Did a comprehensive dive into how well run differential correlates with a team's record both on the season and from the first 81 games to the next 81 games (or in this case, to whatever's been played). Turns out Run Differential is just king still lol and @YankeesFiles comes out with a DOMINANT victory Though taking: OPS (hitting) - OPS (pitching) Is also pretty effective. Though what's about as equally effective, if not slightly (insignificantly) more is Run Differential in games decided by 10 or fewer runs — eliminating those blow outs. I looked at a VAST range of metrics/run differentials to try and exclude blow outs. I took: - run differential against starting pitchers only - run differential against relief pitchers only - run differential in games decided by 2, 3, 4, 5, ...., 10 or fewer runs (excluding true mop-up duty blow outs) - run differential in games decided by 2 to 4, 2 to 5, 2 to 6, ...., 2 to 10 runs (excluding blowouts and 1-run games) - median game hitting OPS minus median game pitching OPS - xwOBA hitting minus xwOBA pitching - win percentages in games decided by 1, 2, 3, ...., 10 or fewer runs . . and some other metrics I looked at how these are predicatively and how these are descriptively within the same seasons. Win percentages in games decided by 5, 6, 7, etc..., 10 or fewer runs correlate as high mostly because there's a VAST proportion of every team's games played that fall within a 5 or less run differential. So in practice, these win percentages in games decided by *insert a lot of runs* or fewer are almost 1 -to -1 images of teams' actual records. What's interesting though (and what was already basically known): one-run game W%s don't tell a good story of how good a team is nor do they do a good job of predicting how a team will be in that same year. In fact, this gets truer (as you'll see) as you look at win percentages in closer and closer games. Win%'s in 4-run or fewer, 3-run or fewer, 2-run or fewer, and 1-run games all have lower correlations to a team's record than actual team's records. However 5-runs or fewer W%'s seem to do a good job of correlating to a team's record, in fact better than the team's record itself.