Clayton Kershaw and Cross-Era Comparisons

Clayton Kershaw is, pretty indisputably, the best pitcher in baseball. He strikes out as many hitters as Max Scherzer or Chris Sale while walking as few as latter-day Bartolo Colon, all while allowing roughly half as many home runs as any of those guys. Most likely, once Kershaw returns from the back injury that has sidelined him since June, he will continue to be the best pitcher in baseball.

What if he doesn’t, though? When he first went on the DL, or when L.A. announced he was being shut back down a few weeks later when he couldn’t throw without pain, the question must have crept into the backs of fans’ minds: What if he is never the same? What if he never recovers to pitch again at all?

Of course, this is a ridiculous suggestion. Sure, it’s possible for any pitcher’s career to suddenly buckle under the strain of pushing his arm to speeds that defy even German autobahn standards, but there is no reason to believe Kershaw’s injury is particularly dire. He’s already begun throwing again, and all indications are he is recovering well and should be back soon enough. There’s no reason to consider his demise as a serious possibility, at least not more than any other pitcher.

No, what I’m really asking, in an unnecessarily fatalistic way, is, what if Clayton Kershaw were Sandy Koufax?

It’s a natural comparison—two Dodger lefties dominating hitters with devastating curveballs, each turning would-be batsmen into hapless bystanders like almost no other left-handed starter (Randy Johnson breaks pretty much every comparison) in the decades spanning from one to the other. Kershaw has been evoking Koufax’s name from fans and writers for years already—even the great Vin Scully has gone there:

What’s particularly fascinating, though, is that at the time of Kershaw’s injury, he not only sort of reminded people of Koufax. He practically was Koufax, at least as far as a quick glance at his numbers go.

Since 2009, Kershaw has allowed 448 runs. He’s struck out 1,791 batters and, if we include hit batsmen and exclude intentional walks, allowed 427 free passes. He’s allowed 1,168 hits and 92 home runs.

Those numbers exclude Kershaw’s rookie season in 2008. While it’s impressive in its own right that, at age 20, Kershaw was already an average major league starting pitcher, he didn’t really become Clayton Kershaw as we know him until his second season. And when he did, he quickly transformed from a promising youngster into what a younger generation might only describe as “the left arm of god.”

Since that transformation, Kershaw has thrown 1.624.1 innings. If you ignore the first half of Koufax’s career–when he was roughly average–and likewise focus on only the six seasons after he learned how to channel the stuff Al Campanis once compared to the work of Michelangelo, he threw 1,632.2 innings. Here’s how they compare over those stretches:

It’s remarkable how similar their numbers are. Kershaw has slightly more strikeouts and fewer home runs, and Koufax slightly fewer walks, but they’re in roughly the same neighborhood across the board. Koufax allowed a slightly lower wOBA, but Kershaw makes up for it by inducing more double plays and controlling the running game better. The Dodgers did win more often with Koufax on the mound, but Koufax also got better run support than Kershaw.

Add it all up, and Koufax allowed 11 more runs in 8.1 more innings. If Kershaw retired today, he would essentially be Koufax, minus the mediocre half of his career.

Thinking about it for a second, though, the contexts for these numbers are actually quite different. Kershaw is pitching in a time when pitchers have to worry about seven- and eight-hitters (and nine-hitters if it’s Madison Bumgarner) hitting the ball over the fence and power hitters like Giancarlo Stanton hitting it completely out of Dodger Stadium. He’s pitching in the same park as Koufax but only after the plate was moved 10 feet toward the outfield and the walls lowered a couple feet. And yet Kershaw still has allowed fewer home runs than Koufax.

For his part, Koufax was pitching at a time when 200-strikeout hitters came up only if you were talking about career totals (or possibly speculating about what would happen if somebody let Dave Nicholson play every game for a year). That he still struck out almost as many hitters as Kershaw is just as amazing.

And Koufax did that while routinely facing those (relatively) Sewell-esque hitters four times a game and having to pace himself to go well past 100 pitches on any given night. In fact, Koufax actually struck out more hitters than Kershaw the first time through the order:

So, while they appear nearly identical at first glance, once we account for their eras, we see their strengths start to diverge. Kershaw was, in fact, significantly better at preventing home runs, while Koufax had the more exceptional strikeout ability.

This, I think, is where the Koufax/Kershaw comparison really gets interesting. It’s not that they are so similar in their dominance. It’s how, when we dig deeper, we can start to see and appreciate the nuances their different contexts bring. Precisely because their raw stats are so similar, how you compare one to the other comes down entirely to how you compare their respective eras.

League Averages and Cross-Era Comparisons

A lot has changed since 1966. The mound has been lowered. Hitters have gotten more powerful. The strike zone has changed. Teams have shifted to five-man rotations and rely much more heavily (or at least more regularly) on relievers to finish out games. Dodger Stadium has grown less cavernous. You could go on and on.

The typical panacea for these types of changes is to compare players to their contemporaries. Koufax, for example, struck out hitters about 75 percent more frequently (on a per-plate appearance basis) than the average pitcher from 1961 to ’66. Dodger Stadium appears to have inflated strikeouts slightly, so applying a park factor drops that down to 73 percent, which we’ll call a K+ (as in ERA+ from Baseball-Reference or ERA- from FanGraphs) of 173.

If we convert Koufax’ and Kershaw’s raw stats to “plus” or “minus” type stats by dividing their rates by the league average and park-adjusting, we get the following (100 is average, higher is better for “plus” stats, lower is better for “minus” stats):

(A quick note on the park adjustments: As mentioned above, Dodger Stadium was considerably more pitcher friendly in the 1960s than it is now. However, the park adjustments don’t make as big a difference as you might think, because Koufax also pitched 1961 in the Coliseum, which featured a left field wall 250 feet from home plate.)

This is the heart of most cross-era comparisons: Who stood out more relative to his peers? This process works well adjusting for things like the lowered mound or a smaller strike zone, which affect everyone pretty consistently. Koufax had the advantage of pitching shoulder-high strikes from a 15-inch mound, but so did every other pitcher in the league, so it doesn’t really matter when we compare him to the league average.

(Actually, different stadia had different mound heights prior to 1969, and Dodger Stadium in particular was notorious for having a mound higher than the 15 inches allowed by MLB rules, so the above is not strictly true. However, as long as you use park adjustments, it still shouldn’t matter much since that will be captured in the stadium’s park factor.)

But what about changes that don’t affect the whole league uniformly? The problem is, league average is not an immutable constant. We use it because it is a convenient comparison point, but that doesn’t mean it is without its flaws.

Changes in League Quality

There are many reasons the average pitcher in one season might not be equivalent to the average pitcher in another, but the simplest way to see this is with expansion. In 1960, Koufax was one of 238 different pitchers to appear in at least one major league game. In 1962, after the league had expanded to 20 teams, he was one of 304. Over the span of two years, there were suddenly dozens of new pitchers who otherwise would have been in the minors who now were being added into the major league average.

Because of that, it’s very likely the average pitcher in 1962 was at least a bit weaker than in 1960. What’s not so clear, though, is whether that short-term decrease in league quality holds over the long term, because there are other factors that can offset the effects of expansion.

While baseball has been popular in parts of Latin America since the 1800s, the pipeline for that talent to reach the majors remained fairly primitive through much of the 20th century. The baseball academies that have turned countries like the Dominican Republic and Venezuela into scouting hotbeds didn’t open until the 1970s, and they didn’t start affiliating with major league organizations until the ’80s. Increased scouting and development efforts in those countries have produced a massive influx of talent to fill the extra jobs created by expansion.

Likewise, there was virtually no pathway for Asian talent into the major leagues until the signings of Chan Ho Park and Hideo Nomo (and, bizarrely enough, Alfonso Soriano) in the 1990s opened up changes to the working agreement between MLB and the Japanese leagues, which previously had prevented American teams from signing anyone already playing professionally in Japan.

These two factors affect the overall league quality, which leads to a stronger or weaker major league average. The question, which we’ll have to address if we want to adequately compare Kershaw to Koufax, is whether the watering down from expansion or the influx of new talent from international expansion and population growth has a greater effect.

Another major factor that has altered the meaning of major league average is the evolution of the bullpen. There are now two entirely distinct classes of pitchers, one of which is devoted solely to coming in and putting 100 percent effort into every pitch for a handful of batters before leaving the game. That approach has a huge effect on those pitchers’ stats, to the point that bullpens routinely allow fewer runs per inning pitched than starting rotations despite being made up mostly of pitchers who weren’t quite good enough to make the starting rotation.

And when you have pitchers who actually are elite talents coming out of the bullpen, they do stuff that would give Campanis visions of all sorts of Renaissance artists, like Craig Kimbrel posting a 1.59 RA9 over 289 innings in his five years with Atlanta—a figure that’s lower than any starter has posted in a season since Bob Gibson in 1968, when the whole National League was scoring just 3.4 runs per game.

This creates a benefit that applies only to relievers, so it distorts the major league average Kershaw is being compared against. After all, part of the reason strikeouts are up so much is that there are guys like Kimbrel and Aroldis Chapman who can come in out of the bullpen and strike out 15 hitters per nine innings.

Then there is also a flip side: Changes in bullpen usage mean that Kershaw seldom has to go through the order a fourth time, so he is being compared to other starters who are expected to last right around the same 100 pitches he is.

Those things weren’t true for Koufax, who routinely had to pace himself to last an entire game. Koufax also was being compared to starters who were being pulled well before 100 pitches any time things weren’t going well (which still happens occasionally, but it happened much more frequently in the 1960s).

These are all things we need to consider if we want to properly compare someone pitching in the 2010s to someone pitching in the 1960s. Basically, while comparing players to major league average certainly helps, it is not necessarily the panacea we’d like it to be.

When a league expands and several new jobs open up to players who would otherwise be in the minors, the average player is suddenly a bit worse than before. When new avenues of talent open up or expand, such as with population growth or an increase in international development and scouting, the quality of the average player starts to go up. And when a substantial change affects some pitchers but not others, such as the emergence of bullpen specialists, it can distort how those pitchers factor into the league average.

The Impact of Bullpen Usage

To start, let’s tackle the issue of evolving bullpen usage. We can reduce the impact of these changes by ignoring relievers when we calculate the league average and instead comparing Kershaw and Koufax only to other pitchers in the same role. Here’s how their stats compare to the average starter rather than the average pitcher:

The overall changes aren’t major—both improve a bit, Kershaw slightly more so than Koufax—but Kershaw’s K+ jumps significantly when you compare him only to other starters. This is, of course, what we’d expect, given the most eye-catching numbers relievers put up tend to be strikeout numbers.

This still doesn’t account for Koufax consistently going deeper into games, though. Recall that Koufax was facing batters for the fourth time almost every game. About 16 percent of the batters he faced were on at least the fourth time through the order, compared to less than five percent for Kershaw.

That wouldn’t matter too much if everyone were doing so, because it also would factor into the league average Koufax is being compared against. That’s not the case, though. While most regular starters would pick up at least a few complete games per season, typically only a team’s ace or a good No. 2 was getting to double digits in complete games. Koufax threw 54 of them in his final two years alone.

As a result, comparing Koufax to his peers doesn’t fully account for him having to go deeper into games than Kershaw. We need to go a step further and compare them not just to what other starters did on average, but to what other starters did each time through the order.

The following table shows, on average, how much better or worse a starting pitcher performs each time through the order. This is calculated by comparing each pitcher’s numbers each time through the order to that pitcher’s overall numbers. For example, we saw earlier that Koufax struck out 26.9 percent of batters overall, but that mark was only 20.3 percent the fourth time through the order, which would be a drop of 0.066 strikeouts per PA.

Strikeouts are most affected, with the average starter during Kershaw’s era seeing his strikeout rate drop about 5.3 percentage points from his first time through the order to his fourth. During Koufax’s era, the drop is even bigger at 6.0 percentage points. Walks remain relatively flat, and home runs go up moderately as the batting order repeats itself.

(The home run rate does go back down the fourth time through the order in Kershaw’s era, though this is probably because it’s difficult to get a representative sample of pitchers going four times through the order in recent years. Because pitchers are left in the game long enough to go four times through the order only when they are pitching extremely well, there is potentially a bias affecting the numbers for the fourth time through the order.)

We can use these numbers to adjust our ratings for Koufax and Kershaw. First, we assign the average pitcher the same distribution of plate appearances as those two had each time through the order. This means increasing the number of times they have to go three or four times through the order. Then, we see how much we would expect their performance to drop by having to face the same hitters more often each game.

This doesn’t have much affect on Kershaw’s numbers, because, as noted above, most starters in Kershaw’s era last about 100 pitches per game, give or take a quick hook here and there. It does make a difference for Koufax, though, particularly in his strikeout rate.

I didn’t include RA9- in this table because runs allowed don’t really break down cleanly between times through the order due to the interactive nature of run-scoring events (i.e., if the ninth-place hitter gets on base and is then driven in by the leadoff hitter, which time through the order do you assign the run to?), but the improvement to Koufax’s numbers is worth roughly a couple points off his RA9-. Overall, this is comparable to the effect of removing relievers from the league average on Kershaw’s numbers.

Expansion, Population Growth and International Talent

So far we’ve seen that, while Koufax was the better strikeout pitcher, Kershaw’s performance relative to his peers is a bit stronger overall, just because he put up similar raw numbers in a higher-scoring environment. Adjusting for modern bullpen usage makes both look even more dominant but doesn’t do much to shift their standing relative to each other.

The big question with these sorts of comparisons, though, is always this: Whose peers were actually stronger?

This is where expansion, population growth, and the increase in international talent come into play. The above question mostly comes down to something we asked earlier: Has the watering down from expansion or the ever-increasing talent pool had a greater effect on overall league quality?

Past research suggests pretty heavily it is the latter. David Gassko, for instance, published a series of articles on this website in 2007 that found the overall quality of the league’s hitters had been steadily increasing for most of baseball history before starting to level off in the last couple decades. Different researchers have come to different conclusions about just how much league quality has improved over time, but the consensus is that the direction of the change is toward the major leagues getting stronger.

Before I go on, it’s worth pointing out that this is actually a fairly complicated question, and there isn’t necessarily one clear answer, or even one clear question. It makes a difference, for example, whether we are talking about how Koufax would fare next to modern pitchers if he were transported directly from the 1960s, or if he were born the same time as Kershaw and grew up with the same modern amenities. It makes a difference if we care how much of the increase in player quality is due to there simply being more talent competing for the same major league jobs, and how much is due to factors like nutrition, conditioning, health care, etc. that help each player push his talent further.

These are as much philosophical questions as anything, so there’s no clear right or wrong answer. It’s also not always immediately obvious which question is being addressed by a given methodology, or how different causes factor into the conclusions, so you have to be careful about interpreting results and be aware there are multiple legitimate ways to approach the problem.

That being the case, I can’t pretend to give anything like a definitive answer about how much major league quality has changed since the 1960s. That doesn’t mean we should simply ignore the issue, though, so first let’s establish what exactly we are trying to measure. I’m going to try to focus on the effect of the increased talent pool and whether expansion has kept up with that increase.

The methodology used in Gassko’s articles, as well as by some of the researchers he cites, involves trying to directly track changes in major league quality from year to year using actual performance data, which is a sensible approach. It is difficult to distinguish how much impact different factors have on the results, though. How much of the increase in quality, for example, is from the growing talent pool versus improved conditioning and training versus medical and nutritional advances?

It also can be difficult to untangle the effects of changes in major league quality from other effects, such as aging, survivorship bias, regression to the mean, etc., some of which are measured using similar methodology. That’s why Gassko got such a different magnitude to his results from Clay Davenport, another researcher Gassko cited in his articles.

Rather than use their methodology, I’m instead going to do something similar to what Gassko did in another article, published in the 2007 Hardball Times Annual, which compared pitchers from different eras.

Gassko’s era adjustments for that article were based on comparing the number of major league jobs to the total male population of countries represented among players. If the population is growing faster than the major leagues are is expanding, the group of players competing for those jobs represent a more elite segment of the population. If the majors are expanding faster than the population, the opposite would be true.

The main problem is that there are two separate issues affecting the growing population feeding into the majors. One is simply population growth. The other is the growing amount of international talent coming into the U.S. from Latin America, Asia, etc. While it’s simple enough to account for population growth, the second issue is more difficult.

For example, how much does Japan, a country with a large population and deep baseball roots, but which keeps the vast majority of its talent in its own domestic leagues, add to the talent pool inthe majors? Do you give the populations of Mexico and the Dominican Republic the same weight even though the Dominican Republic produces far more major leaguers despite having less than 1/10th the population?

Gassko handled international expansion by including a country’s population in the talent pool for the major leagues once it had produced at least three major league pitchers in any given season. It’s not easy to come up with a better way to count the international population the majors draw from, but giving Mexico over 10 times more impact on the size of major league baseball’s talent pool than the Dominican Republic, or counting Japan’s and Cuba’s full populations in spite of the professional, cultural, and diplomatic hurdles keeping most of their talent out of the majors, seems off.

The good news is that we can simply ignore foreign talent and focus only on U.S.-born players. (This works because the average American-born player is pretty close to the average major league player overall.)

The idea is, even though major league baseball has expanded faster than the U.S. population, a lot of those extra jobs are filled by the influx of international talent. If the number of Americans in the majors is growing more slowly than the population, that indicates the extra international talent coming in has raised the overall quality of the league and made it harder to get into the majors If the proportion of the American population in major league baseball is growing, it means expansion has gone further than what the growth in international talent can cover, and the league quality has probably dropped.

So which is it?

In 1935, the year Koufax was born, there were 2.4 million births in the United States, with an infant mortality rate of about 55 deaths per 1,000 births. In 1988, the year Kershaw was born, there were 3.9 million births, with an infant mortality rate of about 10 per 1,000 births. That means Kershaw’s generation is about 72 percent bigger than Koufax’s. In order for the overall talent level to remain the same, there should be about 72 percent more jobs for American-born pitchers in Kershaw’s era than in Koufax’s.

For the reasons covered above, I’m going to focus only on starting pitching jobs here. There were 18 major league teams in 1961 and then 20 from 1962-66, or an average of 19.6 teams in the majors during Koufax’s run. There are 30 today. With today’s five-man rotation, that’s 150 current starting pitching jobs. Koufax pitched in the era of four-man rotations, which would be about 78 starting pitching jobs on average during that span.

However, while we generally think of teams of that era using a four-man rotation, the concept of a starting rotation was a bit different then. Today, teams set a rotation of five pitchers and stick to that order pretty rigidly. In Koufax’s day, an ace like Koufax would regularly take the mound every fourth day, but the schedule was much more erratic for the back end of the rotation. Teams would quite frequently shift between a four- and five-man rotation, move starters forward or back in the order, or insert a spot starter to get the main rotation some extra rest.

Frank Vaccaro’s history covering the origins of the pitching rotation gives us a good way to measure this. Vaccaro tabulated the average number of games each team went before repeating a starting pitcher to computer a “rotation number” for that team. Modern teams pretty much all have a rotation number of almost exactly 5.0, but past teams varied quite a bit more.

As it turns out, while pitching every fourth game was the norm for aces like Koufax, it was actually a bit more common during his era for pitchers to be used on a five-game rotation. On average, teams were using about 4.6 starters per time through the rotation.

Using that number instead gives 91 starting pitching jobs in Koufax’s time (I’m defining a starting pitching job here as one spot in the rotation, by the way, which may be filled by multiple pitchers over the course of a year due to injuries, demotions, promotions, etc). Next, we find the proportion of foreign-born pitchers in the majors. From 1961 to ’66, it’s about eight percent. From 2009 to 2016, it’s about 26 percent.

That means there were about 83 starting pitching jobs for American pitchers in Koufax’s era, compared to 111 now:

That’s an increase of 33 percent, much less than the 72 percent we were looking for.

How do we account for this in the stats? In order for expansion to create enough jobs to keep pace with population growth, the number of starting pitching jobs for U.S.-born players would have to jump to about 144. Therefore, in order to create a league approximately the same overall strength as Koufax’s, we’d need to add an extra 32 starting jobs for American-born pitchers (along with a corresponding increase in the number of foreign-born pitchers, but we can ignore that since the math just cancels out).

We can assume these extra starting roles will be filled by replacement-level pitchers. FanGraphs sets the replacement level for starters at a 0.380 winning percentage, which is equivalent to about a 5.50 RA9 in Kershaw’s league environment. So our newly expanded league would have 111 jobs going to existing starters, who collectively have a 4.36 RA9, and 32 jobs being filled out by replacement-level starters, who we expect to have a 5.50 RA9. Overall, that would push the league RA9 up from 4.36 to 4.61 (or 4.42 once we park-adjust it to Kershaw’s context).

Note that this doesn’t include a times-through-the-order adjustment, so you can knock a couple extra points off Koufax’s RA9- if you want.

The math here is based on the number of starting pitching jobs, which overwhelmingly go to an organization’s best pitchers. It’s worth considering, however, that the five starting pitchers are not necessarily the best five pitchers on a team anymore. While teams from Koufax’s era were mostly starting their best pitchers, teams today are more likely to deploy one of their strongest pitchers out of the bullpen as a closer.

It is reasonable to believe the average closer is a better pitcher than the average fifth starter, and it might even be that the average closer is as talented as the average starter overall.

If we go with the assumption that the average closer is roughly equal to the average starter, that would mean the average starter today is actually equivalent to the top six pitchers per team. In other words, you could make a case that we actually should be counting six pitchers per team instead of five to account for the fact that more starting-caliber talent is being diverted to the bullpen today than in the ’60s.

Under that assumption, we’d only need an extra 10 or 11 U.S.-born pitchers to get the today’s major leagues to match up with Koufax’s era. That would drop the adjusted league average down to 4.44 (4.26 with the park adjustment)—still up from 4.36, but not by nearly as much.

This process also assumes none of the 1960s relievers were as good as the era’s starters, which probably isn’t true (Hoyt Wilhelm, for example). So this is probably too small an era adjustment for Kershaw even if you do think closers are as good on average as starters, but it does give us a decent range of impacts. Depending on how much pitching talent you think modern teams are diverting from the rotation to the bullpen compared to teams from the 1960s, the era adjustment (at least using this methodology) could add anywhere from 0.08 to 0.26 runs to Kershaw’s league average.

This isn’t definitive by any means, and there are any number of additional factors that could further impact the results. If American athletes are more or less inclined to pursue baseball, that can change how we interpret population growth. Medical breakthroughs also can have an impact, since they keep more talent in the game longer. For example, pitchers like Stephen Strasburg who have suffered what would have been career-ending injuries in the 1960s would not have remained among Koufax’s peers as they have for Kershaw.

I also haven’t really talked about defensive support, which is another important consideration. Total Zone rates Koufax’s defenses a little below average, while UZR and Defensive Runs Saved rate Kershaw’s a bit above average, for a difference of roughly 50 or so runs (across all games, not just when Kershaw or Koufax were pitching). That works out to about 0.04-0.05 runs per game, or about one extra point in RA9- in Koufax’s favor.

Definitive or not, though, it’s pretty clear Kershaw’s career is already at least comparable to one of the most revered pitchers in baseball history. Even if Kershaw were to walk away from the game today, his legacy already would be secure. Instead, once he returns to the mound, Dodgers fans finally will get a glimpse at the answer to the question that’s been burning in their hearts for the last 50 years: What if Sandy Koufax had more time?

References & Resources

• Retrosheet
• Lahman Database
• FanGraphs
• Baseball-Reference
• David Gassko, The Hardball Times, “Measuring the Change in League Quality”
• David Gassko, The Hardball Times, “Measuring the Change in League Quality (Part Two)”
• David Gassko, The Hardball Times, “Measuring the Change in League Quality (Part 3)”
• David Gassko, The Hardball Times Baseball Annual 2007, “The Most Valuable Pitchers of All Time”
• Frank Vaccaro, Society of American Baseball Research Fall 2011 Baseball Research Journal, “Origins of the Pitching Rotation”
• Joseph Werner, ESPN SweetSpot, “Comparing greatness: Kershaw vs. Koufax”
• Dave Anderson, The New York Times, “Sandy Koufax and the Sistine Chapel”