APBRmetrics

I ran a (multi year) ridge regression with the referees being listed as players for the hometeam.
Lambda for the real players was fixed (2000); the lambda for referees was determined through crossvalidation.
Optimal lambda appears to be _infinity_, indicating that all referees have the same influence on home court advantage; you get best prediction results by simply ignoring who's refereeing.

I think it also indicates that HCA does not come from the referees being influenced by the fans. If they were influenced by the fans, some of the ~70 referees would be influenced more or less than others. The data say they either all get influenced the same(highly unlikely), or not at all

So, no real shocking results, just thought someone might be interested in the results

This is interesting. I'd have guessed that influence by the home arena would be one of those things most variable among refs.
Could it be that the league is aware of ref tendencies along these lines? And that their refereeing crews are designed to nullify one ref's high home-team bias with another who has the opposite bias?

It may be that some combinations of refs -- 2 or more with large bias -- have been outrageously biased as a group. And subsequently these refs are kept from officiating in the same crew.

Conversely, refs with opposite tendencies will eventually start to behave more like the consensus within their crews. Otherwise, they'll continually be overruling one another on the court.

Mike G wrote:Could it be that the league is aware of ref tendencies along these lines? And that their refereeing crews are designed to nullify one ref's high home-team bias with another who has the opposite bias?

That would be another possible explanation. The league would have to be pretty good at building balanced ref-triplets, though; otherwise it would still somehow show up.

What I haven't looked into yet, but would be possible to do, is estimate ref influence on FTA or time/possession. I suppose there could be some referees that call more than an average number of fouls, which should lead to more FTA and subsequently reduce the amount of seconds a possession takes. That way, they might not have an influence on offensive efficiency, but still might influence point differential (the more possessions, the higher the point differential in favor of the better team)