On the occasion of this past week’s Major League All-Star Game, when baseball’s absurd abundance of young talent was on full display (Hello, Mike Trout!), the question was worth asking: Where did these two All-Star teams stand historically, in terms of the most gifted groups assembled since the first All-Star Game was held in 1933?
The answer might (or might not) surprise you.
To get a handle on the talent available to each league in each All-Star Game, I used the run-based components1 of Baseball-Reference’s wins above replacement (WAR) to build a projection system that, in terms of complexity, falls somewhere between Marcel the Monkey (as simple as can be) and PECOTA (highly intricate). Most crucially, the projections use multiple years of data, adjust for aging effects and regress a player’s expected performance to the mean; the projections can also be used to estimate a player’s true talent in a given season by taking his projection for the following season and subtracting the age progression.
This formed the basis for yearly talent ratings for every player to appear in an All-Star Game, which I then compiled into team ratings in proportion to how much playing time each player logged during the game itself.2 Finally, using these run-based talent ratings and each season’s average runs per contest — plus a little help from the pythagorean expectation (which converts run differentials into wins) — I generated expected winning percentages for each All-Star roster, which represent the rate of wins that the team would post if it were to face a typical slate of opponents from the season in question.
Unsurprisingly, All-Star squads would demolish the competition if turned loose on regular teams in the context of league play. The average All-Star team since 1933 possessed the talent to go about 103-59 if they played a normal 162-game schedule, with the top teams expected to win a 2001 Seattle Mariners-esque 117 games per season. (And that’s their mean expectation — remember that when the Mariners won 116, they were probably more like a true 90-win team that also benefitted from a tremendous amount of luck.)
But what if we instead had all the historical All-Star teams play each other? Using the aforementioned pythagorean talent ratings and Bill James’ log5 formula to construct probabilities for each matchup, we can simulate how each All-Star squad would fare if forced to play 162 games against a random assortment of other All-Star teams from the past. Here are the standings (check out the raw data on GitHub):
The best All-Star team ever? According to this method, that mantle belongs to the roster that represented the National League in 1966 — and it’s not hard to see why. Sandy Koufax started the game on the mound, to be relieved by Juan Marichal, Jim Bunning and Gaylord Perry — Hall of Famers, all. The starting outfield was outrageous: Willie Mays, Hank Aaron and Roberto Clemente. Ron Santo and Willie McCovey manned the infield corners. Along with second baseman Jim Lefebvre, near-Hall of Fame catcher Joe Torre3 was the weak link of the starting lineup. (And fearsome mashers Dick Allen, Jim Ray Hart and Willie Stargell came off the bench!) Our simulations say the 1966 NL All-Stars would go 97-65 against a random collection of their fellow All-Star squads from throughout history.
This year’s teams, while not on the same level as their Hall of Famer-laden predecessor from 1966, would both be above .500 if forced to take on other All-Star squads — and could do even better than that if we tweak our study’s methodology.
Despite losing on Tuesday, the National Leaguers are considered the more talented of 2015’s All-Star teams according to WAR. Led by Paul Goldschmidt, Andrew McCutchen, Bryce Harper and starting pitcher Zack Greinke, this year’s NL team ranks 15th all-time, and project to go 87-75 in 162 games against other All-Star squads. Meanwhile, the AL team lags a bit behind the NL, ranking 71st with a 82-80 record. Although Trout possesses the best individual talent rating (by far) of any 2015 player, the rest of the AL roster isn’t especially impressive by historical standards.
Then again, both of 2015’s All-Star teams fare much better if we attempt to adjust for the rising tide of athletic talent (in baseball as well as other sports) over time. In “Baseball Between the Numbers”, FiveThirtyEight editor-in-chief Nate Silver attempted to quantify this phenomenon as it pertains to baseball and arrived at a “timeline adjustment” that gently slopes upward from the sport’s major league beginnings in 1871 to the present. Nate’s timeline adjustment implies that the average MLB player from 1966 was about 86 percent as good as the average player today, which means that year’s NL All-Stars — talented as they were — benefitted from facing weak competition.
In fact, if we apply the timeline adjustment to every All-Star team’s talent rating,4 the 2015 National Leaguers emerge as No. 1 all-time, with the AL ranking 14th. The aforementioned 1966 NL powerhouse drops to third, with the 1986 AL squad sliding into second place. (That team was powered by great pitching — specifically Roger Clemens and Teddy Higuera at his peak — and a lineup that included Hall of Famers Cal Ripken Jr., Rickey Henderson, Wade Boggs, Kirby Puckett and Dave Winfield.)
Do I actually think the 2015 National League represents the most talented All-Star roster ever? Probably not. They rank highly — but are nowhere near No. 1 — relative to their era, and the question of how to deal with cross-era comparisons while accounting for absolute changes to talent quality over time is far from settled. But we can still appreciate Tuesday’s exhibition as one of the most talent-laden All-Star contests ever: Even without the timeline adjustment, the combined talent ratings of the two teams ranks 10th among the 86 All-Star Games ever staged.