Bill Connelly's (S&P+) team statistical profiles are now live

Submitted by Bambi on October 2nd, 2018 at 6:53 PM

Bill C., the man behind the famed S&P+, tweeted out yesterday that his team statistical profiles are now live. It gives a bunch of advanced stats like post-game win expectancy, performance percentiles, and more to look at for each team on a week to week basis, plus a whole lotta advanced stats. I'll link the page below and also list some interesting notes in the OP.

Link to Michigan's profile


  • Our post-game win expectancy against NW was 81%. That means if this game were played 100 times with the same statistical profile for each team, Michigan would win 81% of the time. This indicates the game was not as close as the final score from a box score perspective. *Note that this isn't opponent adjusted, so whether it's Michigan vs NW or Bama vs UTEP this 81% would hold true. The 81% is based purely on the box score, nothing to do with the teams playing.
  • This was our first non 100% post game win expectancy since ND, which was 59%.
  • Michigan's overall performance this past week against NW graded out in the 70th percentile, meaning that when adjusted for opponent and put on a Bell curve, we performed in the 70th percentile. This was our worst performance of the year, with SMU at 75%, ND at 86% and Western maxing out at 93%.
  • Offensively we had our second worst performance at 69%, only behind ND at 32%. The other 3 games were all between 77% and 85%.
  • Defensively we had our worst performance at 77%, with the other 4 games between 83% and 94%.

Non Michigan notes:

  • MSU has not had a post game win expectancy of 100% this year. They maxed at 95% against Indiana had their lowest was 52% against ASU.
  • Their lowest overall performance percentile was 49% against CMU and the highest was 95% against Indiana.
  • Their lowest offensive percentile was 29% against ASU and their highest was 93% against Indiana.
  • Their lowest defensive percentile was 48% against Utah State and the highest was 89% against Indiana.


  • OSU has had 3 post game win expectancy's of 100%, while TCU was 62% and PSU was 51%.
  • Their lowest overall performance percentile was 74% against TCU and the highest was 91% against Tulane.
  • Their lowest offensive percentile was 74% against TCU and their highest was 95% against Tulane.
  • Their lowest defensive percentile was 26% against TCU and the highest was 94% against Rutgers. Their percentile against PSU was 55 and against Oregon St. was 69.

Keep in mind that preseason projections are still part of S&P+ until about week 7, with the percent they make up decreasing each week. So things like OSU's defense being a top 25 unit despite 3 bad performances are likely held up by preseason where they were projected to be 8th.

Once again if you have some time go check out the spreadsheet. It has a ton more info with individual player and scenario based ratings. You can use this glossary to help explain what you don't understand.



October 2nd, 2018 at 7:18 PM ^

These have actually already been up. They are not up in SB page format (such as from what I can tell, but I used these last week for some bets.


October 2nd, 2018 at 7:26 PM ^

If I'm reading it correctly, our stats support our bitching about play calling on first down: we rank #97 for "Pct. of first downs coming on 1st or 2nd", supported by the data point that we rank third in "average third down distance". So we slowly plow our way down the field compared to most other programs (= Bo/Harball). And we also suck in the red zone ("11- to 20-yard line success rate = 29.6% = 109th ranked"). 


October 2nd, 2018 at 8:01 PM ^

He’s had em up for a few weeks now, I just hate that’s its a google doc now and not a published website like before. Also I don’t think categorized O and D numbers have take until week 7 to go up on Football Outsiders before either.


October 2nd, 2018 at 8:55 PM ^

These “scientific” methods as to who will win a game are no better than a knowledgeable observer who makes predictions based on that knowledge. S&P (even now with a + now attached) are useless, just like PFF. Fake science. 

Frank Chuck

October 2nd, 2018 at 9:43 PM ^

Ah, yes...

...The typical anti-academic/anti-analytics take that has become so popular within a particular demographic in recent years.

So let's review.

Math helped (among many, many achievements):
- build ICBMs which can go from one continent to another in 20 minutes and wipe out entire cities and large pools of human population within seconds

- build advanced systems which can simultaneously keep real-time, tracking profiles of billions of humans

- create a network that can send a digital object (i.e. email with attachments) from one side of the world to another in the blink of an eye

- build algorithms that can successfully identify and control outbreaks/diseases

- perform impossibly long computations of many variables


But according to you, math is incapable of more accurately explaining or predicting why a set of 11 humans in colored uniforms excels or struggles to move an object from point A to point Z in a rectangle-shaped space against another set of 11 humans in different colored uniforms.

You have to be some kind of idiot.

I've often wondered: we know a smart person is aware of his or her intelligence. But does a dumbass comprehend the depths of his or her stupidity?


October 3rd, 2018 at 9:32 AM ^

Not sure if Yooper forgot the "/s" or not, but . . . there is a way to determine if S&P+ is "no better than a knowledgeable observer who makes predictions". In fact Bill C looks at this exact question every week by seeing how S&P+ does against the spread. Thus far in 2018 it is 133-118-2 (53%) against the spread. See for yourself by following the link below.


October 2nd, 2018 at 8:56 PM ^

Keep in mind that preseason projections are still part of S&P+ until about week 7, with the percent they make up decreasing each week.

Actually, I believe Bill said that he's leaving them in all year now.

Preseason projections will remain in the formulas all season. Fans hate this — it’s the biggest complaint I’ve heard regarding ESPN’s FPI formulas. Instinctively, I hate it, too. But here’s the thing: it makes projections more accurate. Our sample size for determining quality in a given season is tiny, and incorporating projection factors found in the preseason rankings decreases the overall error in projections. So I’m doing it.


October 2nd, 2018 at 10:30 PM ^

Yeah, I mean, I don't think people should hate this as much as they do, especially since the factor still decreases with time, as it should. Should we really be that surprised that quality of recruiting classes, recent team history, and experience level of the team, which are the factors that go into the preseason projections, should continue to provide explanatory power as the season unfolds? No, not at all. People really like to think that each instance of a team is some sort of special magical snowflake, with all sorts of inexplicable, unique phenomena that power it to wins or doom it to losses, at that's certainly true to some extent, but at the end of the day mean reversion is real, and it is a bitch.


October 2nd, 2018 at 10:32 PM ^

As or more importantly, he said he would speed up how quickly preseason projections start getting reduced. So yes, there will still be some of the preseason projections in there at the end, but at this point, there's likely less of the projections in there than there was at this time last season.

Stringer Bell

October 2nd, 2018 at 9:36 PM ^

I have to wonder how OSU had a post-game win expectancy of >50% against PSU, in a game that PSU statistically dominated.  Felt like OSU was lucky to pull that one out.


October 2nd, 2018 at 10:36 PM ^

Well, the post game win percentage is based on the five factors:

If you win the explosiveness battle (using PPP), you win 86 percent of the time.

If you win the efficiency battle (using Success Rate), you win 83 percent of the time.

If you win the drive-finishing battle (using points per trip inside the 40), you win 75 percent of the time.

If you win the field position battle (using average starting field position), you win 72 percent of the time.

If you win the turnover battle (using turnover margin), you win 73 percent of the time.

Now from looking at the five factors box score, I don't see anything for explosiveness, but OSU had a higher success rate, won the drive finishing battle, and had better field position. Neither team turned the ball over. I don't agree at all that PSU dominated the game. They did blow a lead, but that game could have gone either way if they replayed it.

Frank Chuck

October 2nd, 2018 at 9:37 PM ^

Didn't Bill C say that he wouldn't totally phase out the pre-season data starting this season because he retroactively found that using pre-season expectations actually improved accuracy?

I recall him writing a post along those lines.

Edit: Zenogias made the same observation a few posts above.


October 3rd, 2018 at 9:57 AM ^

I'm new to S&P+ but can someone explain the 59% postgame win expectancy for ND?


We had 58 net rushing yards.


October 3rd, 2018 at 11:21 AM ^

Viewers of this thread are probably non-existent at this point, but I wanted to look at some OSU numbers compared to Michigan's:

Standard Down Success Rate: UofM = 55% (Ranked 13th) OSU= 57% (Ranked 5th)

Avg. 3rd Down Distance: UofM = 5.9 yards (Ranked 3rd) OSU = 6.2 yards (Ranked 11th)

3rd Down Success Rate: UofM = 52% (Ranked 11th) OSU = 44% (Ranked 31st)

I thought it was an interesting comparison, especially considering how many vocal Michigan fans seem to prefer an Urban Meyer offense over a Jim Harbaugh offense. I'm so far from an expert that I probably shouldn't even follow up on this sentence, but I have actually been quite impressed with the efficiency of the offense this year. I think this is getting pretty close to Harbaugh's ideal, coupled with few turnovers and an improving/elite defense.