South Carolina went 4-8 last season, and 1-7 in the SEC. The rating this site leans on has them 24th in the country. Both of those are true, and the record is the one that will cost you money.
That rating is SP+. It does the job for college football that Pythagorean wins do for the NFL, and every college preview we publish is built on top of it. Here is how to run it yourself, and where it stops being useful.
The simple explanation
SP+ is a power rating for college football. Bill Connelly built it and ESPN publishes it. He created the system at Football Outsiders in 2008, and he describes it as a tempo and opponent adjusted measure of college football efficiency.
Strip the jargon out of that. It grades how well a team plays on a per-snap basis, then corrects for two things the raw numbers get wrong: who the team played, and how many snaps it ran. A slow team and a fast team can be equally good and look nothing alike in a box score. SP+ puts them on the same scale.
That scale is points. A team rated 26.4 is about that many adjusted points per game better than an average FBS team. Which means the ratings subtract, and that is the whole reason the number is worth learning.
One more thing before you use it. SP+ is predictive, not backward-looking. ESPN says so directly: it measures “the most sustainable and predictable aspects of football, not a résumé ranking.” A 10-2 team can rate below an 8-4 team without the rating being confused. It is answering a different question than the standings are.
A worked example, using two real teams
Take the two teams from our own previews. Final preseason SP+ has Georgia fourth in the country at 26.4. It has South Carolina 24th at 12.6.
Subtract them. That is a gap of 13.8 points. If those two met on a neutral field, SP+ says Georgia by 13.8. Put the game in Athens and you add the home team’s 2.5, so Georgia by 16.3. ESPN spells the method out in plain terms: subtract one rating from the other, then give the home team an extra 2.5 points.
No subscription and no model access. Two numbers and a subtraction, and you have a spread to hold up against the one a sportsbook is posting.
Turning a rating into a win total
This is the part that matters for season bets, and it is where our previews spend most of their time.
Connelly runs that projected spread against every game on a team’s schedule and counts them up. Favored by more than 7.5 points is a projected win. An underdog by more than 7.5 is a projected loss. Anything inside that range is a tossup and counts as half a win. Add the twelve games and you have a projected total to set against the number the book is offering.
Georgia came out at 9.9 projected wins. Texas Tech, eighth overall at 21.8, came out at 10.6. Those are the figures our previews argue with, and sometimes argue against.
I do not use the raw projection. Across the 28 college teams where we hold both an SP+ number and a posted line, SP+ runs about 0.15 wins below the market, so every projection gets adjusted before it goes anywhere near a pick. A rating that sits slightly low is not a problem as long as you know it and correct for it.
Why we do not use Pythagorean wins here
The NFL guide on this site runs points scored and points allowed through a formula and gets an expected record. That works in the NFL because everybody plays seventeen games against other NFL teams. College breaks both halves of that.
Twelve games is a thin sample to begin with. Worse, those twelve are not shared. Georgia’s schedule and a Sun Belt team’s schedule have almost nothing in common, so their point differentials are not measuring the same thing. And the big programs post 45-point wins over opponents they were never going to lose to. Those margins run straight into the points column, and the formula reads them as dominance when they are really a scheduling decision.
SP+ handles both. It adjusts for opponent, so a narrow win over a good team is worth more than a rout of a bad one. And it grades efficiency play by play, so the sample is thousands of snaps rather than twelve final scores.
What it cannot do
Preseason SP+ is mostly a bet on who came back. Connelly is open about the recipe. Returning production accounts for more than 60% of the preseason projection. Recent history is a little more than 20%. Recent recruiting is about 14%.
Read that again, because it tells you exactly when the preseason number will be wrong. It will underrate the team that lost a lot and reloaded well through the portal. It will overrate the team that returned everybody and was not very good in the first place. It has no opinion at all on a coordinator hired in January.
It also will not tell you who wins on Saturday. A 13.8-point projection is the middle of a very wide range, and college football produces more upsets than the NFL does. SP+ is a season-long instrument. Treat it like one.
What trips people up
The rank is not the rating, and people quote the rank. Our own previews show why that matters. Georgia is fourth at 26.4 and Texas Tech is eighth at 21.8, so four places cost 4.6 points. Texas Tech down to South Carolina is sixteen places and 9.2 points. The ranks are evenly spaced and the ratings are not, so read the number.
A gap is not a bet either. A team projected a win and a half above its posted total is worth reading into, not backing on sight. You still have to ask why the market disagrees, and the market is usually looking at something a preseason model has not priced yet.
Then there is the one that catches people every December. A conference title game and a playoff game do not count toward a regular-season win total. A team can be projected at 10.6, win eleven, and still lose your Over if one of those wins came in the title game. Read the book’s own rules before you take a number.
And price comes last, which is the step everybody skips. A two-win edge at a bad number is worse than a half-win edge at a good one.
How to use this
Look up the team at ESPN and write down two things: the rating and the projected win total. Set the projection against the posted total. Half a win apart is noise. A full win or more is worth the next twenty minutes.
Then do the rest of the work. Start with returning production, since that is most of what the preseason number is made of. Look at what happened in the portal. Check the schedule for the two or three games that decide whether the total gets there. Then look at what you are being charged. For how much money to put behind an opinion once you have one, the guide to units covers that side of it.
Every college preview we publish shows this math and names the number it used, so you can see where the edge is supposed to be and disagree with how we read it.
Quick reference
What it is: Bill Connelly’s tempo- and opponent-adjusted efficiency rating, published at ESPN
What the number means: adjusted points per game better than an average FBS team
Game projection: subtract the two ratings, then add 2.5 for the home team
Win total: a projected win at better than 7.5, a loss at worse than minus 7.5, half a win in between
Preseason recipe: returning production 60% or more, recent history just over 20%, recruiting about 14%
Gap to act on: roughly a win or more against the posted total
Does not count: conference title games and playoff games sit outside a regular-season win total
Do not use for: picking a single Saturday winner
One last thing
SP+ will not tell you that South Carolina is going to win eight games. It tells you that a 4-8 team is rated 24th in the country, which is a strange enough sentence that somebody ought to go and find out why. That is the job. The rating is where the question starts, not where it gets answered. When you are ready to shop a number, our sportsbook reviews cover which apps actually price college futures.
21+ · Gambling problem? Call or text 1-800-GAMBLER