Here’s the beatpaths graph for Week 6. Tonight’s game between Chicago and Arizona is essentially meaningless for now, as Chicago has multiple beatpaths to Arizona. Either a Chicago win will reinforce existing beatpaths, or an Arizona win will be declared a beatfluke. The picks stunk this week at 5-7 so far, missing on PIT=>KC, CAR=>BAL, SEA=>STL, NO=>PHI, TB=>CIN, DET=>BUF, and TEN=>WAS . But, upsets also lead to more changes in the beatpath graph, which is always more interesting. One item of note, that rickety beatpath I kept mentioning of BUF=>MIN – it finally fell apart this week.
Well, I’d certainly call that game a beatfluke. Not in the graph of course, but that name certainly fits a ‘victory’ like that.
I don’t know–given that the Cards “should” be 4-2, shouldn’t we see the win as a hard-fought victory against a good team on the road?
Interesting to see how St. L.’s defeat of the Broncos edges Chi and Sea to the top of the graph because the Rams lost this week.
Something I have wanted to ask about for a while–you’ve noted a couple of times that your graphs are accurate at retroactively predicting games from previous weeks as evidence of the accuracy of the beatpaths model. But given that your data is the wins and losses, isn’t this analogous to saying that someone translating “Chargers blow” into French and back again proves the accuracy of the original translation? Or do the retroactive predictions prove the accuracy of the way the model breaks beatloops and exposes beatflukes?
Nope, I never meant that the “retrodiction” accuracy was a good harbinger of future accuracy. I just think it’s interesting.