Every year there are a few teams that manage an impressive injury list by late-season. This year, two examples would be the Redskins and Broncos, which followers of the blog picks would know we are usually on, although these teams managed to play badly enough, and lose players at key enough positions, that we don't have them this week (a shame in the case of the Redskins). In general, we value defensive injuries highly, but if a team is hurt on offense ex-QB, we will be on them. And if you look at the popular Massey-Peabody picks, they are fading injuries every single week.

The fine book Sharper introduced the idea of using Madden ratings to calculate injuries. Oddly enough, I had never used the ratings for injuries before reading the book, but more as a way to measure early-season talent levels before stats really start to work in the NFL. It turns out they are effective in injury calculation as well.

A metric I have developed to measure how injured a team is involves taking the best 10 players by Madden rating on that team on offense (excluding the FB), the best 11 on defense, and the best QB who played in the game, and using these as overall unit Madden ratings in each game. Then, one can take the maximum Madden rating that any such unit has had up to that point in time of the season, and compare the rating in the current game to that maximum. Those teams with a rating in the current game far below their season maximum are likely suffering from a large number of injuries or suspensions/trades, which are for the most part the same thing.

Ideally, we'd then include these injury impacts as part of a fundamental handicapping model that takes into account team statistics and other situational factors. For one example of how to do this, buy my book. But for now, we can look at whether injury impacts have been correctly rated by the market, through an ATS analysis of games between 2012 and 2017.

Each of the tables below are shown in terms of each team's Madden rating for that particular unit in that game, subtracted from the maximum Madden rating that that team's unit has had up to that point in the season (we exclude Week 1). We then take the difference in the two numbers, to determine how relatively injured each team is. For example, suppose in a game the home team is starting an offense with an average Madden rating of 78, whose highest Madden rating thus far was 81, and they are facing a team whose average Madden rating in that game is 82, whose highest Madden rating thus far was 83. The home team has a differential of 78 - 81 or -3, while the away team has a differential of 81 - 82 or -1. Subtracting the two we get -2, meaning the home team's offense is two points more injured, relative to their highest Madden rating, compared to the away team.

We start with offensive injuries:

We can see from the ATS table above that on offense, fading offensive injuries has been a huge winning angle. If the home team is two points or more injured on offense than the away team, they have covered the spread by 1.2 points on average and have won 9 units over 219 games, while if the away team is two points or more injured, they have covered by 0.9 points and have won 14.5 units over 208 games. Overall the angle would have won 23.5 units over 427 games, a very solid return.

Looking at the average point spread, the market appears to value offensive injuries highly, as there is a difference of 3.2 points between the two categories of games. This would indicate that our Madden approach has done a decent job of figuring out when players are hurt. But at least over this sample of games, this hasn't translated to nearly as large of a difference in performance - the market prices in 3.2 points, but the difference between the "away injured" and "home injured" groups is only about 1.1 points. Perhaps this is why systems like the Massey-Peabody that ignore injuries have done pretty well. Although offensive injuries do matter, you would have been much better off completely ignoring them than over-accounting for them has the market has done.

Moving on to defense:

On defense we don't see any obvious winning angle. If anything, defensive injuries seem to be slightly under-valued by the market, as betting against the team that is more hurt on defense would have shown a slight profit of 2.2 units in 337 games, and covered by about 0.7 points per game. But this is well within statistical variance, and it appears these injuries are being priced in correctly.

Finally, quarterbacks, perhaps the most interesting category:

Here we have what may be another winning angle. While there is over a 6 point difference between home QB-injured teams and away QB-injured teams, suggesting the average QB injury is given an impact of 3 points, it turns out this wasn't enough. Betting against the team with the QB out would have won 23.3 units in 417 games, with a very healthy average cover margin of 1.6 points per game.

The teams with the injured QB in these situations play only very slightly worse than the market expects, with the main driver of the difference being that they average about 0.01 more turnovers per difference in 1 Madden injury point after adjusting for the market point spread. One turnover is worth about 4.2 points, so this ends up being about 0.04 points per game, per Madden injury point. There is also a very slight difference in yards per play differential as well. Overall with such small differences it is tough to say whether backups have simply had a tough run of luck in these games, but the angle seems promising.