5 Comments
User's avatar
Michael's avatar

Great post again. What if you left lambda as a free parameter in your regression? Just think it might be interesting to check (and for the football).

Note: this is the question I had last time but I think you misinterpreted it (and instead answered a probably more interesting question)

Expand full comment
Tom Singer's avatar

> Let’s take a closer look at home field advantage. In the 2024-25 NCAA men’s basketball season, the average scoring margin favored the home team by an average of about 9 points.

That's going to be affected by good teams having more home games against bad teams, right? Top schools tend to pay cupcakes to come play them as one-offs, rather than scheduling home and homes. Take Auburn, for example:

Auburn played 15 home games to 10 away games; their non-conference home games were exclusively against teams that they out-classed: a 32 point win over FAU, a 51 point win over Vermont, a 23 point win over Kent St, a 33 point win over North Alabama, a 44 point win over Richmond, a 41 point win over Georgia State, and a 29 point win over Monmouth. They played one non-conference away game at Duke, and the rest of their non-conference schedule was neutral site. Is that inflating the apparent home court advantage?

Expand full comment
Zach Wissner-Gross's avatar

You're absolutely right, Tom. And I admit that at the end of the paragraph: "I acknowledge that both of these margins would probably decrease if I looked exclusively at intra-conference games."

I didn't see a quick way to compute the margin for intra-conference games, so I used all games as a proxy.

I do think the impact of home-field advantage for non-conference games is smaller than you might think, since the major conferences tend to host more games, which would have the effect of deflating the indices of the non-major teams.

Expand full comment
Tom Singer's avatar

Maybe. But nearly half of Auburn's home games were non-conference cupcakes. Same for Duke. It's a big chunk of the data. I don't necessarily think the conclusions are invalid where it counts, though - the teams at the top of your rankings are mostly going to play similar schedules in that regard.

Expand full comment
Zach Wissner-Gross's avatar

At some point before 2026 March Madness I'll determine values for parameters that historically best predict the tournament.

Expand full comment