Revisiting ’03 will go back to reviewing some of the seasons of individual Marlins players, but I also wanted to go through the season and tell the story of 2003 with the sabermetric lens. I wanted to start at the point which I clearly remember, some 10 games after Jack McKeon took over for fired manager Jeff Torborg. At the point when Torborg was fired, the Marlins were 16-22, and since McKeon’s arrival, the team went 3-7 and reached 19-29, ten games under .500.
Of course, the Marlins since that point went 71-42, won the National League Wild Card and eventually took the World Series. So what did the team look like at 19-29? Let’s find out.
At that point, the Marlins were batting .262/.330/.429, which is right around the league average of .262/.333/.417 (league average taken at the end of the season). The pitching staff and defense were not particularly poor either, having allowed a 4.31 ERA compared to the league average 4.29. Still, the Marlins had up until that point allowed 227 runs and scored only 202, for a Pythagenpat win percentage of .446.
The Marlins had just come off a particularly poor six-game losing streak, and another failed season was looking more and more eminent. The team was also 6-9 in games decided by one run at that point, including a 1-2 record in extra inning games. Given that you would expect teams to be around .500 on those games, the Marlins gave up a win or so in that regard.
Here’s something interesting though. As I perused the box scores over at Baseball-Reference, I noticed a couple of interesting game scores. The Marlins’ fifth game of the season was on April 5th, and it was a 17-1 blowout of the Atlanta Braves. Just one day later, the Braves pounded the Marlins back, 13-4. Those are some pretty drastic run swings for a 1-1 record. It turns out the team participated in 14 games in which the run differential was five or greater. Games like that could strongly affect the Pythagenpat win percentage given, because those may under- or oversell the actual wins or losses given.
Let’s take a look at those 48 games as a whole. Below is a graph of the team’s win% compared to their Pythagenpat expected win%.
You can see that other than in the early stages of the season, the biggest differential in Pythagenpat and actual win% was at the 48th game. The primary culprit of that seemed to be the six-game losing streak, as the two percentages did not drop at a similar slope.
But why does this even matter? It turns out the difference between a .396 and .446 win% is a meager two games, one of which we said could be chalked up to randomness in one-run games. Heck, all of it could be random. The standard deviation of a true talent .500 team in 48 games is 3.5 wins to either side, meaning the Marlins performed at less than one standard deviation away from expected.
Which does not mean that they were as bad as they were. Maybe there was some sequencing that got them during some of the closer games. Context could have played a role at the game level, something I won’t look too far into. But for the most part, the Marlins were probably who we (more or less) thought they were at 19-29. The question is, how did they get to 91-71 by the end of the season? I’ll leave you with this: we know that the team was slightly below average overall on the season, and the Marlins hit only .267/.334/.417 the rest of the way, slightly above average. What’s left to consider? Well, I believe having Dontrelle Willis and Josh Beckett making starts in place of guys like Michael Tejera and Tommy Phelps probably had a say in the end.