Simpson's Paradox, Bert vs. Jack, and more sweet B-Ref splits

Bert Blyleven had a better winning percentage than Jack Morris when their teams scored 0-2 runs. Bert had a better winning percentage than Jack when their teams scored 3-5 runs. And Bert had a better winning percentage than Jack when their teams scored 6+ runs.

Yet Jack has the better overall winning percentage.

Sounds funny, right? The explanation is pretty easy to accept, however: Jack's teams scored more runs more often, so when you do a weighted average of all the groups, he comes out looking better. But he wasn't.

almost 2 years ago Sky Kalkman 2 comments 0 recs |

Story-email Email Printer Print

Comments

Little off -- Bert very nearly had a better winning percentage when they scored 3-5.

Morris had a .598 percentage for 3-5 vs Blylevens .595 percentage — could call it equivalent.

Otherwise, pretty cool to see how Morris’ 100 fewer 0 – 2 games created such a different overall winning percentage for him. If he had pitched 100 more games and all had been 0-2, he would get, lets say, 10 wins and 40 losses and completely change the outlook on his career winning percentage.

by Trickman on Mar 16, 2010 1:54 PM EDT reply actions

DON'T RUIN THE STORY WITH FACTS!!!!!!!!!!1

When I taught some AP Stats, the text book’s example of Simpson’s paradox was that Airline A had a higher percentage of on-time flights than Airline B at every airport. But Airline B had the higher percentage of on-time flights overall. The students had to figure out how the heck that could be. Then we had a nice long discussion about whether any claims by Airline B that they were better for on-time flights were legal/ethical. Great discussion. Students sure do love the first month of AP Stats…

Beyond the Boxscore Not a member? Sign up.

by Sky Kalkman on Mar 16, 2010 1:58 PM EDT up reply actions