Like anything involving statistical analysis, the bigger the data set, the less minor tweaks impact the results. For example, as I mentioned a few pages ago, the Patriots offense almost literally gave away 24 total points due to 3 special teams fumbles inside their own 10 yard line and a rare Brady pick inside the NE 15 yard line. Those resulted in 3 TD's and a FG, which really had nothing to do with the defense. I am normally not a fan of "changing" results or filtering data, but if those 24 points were removed from NE's ledger on defense, the last 8 games model would look like this instead:
NE OFF 30.0NE ADJ. OFF +8.5ATL DEF 20.5ATL ADJ DEF -2.7NE OVERALL = 28.2ATL OFF 37.5ATL ADJ OFF +14.3NE DEF 9.9NE ADJ DEF -10.3ATL OVERALL = 25.7
Those plays swung the pendulum on the Atlanta offense side by 3 points. But those plays were really bizarre. The punt returner that had the ball bounce off his foot at the 2 yard line is no longer returning punts. One of the kickoff fumbles was from a returner that only returned that one kick the entire season. And the INT came off of Michael Floyd's hands, who probably will be inactive for the SB. Weird, quirky plays happen from time to time, but THOSE particular quirkly plays are very unlikely to happen in the SB. I would have bet that it would be years before we saw NE fumble on back to back plays (and I would have lost) . . . and then they go and do it again within a month. Go figure. A trifecta in another month? I would be shocked.
The point being, those points technically get hung on the NE defense, but overall the defense has performed even better when you look at the mess the offense left them to deal with.