A Beginner's Guide to Reading a College Football Spread (From 8,000 Games)

Analysis · by Rick's Picks Analytics

If you're new to college football analytics, the spread is the first number you have to learn to read fluently. It looks simple -- "Georgia -6.5" -- but packed inside that number are a sign convention, a pricing structure, a distribution of final margins, and a market that has been sharpening itself for decades.

We maintain a database of more than 40,000 historical college football games. Everything in this guide comes from the 7,962 FBS-vs-FBS games in that data with closing spreads -- 5,649 from 2015-2022 (our training window) validated against 2,313 from 2023-2025 (our test window). No intuition, no tout-speak. Just what the numbers say.

What the number actually means

A spread is the market's projection of the final margin, expressed from the home team's perspective. The convention that trips up every beginner:

Negative means the home team is favored.

• Home -6.5: the home team is projected to win by more than 6.5. They "cover" if they win by 7 or more.
• Home +3: the home team is the underdog by 3. They cover by winning outright or losing by 1 or 2.
• If the game lands exactly on the number (home favored by 3, wins by exactly 3), that's a push -- nobody wins, stakes are returned.

Pushes can only happen on whole numbers. A spread of -6.5 can never push, because no football game is decided by six and a half points. That half point is called the "hook," and as you'll see below, where the hook sits matters enormously.

The vig: why 50% isn't good enough

Spread positions are typically priced at -110: you risk $110 to win $100. That extra $10 is the vig -- the market's fee for taking your action.

The math falls out directly: to break even at -110, you need to win 110 / (110 + 100) = 52.38% of your picks. Not 50%.

This is the single most important number in all of spread analysis, and it's why we treat 52.38% -- not 50% -- as the null hypothesis in every backtest we run. A model that picks winners 51% of the time sounds like it "beats a coin flip." It doesn't beat the vig. It loses money slowly, which is arguably worse than losing it fast, because it takes longer to notice.

Why -3 and -3.5 are completely different lines

Here's where the data gets interesting. Football margins are not smoothly distributed -- they cluster hard on a few numbers, because points come in chunks of 3 and 7. From our 5,649-game training set:

Nearly one in ten college football games ends with a 3-point margin. Another 8.5% end on exactly 7. Together, those two numbers alone account for roughly 18% of all outcomes. These are the key numbers.

And this isn't a fluke of one era. When we validated on 2023-2025 data, the top-10 most common margins overlapped 90% with the training set. In the 2025 season alone (807 games), margin 3 hit 11.4% and margin 7 hit 8.3%. Key numbers are stable because they're a property of how football scores points, not a market quirk.

Now read those two lines again:

• -3 means the game can push on the single most common outcome in the sport.
• -3.5 means the favorite must clear that outcome entirely. If they win by exactly 3 -- which happens 9.6% of the time -- the -3 position pushes while the -3.5 position loses.

That half point is worth roughly a 9.6% swing in one specific outcome. Compare that to the move from -4.5 to -5, which crosses margin 5 -- a number that shows up in only 2.6% of games. Half points are not created equal. Their value is exactly the probability mass of the integer they cross.

One honest caveat, since some books let you buy that half point for extra juice (typically -120 instead of -110): we computed the expected value of buying through every integer margin, and at -120 pricing, buying through 3 is still negative EV (about -5.5 cents per dollar) and so is 7 (about -6.7 cents). Only one obscure margin (28) cleared the bar in our data. The market knows this distribution as well as we do, and it prices the half points accordingly.

"This team is 9-2 against the spread!" -- why sample size is everything

You will constantly see cover-rate claims: a team is 9-2 ATS at home, a coach is 7-1 ATS as an underdog. Here's the honest way to read them.

A cover rate from a small sample is a guess with a huge error bar. The tool we use to size that error bar is the Wilson confidence interval -- in plain English: "given this record, what's the plausible range for the true underlying rate?"

A real example from our own research. We tested whether a line crossing the key number 3 toward the home team (from open to close) predicted covers. In training data, that angle went 9-2 (81.8%). Sounds incredible. But the Wilson 95% interval on a 9-2 record runs from 52.3% to 94.9% -- the honest statement isn't "this hits 82%," it's "this hits somewhere between barely-break-even and amazing, and we genuinely don't know where."

Then we ran it on the test window. The same angle went 17-21 across 38 games -- a 44.7% cover rate. Below a coin flip, well below the 52.38% break-even. The 82% was noise wearing a costume.

The rule of thumb: an 11-game ATS record tells you almost nothing. Even a 100-game record at 55% carries an interval that dips near break-even. This is why every result we publish comes with its sample size and interval attached, and why we run Bonferroni corrections when testing many angles at once -- if you test twenty ideas at the standard significance threshold, one will look "significant" by pure chance. Most published ATS trends are exactly that.

Opening lines vs. closing lines

The number posted Sunday night (the opener) and the number at kickoff (the closer) are often different. The closing line is generally the sharper number -- it has absorbed a week of information, injury news, and market pressure.

A popular theory says you should follow that movement: if the line moves from -3 to -4.5, the market is "telling you" to take the favorite. We tested it. Across 1,727 test-window games with recorded open and close, siding with the direction of line movement covered just 48.9% of the time (95% interval: 46.6% to 51.3%). In training data it was 49.4%. Not just short of break-even -- short of a coin flip.

The honest interpretation: the closing line being accurate is not the same as line movement being predictive. By the time you can see the move, the new price already reflects it. What line movement is genuinely useful for is context -- knowing whether the number you're evaluating sits on, near, or safely past a key number.

The bottom line

• The spread is a margin projection; negative = home favored. Half-point lines can't push.
• Break-even at -110 pricing is 52.38%, and that's the bar every claim should be measured against -- including ours.
• Margins pile up on 3 (9.6%) and 7 (8.5%), so a half point matters enormously at those numbers and barely at all elsewhere. That structure held at 90% overlap across a decade of train/test data.
• Any cover rate without a sample size and a confidence interval is a story, not evidence. We watched an 82% angle collapse to 45% out of sample.
• The closing line is sharp, but chasing its movement covered 48.9% in our test data -- a negative result we're publishing anyway, because that's the point of this site.

Everything above comes from the same methodology we apply to every hypothesis: train on 2015-2022, test on 2023-2025, correct for multiple comparisons, and report the result whether or not it flatters us. The market wins most arguments. Learning to read the spread properly is how you figure out which arguments are even worth having.

Rick's Picks publishes statistical analysis of college football for informational and entertainment purposes. Nothing here is betting advice. 21+. If gambling is affecting you or someone you know, call 1-800-GAMBLER.