Saturday, December 12, 2020

Early look at Home Court Advantage in Covid Season

With some frankly disturbing exceptions, most college basketball games this year are being played in empty or near-empty arenas. This raises an obvious question: how much does this affect home court advantage (HCA)?

It's a reasonable assumption that lack of fans will reduce home court advantage, and I lowered the HCA factor in my ratings from 1.3% to 1.0% coming into the year. But I am tracking it (or at least attempting to)to see whether it needs to be further adjusted. Here's what I'm looking at so far.

I calculate and apply HCA by looking at how many more net points home teams score than my ratings project they should. For example, if two perfectly evenly matched teams play, and the home team ends up winning by 3 points in a 50 possession game, that's an HCA of 6 points per 100 possessions. Here are the actual results of that method over the last three years, plus this year so far:

Year Games HCA/100
2018 4873 5.14
2019 4934 4.93
2020 4877 5.22
2021 438 5.53

Looking at this, you might conclude that HCA is as strong as ever so far this year. But that would be wrong. Things look quite a bit different if you look at just the first 17 days of each season:

Year Games HCA/100
2018 662 6.24
2019 555 6.33
2020 666 6.42
2021 438 5.53

On this view, HCA is about 13% reduced from the average of the last three years. Basically, HCA is usually higher at the beginning of the year, probably because it is generally higher in mismatches, and mismatches are common early in the year.

In fact, if we break this down and look at only games in the first 17 days involving teams who are ranked within 100 spots of each other, there is a huuuuge drop-off this year:

Year Games HCA/100
2018 293 5.59
2019 236 5.69
2020 300 5.89
2021 215 2.18

Here, in a fairly similar number of games as previous years, HCA is down by more than half! Given that the overall number is only down about 12%, that means that HCA in uneven matchups (where the difference in team ranks is more than 100) is actually UP quite a bit:

Year Games HCA/100
2018 369 6.75
2019 319 6.8
2020 366 6.86
2021 224 8.78

What is the difference in these games? Well, one possibility is officiating. In the evenly matched games, home free throw rate is just a 1.18% advantage for the home team (versus an average advantage of 3.4% in the previous years. Meanwhile, in mismatches the home team is mostly retaining its usual advantage in getting to the line: 3.2% versus 4.3% in those kinds of games in previous years.

Where does that leave us? It's too early to make any firm conclusions. Another couple caveats I want to add: (1) this analysis is kind of down of dirty because I'm not fully backing out the effect of HCA in the first place when considering team quality (which could especially be skewing the results this year of teams that have not played road games), and (2) the results for this year are obviously hard to disentangle from errors in the preseason projections.

But a couple things I'm confident in saying: HCA still exists, even with few fans in the arena, but it is attenuated to at least some degree. Maybe a lot in real games.

Now for some speculation. Perhaps the difference that shows up between more evenly matched games and mismatches reflects that Covid-related travel and other restrictions are particularly rough on low major teams. And perhaps the vast majority of HCA that remains is more related to this kind of travel fatigue, and has largely replaced implicit officiating bias as the main driver of HCA for now.

Update:

I did some more work on this to back out HCA for team quality. The raw numbers are different (slightly less HCA overall), and make this year even closer to a normal year so far, but the basic ratios / conclusions noted above still hold. In particular, the marked split between closely matched and mismatch games still holds, although the "through day 17" split is now close enough to be reasonably considered possibly just noise.