On October 21, 1997, the Florida Marlins beat the Cleveland Indians 14-11 in Game 3 of the World Series. The game was tied 7-7 entering the 9th inning. Florida scored 7 in the top of the 9th - fueled by three Cleveland errors, three walks, and a wild pitch - to take a 14-7 lead. Cleveland scored 4 in the bottom of the 9th off closer Robb Nen. Omar Vizquel grounded out to end it with the tying run one batter away.
That game happened. That is history.
In 69 simulations of that game, Florida wins 18 times. Cleveland wins 51.
The multiverse is telling us something: the real outcome was the exception. The real Game 3 - the one that actually happened on October 21, 1997, in front of 44,880 people at Jacobs Field - was the 26% scenario. It was the coin landing on its edge. Across the infinite expanse of possibility, Cleveland wins this game almost three-quarters of the time.
And it isn't close.
Florida can lead. Florida can score first. Florida can build a 5-run cushion. It does not matter. Cleveland comes back. In 30 of 51 wins, Cleveland trails at some point and erases the deficit. They overcome 5-run leads five separate times. They overcome 4-run leads nine times.
Florida does not lose because it cannot score. Florida averages 6.1 runs per game. In the real Game 3, Florida scored 14. In the simulations, Florida scores 10 or more runs in nine different timelines. They still go 5-4 in those games.
Florida loses because Cleveland is unkillable.
In 30 of Florida's 51 losses, Florida held a lead at some point. Here are the worst:
Timeline 1: Florida leads 9-5 after 7 innings. Cleveland scores 7 in the bottom of the 8th. Florida scores 4 in the top of the 9th to retake the lead 14-13. Cleveland scores 2 in the bottom of the 9th to walk off 15-14. This game is a near-perfect echo of the real Game 3 - except in this version, Cleveland finishes the rally that real-Cleveland couldn't.
Timeline 33: Florida leads 5-0 after 2 innings. Cleveland scores 7 in the 4th. Wins 12-9.
Timeline 63: Florida leads 5-0 after 3 innings. Cleveland scores 5 in the 5th. Wins 10-5.
Timeline 67: Florida leads 22-19 entering the 9th. Cleveland scores 4 to walk off 23-22.
Timeline 68: Florida leads 25-20 in the 9th inning. Cleveland scores 5 to walk off 25-24.
Five times, Florida builds a 5-run lead. Five times, it evaporates. Two of those five are games where the total combined score exceeds 45 runs. The lead should feel safe. It is never safe.
In the real Game 3, Florida's 7-run 9th inning was the decisive blow. In the multiverse, Florida scores in the 9th inning of 18 games, totaling 45 runs. They score 4 or more in the 9th five times. They have their big 9th-inning rallies. It doesn't save them.
Cleveland leading entering the 9th inning: 45 games. Record: 44-1. They lose that lead exactly once - Timeline 32, a 10-inning game where Florida scores 1 in the top of the 9th to tie and wins in extras.
Florida leading entering the 9th inning: 18 games. Record: 16-2. They blow it in Timeline 53 (CLE walks off 3-2) and Timeline 67 (the 23-22 game).
And when the game is tied entering the 9th? Six games. Cleveland wins five. Florida wins one.
The lesson: Florida can build a lead. Florida can maintain a lead. But if the lead cracks, if it slips for even one inning, Cleveland pounces. And leads crack all the time.
Al Leiter starts all 69 games for Florida. He pitches 13 quality starts - 6 or more innings with 3 or fewer earned runs. These are the games where the ace does his job. These are the games where the pitcher gives his team a chance.
Florida's record in Leiter's 13 quality starts: 6-7.
He throws 8 shutout innings in Timeline 10, striking out 8. Florida wins 1-0. He throws 7 innings of 1-run ball in Timeline 39. Florida wins 14-1. He throws 8 innings of 1-run ball in Timeline 53, striking out 3. Florida loses 3-2 on a Cleveland walk-off.
He can pitch a masterpiece and it might not matter. In 7 of his 13 best starts, Florida still loses. In his other 56 starts - the ones where he's average or bad - Florida goes 12-44.
Leiter's overall record across the multiverse: 6-38. Six wins. Thirty-eight losses. He is not a bad pitcher. His quality start rate is 13/69 (19%), which is low but not shocking for a guy facing this Cleveland lineup. The problem isn't Leiter. The problem is that Florida's margin for error is zero, and even Leiter's best work only gives them a coin flip.
The real Game 3 shares DNA with Timeline 1. Both feature Florida building a lead, Cleveland storming back in the 8th, Florida scoring big in the 9th to retake the lead, and Cleveland rallying in the bottom of the 9th. In the real game, that final rally fell 3 runs short. In Timeline 1, Matt Williams hits a walk-off double to complete it.
The same players appear in both scripts. In the real game's bottom of the 9th: Bip Roberts hits a two-run double. Omar Vizquel makes the last out. In the multiverse, Bip Roberts bats .370 in close-and-late situations. Vizquel hits .553. The hitters who nearly saved the real game are the ones who actually save it, over and over, across 69 timelines.
The real game's 9th inning - Florida's 7-run explosion, Cleveland's desperate 4-run answer - looks like exactly the kind of thing the simulation produces. It looks like something that could go either way. The multiverse confirms that it could. It just usually goes the other way.
In 69 versions of October 21, 1997, Florida's victory is the exception. Not the rule. Not even particularly close to the rule. Cleveland wins 74% of the time. They come back from any deficit. They win 15 of 19 one-run games. They are, across every possible version of this night, the better team.
Florida won the game that counted. The multiverse has spent 69 timelines disagreeing.