12-15, struck out 180 batters with a 3.64 ERA.   I would have to say that’s a pretty good year.   My judgment, that’s a good year for a starting pitcher.   We can’t do the study that way, of course, and the system that I set up says that it isn’t quite good enough, so.. .no Ken McBride.

              Anyway, of the 14 starting pitchers who had their first good seasons in 1961, eight, in the judgment of the method, had significant declines in performance in 1962.    Five of those no one would quarrel with.   Certainly Don Cardwell, Joe Gibbon, Mudcat, Sadecki and Schwall had significant declines in performance in 1962:

Juan Pizarro. . . well, you can argue it, but he had fewer wins, twice as many losses, more walks, fewer strikeouts, and an ERA 76 points higher: 

              More troublesome is Jim O’Toole:

              I would be happier if O’Toole DIDN’T register as having a significant decline in performance, since. . .well, it doesn’t seem all that significant.   And then there is Steve Barber:

              Barber pitched with the same effectiveness in 1962 as 1961, but, as I recall, was in the military reserve all season, doing training, and was only able to pitch when the Army didn’t need him.  Still, what we are asking is how many pitchers take a step backward after their first good season as a starter, and 1961 was Barber’s first good season as a starting pitcher, and 1962 was a step backward for him, so. .  ..no special pleadings.    The other six starters who had their first good season in 1961 (Bob Gibson, Joey Jay, Sandy Koufax, Jack Kralick, Bill Stafford and Ralph Terry) did not lose value in 1962; you can take my word for it or look them up yourself, I guess.   I am just trying to make sure that you understand what I have done in this study to reach the conclusions that I did.  

              OK, there are 1,337 pitchers in history who have at some point had their first good season as a starting pitcher.    However, ten of those did so in 2016; those aren’t useful to us, because at this point we can’t say for sure whether they did or did not have significant declines in performance in 2017, so that cuts us down to 1,327

              Of those 1,327, 718 (or 54%) had significant declines in performance in the following season.   These numbers, however, include the 19^th century, which wasn’t actually major league baseball, and the Dead Ball era, which was so remote in time from the present as to be not very instructive.   This chart breaks down the number of starting pitchers having their first good season by decade, more or less:

              There were 107 starting pitchers having their first good season in the 1980s, and 86 in the 1990s.   The number probably dropped in the 1990s because the 1994-1995 strike prevented a certain number of pitchers from reaching the standards necessary to qualify as a good season.   Then we went back to 113 in the 2000-2009 era, and seem likely to pass that total in our present decade.  

              In the 19^th century, a starting pitcher having his first good season pitched an average of 368 innings.   This makes all of the other averages for a 19^th century pitcher so different from the modern game as to be irrelevant.   The average number of innings pitched by a starter having his first good season declined to 261 in the 1900-1909 decade, and to 242 from 1910 to 1919.  

              The average performance of a starting pitcher having his first good season has not really changed very much since 1950.    The average won-lost record for these pitchers was 17-11 in the 1920s, 17-10 in the 1930s, 16-10 in the 1940s, 15-10 in the 1950s and 1960s, 14-10 in the 1970s, 15-9 in the 1980s and 1990s, 15-8 in the last decade, and 14-8 in the current decade:

              The winning percentage has gone up slightly over time because, as pitchers make fewer starts, a better performance is required to qualify as having had a good season.   Strikeouts have gone up; walks down.   Complete games within this group of pitchers have declined by more than 90%:

              In the 1950s first-year starting pitchers typically made 27 starts and 8 relief appearances, because in most cases they moved from the bullpen into the rotation.   Now first-year starters typically make less than one relief appearance, as starters and relievers are no longer interchangeable, but are now regarded as separate species, housed in different cages in the local zoo.  

              In terms of ERA relative to the league, there has been little change; that’s been about 83% of league ERA all along.   Strikeouts per nine innings have almost always been 3 to 5% above league norms.  Since the Cy Young Award was introduced (1956) about 1% of starters having their first good season have won the Cy Young Award, or about one in each decade except the 1980s, when four starting pitchers won the Cy Young Award in their first good season (Fernando, Clemens, Saberhagen and Steve Stone.   1980 was the only season in which Stone qualifies as having had a good season, by our standards.) 

              From 1910 to 1919, 35% of starting pitchers having their first good season were also rookies.  This number has dropped steadily over time, so that in the current decade it has been only twelve percent.   Very few rookies anymore move into the starting rotation at the start of a season and stay there throughout the season.   This chart tracks the percentage of starting pitchers having their first good season who were destined for the Hall of Fame, by decade: 

              There were 119 first-good-year starters in the 1960s, and 9% of them are in the Hall of Fame, so that’s eleven.   Can you name them?  (Nosbig, Xafuok, Revaes, Lahciram, Yrrep, Retnuh, Notlrac, Orkein, Sniknej, Nottus, Remlap). 

              OK, I have stalled as long as I can.   Through 2015 there were 1,327 starters in history having their first good year, as I mentioned, and 718 of them had significant declines in performance the next season.  These numbers, however, include the 19^th century and other seasons so remote in time from the present as to be not believably relevant.   This is the full chart by decade:

              Basically, 57 or 58% of starting pitchers have had significant declines in performance after their first good seasons.   The percentage has been a little bit lower in this decade, but (a) that’s a small enough sample that it’s probably just random, and (b) of the ten pitchers who had their first good seasons in 2016, at least seven have are having not-very-good seasons, so that’s going to push us back up.  

              Significant declines; that’s the key.   If 50% of players declined in a season, that would be nothing; there’s a baseline, half the guys go up, half go down, doesn’t mean anything.   But if 57% of players have significant declines, that’s a very different thing.   A few of those may be borderline cases like Jim O’Toole in 1962, but the majority of them are NOT borderline.   The great majority of them are serious declines—like Aaron Sanchez, Kyle Hendricks or Steven Wright this season.  

              Oh, I think I forgot to give you the age spectrum of starting pitchers having their first good years.   In the 19^th century starting pitchers having their first good season were, on average, 23.5 years old.   This number went up steadily through the 1940s:

              And then nose-dived in the 1950s and 1960s:

              People will think that the low figure in the 1960s is due to expansion, but it definitely is not.   Expansion actually drives UP the average age of rookies, as an expansion pushes into the major leagues a certain number of players who have been trapped in the minor leagues, and into the starting rotation a certain number of pitchers who previously were in the bullpen.  What actually happened was the Kiddie Corps.

              Over time, baseball men had adopted a set of ideas about developing starting pitchers.   Starting pitchers in those days threw a great variety of pitches, and from different motions.    It was ASSUMED, about a starting pitcher, that he would be able to drop down and throw sidearm against a same-side hitter.   It was assumed that he could throw a fastball, a curve, a slider, a change (which was then called change-up), and usually some other pitches.   As late as 1958 many, many pitchers threw a knuckleball as a part of their repertoire.   If a pitcher couldn’t do these things, he wasn’t ready to pitch in the major leagues.  

              Paul Richards, managing Baltimore, had a different idea.   His different idea was that, because we require a pitcher to be able to do all of these different things, we’re keeping starting pitchers in the minor leagues until their fastball is gone.   Maybe overstated, but let’s say that a pitcher has his best fastball from the time he is 20 until he is 28; why do you want to keep him in the minors until he is 26?   If he’s got two pitches to work with and a good fastball, get him to the major leagues, tell him to challenge the hitters, put a good defense behind him and he’ll be OK.

              In 1960 the Orioles’ starting rotation was Milt Pappas, Steve Barber, Chuck Estrada and Jack Fisher.   Estrada was the old man of the group; he was 22.   The others were 20 or 21.   Jerry Walker also made 18 starts for them; he was 21, and then they had a couple of older guys who made spot starts as well.    They were known as the Kiddie Corps.

              This actually started in 1959. . ..I’m simplifying.   Also, the Orioles weren’t on an island here; there were other organizations which were experimenting with younger starting pitchers at the same time, although not to anything like the same degree.  Anyway, the Orioles led the league in ERA and Complete Games in 1960, led the major leagues in ERA by a wide margin in 1961, and led the American League in ERA again in 1962.   It was an "Oh, shit" moment for organizations around baseball.   They realized that young, inexperienced starting pitchers COULD pitch in the majors, if they had a fastball and control.   

              Well, you know how baseball imitates whatever is working, right?  That was what was working.   So the average age of a first-year starting pitcher dropped by two or three years overnight.   

              Since then, we have sort of stabilized in the middle, trending slowly upward:

              So that brings up an issue:  Is a younger starting pitcher having his first good season more likely to have a relapse the next season, or an older starting pitcher?

              You can look at it many different ways; you can make up reasons why it should be one or the other.   But the answer is that an older starting pitcher, having his first good year, is more likely to have a relapse in the next season.   Of pitchers aged 19 to 22 when they had their first good season in the majors, only 52% had a serious decline the next season; ages 23 and 24, 53%.   But of pitchers aged 29 and up, 66% of starting pitchers had a serious decline after their first good season.   This is based on the pitchers since 1950, group averages:

              At this point I haven’t addressed a large concern with the study.    The percentage of successful first-year starting pitchers who have a significant decline in production in the second year is certainly large, you might think.   But is it actually larger than the percentage of second-year or third-year starting pitchers who have significant declines in production?

              It’s a critical question, because what I am essentially saying is that Justin Verlander should rank ahead of James Paxton because Justin Verlander has proven that he can carry the load of being a starting pitcher year after year, while James Paxton has not proven that.   But this requires that we document that the relapse right for pitchers having their first good season is higher than for veteran, established starting pitchers.   But is it really?
              Yes, it is higher. . . not all that much higher, but higher.    In all of baseball history through 2015 there have been 1,327 pitchers having a first good season, 906 having a second good season, and 642 having a third good season.   Through all of history, 54% of first-year pitchers had a significant decline in value after their first good season, while only 48% had a significant decline in the season after their second good season, and only 44% had a significant decline in the season after their third good season.   Focusing on the seasons since 1950, the decline rate is 57% after the first good season, 52% after the second good season, and 47% after the third good season.   So the washout rate remains very high, but it is higher after the first good season than after a pitcher is more established.   Therefore, it is reasonable to have higher expectations for a veteran pitcher than for a less established pitcher. 

Commentary

              In a sense, my study doesn’t QUITE get to what I believe to be true.   What I believe to be true is that the performance of a young pitcher, a pitcher in the Aaron Nola/James Paxton/Robbie Ray/Luis Severino/Jordan Montgomery/Paul Blackburn/Eduardo Rodriguez/Jake Junis portion of his career, should not be regarded as equal to that of a veteran starting pitcher, and I believe this to be true because I believe that it is likely that, before he becomes an established starting pitcher, Aaron Nola is going to hit a bump in the road.   Tom Seaver never did and Don Drysdale never did, but Bob Gibson did and Sandy Koufax did and Roger Clemens did and Randy Johnson did and Jim Palmer did and Max Scherzer did.   More do than don’t.  

              My study does not disprove this belief or deny it, but I believe it to be MORE true than the study shows.   I studied pitchers who had a COMPLETE (or more or less complete) good season.   Most of these guys hit that bump in the road BEFORE they have that complete good season.   I wonder, in retrospect, whether I shouldn’t have studied game logs, rather than career records.   My game log records don’t have pitchers ages in them, but I could have studied by career games started, perhaps.   I could have studied pitchers who had a streak of very good games in their first 40 career starts versus those who had a similar/equal streak of very good games after they had made 100 career starts.   In retrospect, I wonder if that might have been a better way to study the issue. 

              Explanation of the process follows. . . thanks for reading. 

Process

OK, the process of doing this study involves three essential steps:

1)  Designate those seasons which qualify as good seasons by starting pitchers,

2)  Identify the FIRST good season by each starting pitcher, and

              3)  Decide whether or not that starting pitcher had a significant drop in performance in the following season. 

              I designated as a good season by a starting pitcher any season in which:

              1)  A pitcher made at least 15 starts,

              2)  The pitcher made at least 40% of his game appearances as starts, and

              3)  The pitcher had a Season Score of 150 or higher.

              Why 40%, rather than 50%?    If you mark it at 50%, Ernie Broglio in 1960 isn’t considered a starting pitcher.   Broglio made 24 starts, pitched 226 innings, went 21-9 with a 2.75 ERA, pretty obviously a starting pitcher—but he made 24 starts, 28 relief appearances.   A full-time reliever typically pitches 60-70 times a season; a starter, 30 or 35.   You can’t treat starter and relief game appearances one-to-one.    

              I’ll have to explain Season Scores.   The first thing you do to figure a Season Score is to figure the leverage index, or "CLI" as I call it, for "Crude Leverage Index".   The Crude Leverage index is

              Innings Pitched  +  (1.5 * (Games Finished + Saves))

              --------------------------------------------------------------------

                           &n​bsp;               Innings Pitched

              So that if a pitcher has 60 game appearances, 25 saves and 35 games finished, then his CLI is 2.50.     

              But basically, the CLI is mostly irrelevant to this study, since this is a study of STARTING pitchers, and starting pitchers have CLI of 1.000.   It’s not usually a factor; I just have to explain it because it is part of the system.   Occasionally it matters.   Jim O’Toole calculates as having a significant decline in performance in 1962 because, in 1961, he also had a couple of Saves.   Without those Saves, it’s not a significant decline in performance.  

              Next you take the outs recorded by the pitcher, multiply that by .425, subtract runs allowed, and subtract earned runs allowed.    Basically, what you are doing is figuring how many runs the pitcher saved as opposed to a pitcher with a 5.50 ERA.   Runs Saved above replacement, basically.   Let’s say a pitcher pitches a shutout; 27 outs, no runs.   Multiply the 27 times .425, you have 11.475.   His result is 11.475.   If he allows one run, which is earned, then he would be at 9.475; two runs, earned, he would be at 7.475; three runs, earned, at 5.475; four runs, earned, at 3.745; five runs, earned, at 1.475; six runs, earned, at negative .525.   If he allows six runs, one UN-earned, he’d be at negative .025.  

              So anyway, that number you multiply by the CLI.    To this, you add eight points for a win, subtract five points for a loss.   Subtract his walks allowed from his strikeouts, divide by five, and add that, and then add his saves.    Basically, all we are doing here is taking the pitcher’s basic stats—wins, losses, saves, strikeouts, walks, innings pitched and ERA—and putting them all on a common scale, asking "how good are these numbers?"

              A pitcher who has a Season Score of 150 (and was a starting pitcher) was considered for this study to have had a good season as a starting pitcher.   As to what exactly that means, there have been ten starting pitchers since 2000 who have had Season Scores of 154.5 to 155.499.   These are those ten pitchers:

              Jacob deGrom in 2014 went 9-6 with 3.66 ERA, 144 strikeouts in 140 innings; that scores at 155.   Tim Wakefield in 2005 went 16-12 with a 4.36 ERA, 151-68 strikeout to walk ratio in 225 innings; that also scores at 155.   These qualify as good seasons by starting pitchers.  These seven pitchers, on the other hand, score at 145:

              If you are thinking that the second group of pitchers look an awful lot like the first group, well, yes, they do; that was my point.   There isn’t much difference between the two groups of pitchers, but you have to draw a line somewhere.   The first group of pitchers pitched an average of 184.7 innings, had a winning percentage of .588 and an ERA of 3.56, with a strikeout/walk ratio of 148-51.   The second group of pitchers pitched an average of 184.8 innings, had a winning percentage of .570, an ERA of 3.82 and a strikeout/walk ratio of 145-58.   They averaged more wins than the first group (12.9 to 12.0), but had a higher ERA, a lower winning percentage, more walks and fewer strikeouts.   The differences are small, but you have to draw the line somewhere, and that is where I drew the line—halfway between those two groups of pitchers.  

              OK, so now we know whether a starting pitcher has had a good season, but how do we know if it is his "first" good season?    It has to be the first season in which the pitcher has a season score of 150 as a starting pitcher, but also, it has to be the first season in which he has had a season score of 150, period.   A few pitchers have had a season score of 150 as a reliever, then later did so as starters.   He doesn’t qualify as a first-good-season starter for purposes of this study.

              I’m trying to study emerging starting pitchers here.   If a guy has been around a few years as a reliever and then moves into the starting rotation, that’s a different thing.    Also, I didn’t consider a pitcher eligible for this study as a first-year quality starter unless his season score was at least 10% higher than any previous season.    If a pitcher is just a little bit better one year than he was the year before and strays over the qualifying lines, that’s not really what I am talking about.  

              Finally, how do we decide whether the pitcher has had a significant decline in performance after his first good year?

              Well, first, he has to have a 25% decline in his season score.   If his season score drops from 250 to 180, that’s a 28% decline; he qualifies.   If his season score drops from 250 to 190, that’s a 24% decline, so he doesn’t qualify.  

              Second, no pitcher is marked as having a significant decline in performance if he has a season score of 200 or more.    Dwight Gooden is 1987 was 15-7 with a 3.21 ERA, a season score of 201.   Yes, that represents more than a 25% decline from his 1986 season (17-6, 2.84 ERA), but I’m just not comfortable saying that a pitcher who is 15-7 with a 3.21 ERA has had a serious decline in performance.  

              Third, a pitcher has to meet one of two standards.    He has to have either an increase in his ERA, relative to the league, or he has to have a 40% or larger decline in innings pitched.   If his ERA doesn’t go up relative to the league and his innings pitched are kind of the same. . .well, that’s not a significant decline in performance, regardless of the season score.