NFL Win Totals vs Polymarket Championship Odds

The Detroit Lions looked cheap to us all through the prep for Episode 10.

BetMGM had them at 10.5 wins with the over favored at −120, which is roughly fourth-best in the league on expected record, running one of the softer schedules in football. Polymarket was selling a Detroit championship for 4.25 cents. We talked ourselves into buying them on air, along with the Bengals. It took about four minutes.

The schedule angle wasn't new to us. I'd flagged Detroit's soft strength of schedule back on Episode 9, when they were going for 4.2 cents, and bought a little then. But four cents is a long way from what that win total looked like it was worth, and markets are usually cheap for a reason. I wanted to dig deeper independently and validate our simple observations during show prep.

The model I built

A win total and a championship price aren't in the same units, so there's no way to look at 10.5 and 4.25¢ and say which one is wrong. One is about how many games you win. The other is about winning the last one. Converting between them means running the season, so I built something that does it in four steps.

Step 1: Strip the vig out of the line

A sportsbook price isn't a probability until you take the margin out. Detroit's 10.5 was priced −120 on the over and +100 on the under, which implies 54.5% and 50.0%. Those sum to 104.5%, and that extra 4.5 points is BetMGM's cut. Normalize it away and it's 52.2% to go over. Our no-vig calculator does this on any two-way market.

Step 2: Turn that into an expected record

A 52.2% chance of clearing 10.5 wins isn't the same as a prediction of 10.5 wins. Season records scatter around their true mean with a standard deviation of about 2.9 wins, so working backwards from that 52.2% gives Detroit an implied mean of 10.66 wins. That's fourth best in football. This is also the one number in the whole model that's a judgment call, which is why it gets stress-tested below.

Step 3: Play out 400,000 seasons

Each team's expected record becomes a per-game strength rating. Then the simulation draws 400,000 full seasons and plays each one properly: four division winners and three wild cards per conference, seeded by record, the bracket reseeded every round, home field worth about 57% to the host, and the Super Bowl at a neutral site. Then it counts how often each team finishes holding the trophy.

Step 4: Scale to Polymarket's own book

Polymarket's championship prices sum to 104.05% rather than 100%, the same kind of overround we measured at 102.2% on the World Cup field. The model's output gets scaled to Polymarket's book before any gap is calculated. Skip this and that 4% shows up as fake edge spread across all 32 teams, which would make every single one of them look like a buy.

Detroit Lions Super Bowl odds: our model vs Polymarket

Detroit came back at a 7.15% title chance: 10.66 projected wins, 66% to make the playoffs, 12.9% to win the NFC. Polymarket was charging 4.25%.

Put as a price, you're being laid 22.5-to-1 on something the conversion calls a 13-to-1 shot. A thousand dollars at 4.25¢ buys 23,529 shares, which pay $23,529 if Detroit wins. That happens 7.15% of the time, so the expected value is $1,682, or +$682 on average. It also means losing the entire thousand about 93 times in 100. Positive expectation and a good night are different things.

My first instinct was that the market must be pricing the path. Detroit shares a conference with the Rams at 11.91 projected wins, the Seahawks at 10.74 and the Eagles at 10.02, which is a rough road to a Super Bowl. But the simulation runs that exact bracket, with those exact teams standing in the way, 400,000 times. Detroit is cheap despite the path rather than because anyone forgot it was there.

So I bought more.

The teams Polymarket underprices

Sorted by excess return per dollar, which is the model's probability divided by what Polymarket charges. Implied wins is the line and its price read together, and it's often a different number from the line itself.

* Implied wins is what the line and its price say together, not the line itself. Detroit's 10.5 is priced −120 to the over, which de-vigs to 52.2% and works back to a mean of 10.66 wins. Cincinnati's 9.5 is priced −155 to the over, so it implies 10.11. The posted number is only half the information.

The Rams row is the one I find most reassuring. They're the model's clear favorite at 16.75% against 14.50% quoted, and the return is only +15.5%, because Polymarket has broadly got them right. If the model were producing free money on every team I'd assume it was broken.

Cincinnati is the row that reads wrong at a glance. A 9.5 line looks a full win below Baltimore's 11.5, but the Bengals' over is priced at −155, which is heavy, and the implied total comes out at 10.11. That's within 0.8 wins of the Ravens. Reading a win total without its price is how you misjudge a team.

The teams Polymarket overprices

* See the note above: implied wins reads the line together with its price.

Every NFL team's 2026 implied wins

Those two tables are the extremes. Here's the whole league, with both sides of every line shown so any row can be checked: take the two prices, convert each to a probability, normalize them back to 100%, and see what record the result implies. Click any column to sort.

Sorted by implied wins. BetMGM prices as of August 28, 2026. Implied wins is derived from both sides of the line, not the over alone.

Strip the vig out of all 32 lines at once and something falls out. The posted lines add up to 275.0 wins against the 272 that actually exist in a season, which looks like BetMGM being three wins long across the league. De-vig every line and the implied totals sum to 272.51.

The extra 2.5 wins are part of the vig. Always good to know where you're paying.

The other thing the posted lines hide is the eight-team pileup at 10.5, which splits hard once you read the prices. Buffalo, Seattle and Detroit lean over. The Chargers, Chiefs, Patriots, Eagles and 49ers lean under. Same headline number, a full win of difference underneath it.

Dallas Cowboys odds: why Polymarket overprices them

Polymarket ranks the Cowboys the sixth most likely champion in football at 5.25%. BetMGM's line ranks them fifteenth.

What makes it stark is that BetMGM has no opinion on Dallas at all. Their 9.5 is priced −110 both ways, the only true pick'em on a board of 32 teams. Every other team carries some lean. The book that sets numbers for a living looked at the Cowboys and shrugged, while the prediction market made them a contender. The model says 3.08%, so the quoted price gives back about 41 cents on the dollar in expectation.

The Patriots and Chiefs have the same line and different prices

New England and Kansas City both sit at 10.5 wins. Both prices imply the same 10.02 expected wins, to two decimals. As far as the sportsbook is concerned, these are the same team.

Polymarket pays 3.45% for the Patriots and 6.25% for the Chiefs, so Kansas City costs 1.81 times as much off an identical underlying number. Some of that's probably a real read on January experience that a season-long total can't capture, and I'd believe a version of this gap. I have a harder time believing it's worth 81%.

The rest looks like sentiment. Mahomes is a brand, and a lot of people are buying the story of him coming back rather than a number. Andrew made this exact point on Episode 9, when the Chiefs were sitting fourth on Polymarket at 6.7 cents: he wasn't convinced Mahomes would return worth anything, and thought he'd have to change his game completely to manage it. New England is the same trade in reverse. Young talent nobody has an emotional position in yet, and a price that reflects it.

Longshots cost about 15x what they're worth

The six cheapest teams on Polymarket are quoted between 0.35% and 0.95% each, adding up to 3.80%. The model puts them together at roughly 0.25%.

That's a markup of about fifteen times, and it isn't a Polymarket quirk. It's the favorite-longshot bias, and it turns up in racetrack data around the world with very few exceptions. Snowberg and Wolfers put a number on it: betting horses at 100-to-1 or longer returns about −61%, against −23% for betting at random and −5.5% for backing favorites. A share that costs a penny feels free, and a 200-to-1 payoff is easy to picture.

The individual numbers down here round to zero and I wouldn't quote any single one of them as precise. The aggregate is the finding.

Stress testing the model

Everything above rests on that 2.9-win standard deviation, so I re-ran the whole thing at 2.5 and at 3.3 to see which findings held and which were artifacts of the setting.

The same five teams at every setting, with only the internal order moving, and Detroit first or second throughout. That stability is the part I'd actually put weight on. The decimals are illustration; the ranking is the finding.

Polymarket also runs conference championship markets separately from the title market, and the two don't agree with each other. Detroit is quoted at 8.80% to win the NFC against a modelled 12.87%, which is a wider gap than the one in the title market. Same team, same underlying line, two books on the same venue that nobody has reconciled.

How to run this check yourself

You don't need 400,000 simulated seasons to do a useful version of this. Most of the value is in the first step, it takes about a minute per team, and it works against any venue quoting the same outcome, Polymarket and Kalshi included.

Strip the vig from the win total. Take both sides of the line, convert each to an implied probability, and normalize so they add to 100%. Our no-vig calculator does it on any two-way market, and the odds converter handles the American-to-probability step on its own.

Compare teams on the same line. This is where the easy money hides. Eight teams sat on 10.5 this year and the market split them hard, some leaning over and some under. Any two teams with the same headline number and very different championship prices are worth a closer look, which is exactly how the Patriots and Chiefs turned up.

Watch for the price, not the line. Cincinnati at 9.5 looks like a mid-table team until you notice the over is −155. A heavily priced line is telling you the number is stale, and the implied total is what matters.

Check the book against itself. Add up every team's championship price. If it comes to a lot more than 100%, that's the overround, and you're paying it on every position. Then check whether the conference markets agree with the title market, because sometimes they don't.

Variance is large: what to know before you bet

The 2.9-win standard deviation is a choice, not a fact. Converting wins into per-game strength by dividing by 17 is crude, and a 10.7-win team isn't necessarily 62.7% per game against everybody. Home field is a single number applied to all 32 stadiums. Team records are treated as independent when divisional games are zero-sum and schedules interlock. A season-long win total encodes a roster and a schedule, but a title run also turns on quarterback health in January and one-off coaching calls that a season number smooths away.

One practical caveat bothers me more than any of those. The Polymarket figures here are midpoints, and I couldn't verify executable spreads or depth, so the price you'd actually get may be worse than the one quoted. Polymarket also charges a fee on market orders, which comes straight off the top of any edge, so a gap this size is real but it isn't quite as large as the table makes it look.

We're long the Lions and Bengals, it's on the public scorecard with every other bet we've made on the show, and it resolves in February 2027. You can watch it fail there in real time if it does.

Although holding to February isn't the only way this pays. A high win total is really a bet on making the playoffs, and a team that starts 5–1 gets repriced long before anyone lifts a trophy. We can sell into that and take the profit without ever being right about the Super Bowl. Andrew put it well on Episode 8: a 20-to-1 shot doesn't have to win, it just has to get more likely for a while. The optionality is a good chunk of what you're buying.

Disclosure: we hold an open position on the Lions and Bengals, listed on our scorecard. Our Polymarket links carry a referral code. None of this is betting advice.