NFL Betting Angles Fact or Fiction #6: Black Coaches

The "fade black coaches" angle is well-known across the internet. If you don't believe me, take a look here:

https://www.sportsbookreview.com/forum/college-football-betting/3139651-cut-shit-black-coaches-do-not-win-higher-rate-than-non-black.html

https://www.sportsbookreview.com/forum/saloon/3493456-its-not-racism-its-realism-nfl-ncaaf-betting-strategy.html

There was also a long thread on rxforum in the recent past; perhaps it was taken down as I can no longer find it there. In any case, since this angle basically went 2-0 in the same game yesterday - the Browns having a black coach earlier in the season and last season only to switch to a white coach with improved performance thereafter, facing a team with a black coach who has under-performed all year - I figured it was worth investigating. I expected nothing to come of this angle, as it just seems racist and stupid. However, I was wrong, and this is one of the strongest betting angles I have ever seen.

To start, I noted the race of each coach in each game back to 2009, including mid-season coaching changes. We can then perform the simplest test possible - black coach ATS record since 2009:

We can see that fading black coaches - betting the white coach whenever a white coach was facing a black coach - would have shown a very slight profit of 4.2 units over 594 games. Enough to beat the vig, but well within normal statistical fluctuation. Of note is that black coaches tend to be coaching teams that are worse on average, as the home team in home black coach - away white coach games is favored by 1.6 points less on average compared to the opposite type of game. Such a difference is much larger than could be caused by chance alone. Whether this is due to bad coaching, bad rosters, or a combination of the two, we can't tell at this stage.

Not a particularly interesting angle so far. But let's take it one step further. Theoretically, if black coaches are worse than white coaches, we'd expect the coaching influence to be reflected in the team's stats. While most people tend to think of bad fourth down or challenge decisions when they think of bad coaching, almost all NFL coaches are absolutely terrible at fourth down. Instead most of the difference between coaches comes from better play-calling, better staff, and better roster management, all of which should flow through to the stats. Over time, the betting line, which is primarily a measure of past team performance, would then reflect those stats. This would suggest that we would not expect this angle to work on teams who have had the same coach for multiple seasons in a row regardless of race, as the betting line which focuses primarily on stats and injuries would price in the level of coaching through past stats.

Where this angle might have value is when there is a coaching change, from white to black or black to white. If the theory is true, teams that switch from a black coach to a white coach would be expected to improve their performance, since while they were being held back before, they can now benefit from better coaching. Past stats, and probably the market as well, would under-rate such teams, particularly early in the season when the stats have not yet had the chance to "catch up" to the change in the level of coaching. The opposite would also be true - teams that switch from white to black coaches would potentially be worth fading.

To test for this concept, one can use a statistic known as "coach race differential". This statistic measures the percentage of games from the team's last season that were coached by a black coach, then subtracts this percentage from the current race of the coach, where black coaches are given a value of "1" and other coaches a value of "0". For example, if the current coach is white but the team was coached by a black coach in all 16 games last year, the statistic would have a value of (0 - 1) = -1. If the team was coached by a black coach in half the games last year, and this game is coached by a black coach as well, the statistic would have a value of (1 - 0.5) = 0.5. Note that we are accounting for in-season firings, so if a team had coaches of two different races in the past season, they could have had  "half a white coach" in the last season.

We can then calculate the difference in this statistic between teams, to test the power of this betting angle. In the eyes of the angle, the most favorable setup for the home team would be -2, where the away team switched from a white coach in the past year to a black coach, while the home team switched from a black coach to a white coach, and the least favorable would be +2. Most games will fall 0 as coaching changes are semi-uncommon and only about 1/5 of coaches are black on average, but there were over 250 games where there was a difference of at least 0.5 between home and away:

 Black Coaches YoY Change in Race ATS Record

We can see the makings of a very powerful betting angle here. Over 255 games, betting against teams that went from white to black coaches, or on teams that went from black to white coaches, compared to the opponent, profited a cool 26.4 units (actually 27.4 since CLE-DEN is not included above). Both the home and away sides of this angle were highly profitable. This was backed by a very strong average cover margin of 2.2 points per game over these games suggesting that this was not just due to teams having a lucky run in close games.

Black Coaches Regression

We can see this further from a regression test, where we have regressed home margin of victory in the game versus the game's point spread and the race differential stat. The coefficient of -1.92 implies that, for example, assuming no away coaching change, a home team with a black coach in the current game who had a white coach in 100% of the previous season would be expected to perform 1.92 points worse in the current game, after the market point spread was taken into account (if we run the same model as above without the market point spread, the value increases to 2.35).

One potential issue with this angle is that it turns out that black coaches have been far more likely to be hired as interim coaches after mid-season firings, which as I have mentioned here before, have done very poorly in the past, although that theory has not done well this season. To rule this out, we can look at early-season performance, where no coaches have yet been fired. We find that the angle persists and in fact that average cover margin was actually better in early-season games, although the win-rate was slightly lower:

In the end, I have to go with a Fact here as this seems to be an extremely powerful angle. The black coaches really have done worse even after accounting for the market, which is the sort of thing that is hard to find in this day and age. Just going through the hirings and firings, I suspect that black coaches have been hired into worse situations on average than similar white coaches, but that may just be confirmation bias.

Either way, it should be noted that the results of this angle were based on a small subset of perhaps 10 to 15 coaches that sucked. The effective sample size of the statistical tests is lower as a result. In addition, successful coaches like Tomlin and Lewis are not part of this angle since they had their job over the entire sample period. Rather than betting money based on the change in race of the coach, in real-life NFL betting, one is probably better off evaluating a new coach on their merits, not just race, to determine if they are likely to be better or worse compared to the previous coach. The issue is that doing so probably requires an understanding of football and human psychology that I don't have, and not many other people have, as well as knowledge of what goes on in practice and the quality of the team's staff.

Another possibility would be to set up a sort of first-year coaching prior model based on the characteristics of the new coach coming in, such as age, past experience, education, race, and other factors. My guess is that there would be enough of a sample to make some interesting conclusions. If I had to gamble on which coaches were best, young successful coordinators would be the best choice, and old coaches with mediocre records would be the worst choice.