I did a study this year comparing the two and came up with a similar, if predictable finding.

What’s the correlation for a quadratic regression?

That way we can limit the luck/unlucky bias.

1. No team with a combined 70 WAR or below has had more than 84 wins in the past three decades.

2. No team with a combined 90 WAR or above has had fewer than 86 wins in the past three decades.

That’s fairly significant. Cross reference that with a study on the average number of wins it takes to make the playoffs, and BLAMMO—an all purpose guide for telling teams how much WAR they’ll need to make the playoffs the next season.*

*provided we have accurate WAR projection models in place.

What I think would be interesting is to look at how predicative WAR is vs actual wins. So run a trendline of wins from one year vs wins the next, and WAR one year vs. wins the next. I know there will be problems with players leaving or coming to the organization, and players performing better or worse, but those should affect both the same and it would be cool to see if there’s a major difference between the R squared for each.

Also, Stevesommer mentioned looking for trends in teams that under or overperformed. I have always thought that teams that bunt/situational hit might be undervalued a bit because moving a runner over counts as a normal out, when it does have more value than that (or less negative value). It would be interesting to see if this made any difference. I would guess that a lot of this would be captured in a baserunning statistic, as the runner would get credit for moving up the base, but I’m not sure exactly how those are calculated, and if the WAR you are using takes those into account (I think fangraphs just uses sb & cs, not an overall baserunning score).

These are things I’d look at if I had more time (and was better and this type of thing), and probably will when things slow down here. If someone does this, or has done it, let me know!