But it also shows the Cardinals as having performed far above expectation.

One big problem with this is that every team does not play every other team an equal amount. So even if simply summing a team’s WAR was indicative of a team’s talent level for a given season, some teams would be favored in this because they had an easier schedule.

Case in point, there’s a big red flag that I noticed when looking at the numbers: every team in the NL central overachieved last season? That seems unlikely. And the same thing in 2008 (only 1 team below 0). Meanwhile, on the other side of that coin, the AL East, for 3 years in a row the majority of the teams underachieved according to this measure of true talent. At first glance it looks like division strength correlates negatively with this measure of Wins vs Expected, which is what I would have expected.

Beyond that it’s dumb to attribute all of this difference strictly to the manager. There are a lot of reasons why a team’s record may not match its level of talent, and a manager is just one of them.

I like the set-up, it looks like a simple regression equation from 4 years worth of correlational data. Looks like it was a good rainy Sunday activity.

I guess the first two questions would be to investigate the covariance between WAR and win totals; I’m not so certain 4 years worth is enough to create this regression equation. I would have done at least 5 or 10, I’m not sure why 4 was chosen.

Can we use individual-level performances to predict group-level wins? I myself don’t know if this is an appropriate analysis; perhaps something like a HLM approach would be easier, with players nested within teams.

Conversely, and simply enough, can a manager be to blame for a loss? He doesn’t hit, field, or pitch. Perhaps it means for as good some players are, they are poor on other days.

Fun analysis, I liked it.

Dan