Nick Bosa

The Interesting Case of Adjusted Games Lost

Arjun Menon of the Michigan Football Analytics Society took a closer look at some of our adjusted games lost numbers over the last decade, identifying the teams with the best and worst injury records. Among his findings: injuries correlate somewhat with wins and they correlate somewhat from year-to-year but interestingly they don't correlate with snap-weighted age, meaning that older teams don't particularly suffer more injuries to their starters.

View Full Article

Comments

4 comments, Last at 10 Jun 2021, 10:22am

1 That's a great article, and…

That's a great article, and I really like the style of the team-by-team graphs. One question that comes to mind is whether there is any "reverse" causation in the relationship between AGL and wins. It seems possible that if a team is out of the playoff picture in the last month or two of the season, veterans that are dinged up would be more likely to sit out entirely than if a team is a contender, in which case they might tough it out and play. In other words, a team's winning percentage may have an effect on AGL, in addition to AGL having an effect on winning percentages.

2 So I would be justified!

in wanting Julio despite people clamoring injuries. Doesn't have to do with age really but medical staff of the Falcons likely more so.

Always nice to have priors backed.

3 but interestingly they don't…

but interestingly they don't correlate with snap-weighted age, meaning that older teams don't particularly suffer more injuries to their starters.

One long-standing idea is that some players are just injury-prone ('the best ability is availability'). If that were true, and you have a bunch of rookies entering the league, you'd expect a rash of injuries as the injury-prone get injured. Then, over time, survivor effect would tend to reduce the injury rate of veterans. As veterans aged, however, you might expect the effect to reverse. It's not clear whether such a trend could be detected in the data, given all of the other confounding variables.

4 I think there could also be…

I think there could also be a simpler effect: the older, injured players are being explicitly down-weighted in snap-weighted age. Assuming backups are younger than starters, a higher AGL would lead to a lower snap-weighted age (for a particular team). So even if older players are more likely to get injured, this effect is covered up when looking at age through the window of snap-weighted age. Right?