Every year, MLB’s All-Star game brings together the best players from each league to form two superteams. For one game, we get to see Jose Fernandez as a reliever against lineups in which Mike Trout and Miguel Cabrera hit back to back. But unlike the NBA’s fantasy rosters made real, we never get to glimpse how dominant such a talent-laden squad would be against normal MLB competition. So with a little statistical analysis and some conjecture, I took a guess at how well an MLB All-Star team would fare in a regular season — and even how often they’d go a perfect 162-0.1
To get an idea of how good each All-Star team would be, I added up the wins above replacement2 for every All-Star team’s best player at each position since 1933 (the first year of the All-Star Game). To further make things comparable to regular-season teams, I summed the top five pitchers’ WAR totals to get a rotation’s worth of pitching WAR.3 I also did not consider any WAR contributed by the designated hitter for each All-Star and regular season team, since the DH did not exist before 1973. The result of all this was a predicted WAR total for each All-Star team, which I could use as a comparison against real regular-season teams.
Not surprisingly, All-Star teams tend to carry far more talent in their ranks than most normal teams. The average All-Star squad put 60 full-season WAR on the field, which is about the same as the 1976 Reds — widely regarded as one of the best teams in MLB history. No regular-season team in history exceeded the 1927 Yankees’ 66.3 WAR, but about 30 percent of All-Star teams would have if given the chance to play together in the regular season.
But 66.3 WAR is kind of an abstract idea; what most fans care about is Ws and Ls. To establish how well these All-Star rosters might have fared in the standings, I used regular-season teams as a guide. I regressed their winning percentages against the total WAR on their rosters to get a sense for how much each additional WAR was worth.4 As expected, each win above replacement contributed to a team increases winning percentage by roughly 0.7 points, or the equivalent of about one win in a 162-game schedule.
By this method, practically every All-Star team would be predicted to have a winning record, and the average All-Star squad would be predicted to win 73.4 percent of its games.5 In a 162-game schedule a .734 winning percentage would lead to 120 wins, a feat no real-life team has ever achieved. And that’s just the average; the very best All-Star teams — the top 10 percent — would be predicted to win more than 81 percent of their games, or 132 contests in a regular season.
Standing atop that group as the best All-Star team ever was the 1997 National League squad. (Which, by the way, lost to the AL 3-1, a reminder that in any one game a superteam can lose to a merely great team, especially if there isn’t much at stake.) Seven players from that roster have already made the Hall of Fame, with two more (third baseman Chipper Jones and pitcher Curt Schilling) likely to reach induction in the near future, and a few others (such as outfielder Barry Bonds and first baseman Jeff Bagwell) mainly excluded over performance-enhancing drug concerns. (By comparison, only five players from the AL team that opposed them have made the hall so far.) Combined, the top players on the ’97 NL team produced 86.2 WAR; six of them reached the MVP level threshold of 6 WAR; their worst position player, Jones, ended up producing 3.7 WAR — still 23rd best in the NL.
We can’t say for sure how such a team might have fared over a 162-game schedule; the assumptions of any model can break down at the extremes, particularly since we’re trying to extrapolate from a sample of regular-season teams that have never been anywhere near as powerful. But by the model outlined above, the 1997 NL All-Stars would have been predicted to win 87 percent of their games, or 140 times in a season. Even given the amount of luck in baseball records,6 the ’97 NL would hypothetically go undefeated only once every six billion seasons. (So the best team in baseball history — by a huge margin — would still be the longest of long shots for a perfect record.)
I can, of course, take things a step further and assemble the all-time greatest All-Star team. By assembling the greatest single-season performances at each position throughout history, I can build a team with almost 137 WAR, more than 50 percent better than the greatest single All-Star team ever. This team — with Lou Gehrig from 1934 manning first base, Barry Bonds from 2002 in the outfield, and ’99 Pedro Martinez sharing a rotation with ’72 Steve Carlton — would be predicted to win 96.8 percent of its games, becoming the first 157-game winner. With a lot of luck, it could eke out an undefeated regular season, but even for them it would be far from a certainty. (According to the binomial distribution, it would happen once every 200 or so seasons.)
Obviously, no such team will ever play the regular season, and this simplified approach ignores many factors that limit teams from such otherworldly performance, such as injuries and the grind of the long schedule. Even so, it’s intriguing to consider how overwhelming an All-Star team likely would be in the face of regular-season competition. For a game in which the default is to fail seven times out of 10, most All-Star teams would suddenly make baseball look quite easy.
