I’ve said this before, but one thing I like about the advanced stats gurus is that they’re open to tweaking their metrics in pursuit of finding more accurate ways to correlate the raw data to things like wins and losses as well as the relative strength of programs. The tough part, of course, is the relatively small sample size of the regular season, and that’s not gonna change.
In any event, Bill Connelly has a go at bringing more meaning to strength of schedule with this post. I find it interesting; see what you make of it.
9 responses to ““Team A ain’t played nobody!””
Before these games there’s always doubt but once you boat race them they’re a nobody. When you are good, performing well and executing, everybody becomes a nobody. Isn’t that the point?
LikeLiked by 2 people
Aside from the acknowledged issue of small sample size, there are two things I don’t agree w/ regarding his formulas.
First, it wanders too far into the land of the transitive. I get it to an extent, but looking better against certain teams by a wider margin against various opponents really doesn’t do anything for me.
Second, and maybe I missed something he addressed, as I didn’t carefully read it from top to bottom, so correct me if I’m wrong, but I don’t think clobbering the #80 team holds any more value than clobbering the #120 team. They’re obviously both awful teams, and at some point, I feel like you have to almost lump all those together. #1 would be considered a far better team than #15, but the disparity would likely be far smaller for the #41 team and the #55 team. Again, though, he may have accounted for that and I just missed it.
Similarly, I think conference games should be weighed much more than non conference. I know each season has a different team but there absolutely is value repeat matchups. The sample size can be doubled by taking a 100% value from the current season, then making a curve that gives a 80% for last year and then maybe taking two seasons ago as a 40%. There is carry over and sample size can be less of a problem.
In that case, Notre Dame would return a (null) and blow up Bill’s computer.
I got a smile out of one stat in Bill’s post: Alabama’s 114 claimed national titles
LikeLiked by 1 person
If you were going to do something like this, it would make more sense to use several models instead of just S&P+. You’d still need to apply an eye test, because the sample size is eternally too small.
FSU in 2013 and 2014 was a poor choice to prove his bona fides, too. Without looking at advanced stats I knew they were going to beat Clemson in 2013 and lose in the playoff in 2014.
It is interesting, and I applaud another way to measure relative strengths. We can all nit pick the S&P but if you look at the additional depth it provides from, say a Sagarin computer model, it adds to the discussion with additional data points. I have always had issues with Sagarin and the ESPN index as well. Somewhere in a blended fashion may have some value. Of course the best solution is take the 5 conference champions and only apply the “subjective picks” to the last three teams added to complete the 8 team playoff we sorely need. 🙂
That guys numbers and formulas make my head hurt..