The data set used for this study is every team’s wins, losses, win percentage, runs scored, runs allowed, run differential, expected winning percentage, expected wins and luck. All of these specific stats have been tracked for every team in baseball from 2000-2018, giving us 19 years of data for 30 teams each year. This data analyzes 570 different teams from this century.
Not only are the listed stats analyzed from the beginning of the year, but they are also analyzed for the end of the season, as well as for the season after the month of April. It is, essentially, an insane amount of data.
As for an explanation of some of the more complex stats listed, run differential is quite simply the difference between runs scored and runs allowed. Expected winning percentage (or Pythagorean win percentage) is calculated on Baseball-Reference.com and uses a team’s runs scored and runs allowed to determine how many games a team should be winning. Expected wins is simply expected winning percentage times the number of games played.
“Luck” is a little bit of a strange stat, accounting for the difference between a team’s actual wins and a team’s expected wins (as calculated by run differential). One of the strongest examples of luck is the 2018 Seattle Mariners. At the end of June 2018, the Mariners had a 53-31 record, giving them the fourth-best record in the American League. Their 63.1 percent winning percentage would break their playoff drought very easily.
For anyone that paid attention to run differential in 2018, the Mariners at that point of the season had only scored 21 more runs than they had allowed. Based on run differential, they would be expected to win 44 games at that point. But their 53-31 record showed they were nine games better than their run differential would expect them to be. This is typically not sustainable. By the end of the season, their winning percentage slid to 54.9 percent and they did not make their way into the playoffs.
To make the data a little bit easier to digest, hyperlinks that show data and graphs are linked within the article here.