Yeah, the average correlation between overall offensive and defensive output (measured by points) is 0.289 since 1978 so in statistics the "variance explained", or how much of the variation in one is explained by the other, is simply the square of that correlation = 0.289^2 = 0.0835 = 8.35%. This means that on average less than 10% of defensive output is influenced by an offense's ability to score points and vice versa. Having said that, mooseguts does point out something interesting which is that certain aspects of an offense might correlate real well with defensive output and we just might have been real bad in specifically those areas. For example, in 2017 the correlation between "percent of offensive drives ending in turnovers" and points allowed is a whopping 0.629! And as mooseguts points out we were 4th worst in that stat in 2017. Here's the problem. Even if that's the case, it's still true that whatever influences offensive output is on average not influencing defensive output that much. This means that there MUST be other areas of the offense that are compensating for being bad in a particular area that happens to be highly correlated with defensive output. If there weren't such a compensating factor then the correlation between overall offensive and defensive output would also be high, but it isn't. Another problem is that it's not clear what's causing the observed correlation. For example, it's true that we were dead last in 3rd down conversions which has a -0.32 correlation to points allowed in 2017. But guess what? 3rd down conversion RATE has a -0.08 correlation to points allowed! And it's the rate that really matters. In other words the total number of 3rd down conversions isn't a purely offensive stat! You look at the rate of 3rd down conversions and you find that even if we're dead last there too it doesn't affect points allowed in any way worth mentioning. And a similar argument can be made for time per offensive drive which has a pretty high -0.539 correlation to points allowed. But how much of that is having to pass more when behind in the 4th quarter for example? That is, how much of the high correlation is due to being behind for OTHER reasons? That's why in the end it's best to look at correlation between overall offensive and defensive outputs. Correlation doesn't tell you causation, but you HAVE to have a correlation for a causation, unless there's some other mechanism that cancels the one you're looking at out. And when you look at things with a scalpel (correlations for individual component stats) you really don't know what those other compensating factors are, if any.
Well since turnovers either directly result in points scored by the opposing D or indirectly result in extra points by giving the other team more opportunities and usually in better field position, there should be a high correlation between turnovers and points allowed. I was thinking however, that maybe the better thing to measure is offensive points scored and defensive points allowed, i.e. removing returns. By keeping the return TDs in the mix you could be giving credit where credit isn’t due. It may all wash out in the swings and roundabouts in the end though.
Yeah it's so stupid that they count pick 6's as offensive points scored, etc... It would be great if they actually gave us NET points scored by each unit but no one does that apparently. Only way to get the data you're suggesting is play-by-play and even then that takes some scrutiny to make sure you're not missing something. Probably does wash out.
Well on this page https://www.pro-football-reference.com/teams/mia/2017.htm there is a table labelled scoring summary. From that you can extract passing TDs rushing TDs, PR, KR, int, fumble and other (blocked kicks ?). It also gives the summary for the team’s opponents. From there you could then track offensive TDs against defensive/special team TDs. Maybe if you want to be extra fine you could track the XP attempt to each TD, or assume each TD had a made XP, or even leave XPs off the total scoring summary. For me I’d have to do it manually but you mentioned you had some scripts that could do it automatically. On this page https://www.pro-football-reference.com/years/2017/ there is a table called scoring offense, but without a matching scoring defense table somewhere you can’t adjust offense and defense. For example Jacksonville had 8 of their 47 TDs come from returns for 39 offensive TDs, and Oakland had 0 of their 36 TDs come from returns, which would suggest Jacksonville and Oakland had similar offenses, yet their points scored difference was 415 to 301, and at least half of that difference cam from return TDs.
Hey thanks! I actually never looked down that far on those pages even though I scraped data from them lol. In any case, PFR usually has their defensive stats on a URL that ends in opp.htm so the corresponding defensive stats (including scoring) are at: https://www.pro-football-reference.com/years/2017/opp.htm
