Do Polymarket’s spread and moneyline markets for the same game disagree — and which one moves first?
Comparing the spread and moneyline markets for the same NFL, NBA and NHL games: how far apart their implied win chances are before the game, whose odds move first, and which is closer when they disagree.
- Result status
- Preliminary results
- Last updated
- Start date
- Duration
- Day 1 of 1
- Research team
- Mysports.AI Research Team
- Plan registered
Spread vs Moneyline Odds: Key Findings
169 games from week 8 of 2025 to the Super Bowl; 31% of games differed by more than 3 points (range 24%–38%), 3% by more than 5. Conversion uses NFL margins from 2010–2024; other methods give 2.2–4.1 points. Retrospective, not blind.
- Consistent direction: the spread put the favorite about 1.5 points lower than the moneyline (e.g. 60% vs about 58.5%, odds 1.67 vs 1.71).
- No detectable leader: for the three leagues combined, the lead-lag asymmetry’s range includes zero; the two markets’ 5-minute changes barely move together.
- Spread above win price: never in the NFL; 0.08% of the time in the NBA (50 games), median 3 minutes, usually the spread price quickly fell back; one 4-minute case in the NHL.
In plain words
Why study this?
If the spread market often moved first, it would be an early signal for moneyline odds; if the two often disagreed, a model would gain a second price to compare. The results also show readers that the two markets on the same game rarely truly contradict each other.
What do we want to know?
Polymarket often lists two markets per game: the moneyline asks who will win; the spread asks whether the favorite will win by more than X points. Both are about the same game, so they should agree — and “win by more than X” is always also a “win”, so it should never cost more. How far apart are they, whose odds move first, and who is right when they disagree?
How do we study it?
We used every trade in the public Polymarket-v1 dataset for the 2025-26 seasons: 172 NFL, 1,048 NBA and 997 NHL games. We read a Polymarket price as the market’s implied win chance — $0.60 means about 60%, or decimal odds of about 1.67 (not the true chance). We turned spread prices into win chances using the history of NFL final margins from 2010–2024 and compared them with the moneyline. For “who moves first” we cut the 24 hours before each game into 5-minute bins, checked whether one market’s change in one bin predicts the other’s change in the next, and used simulations to remove the illusion created by the two markets trading at different moments. The method was fixed and timestamped before we looked at the data.
What have we found?
One hour before NFL kickoffs, the home team’s win chances implied by the two markets differed by a median of 2.3 percentage points, and by more than 3 points in 31% of games (e.g. 60% on the moneyline, odds about 1.67, versus about 57.7% or 62.3% from the spread, odds about 1.73 or 1.61). The gap had a consistent direction: the spread put the favorite about 1.5 points lower than the moneyline (slightly longer odds). We could not tell which market moves first — their 5-minute changes barely move together — nor which one is more accurate when they disagree. “Win by more than the spread” priced above “win” never happened in the NFL; in the NBA it happened 0.08% of the time and usually vanished within minutes.
What does this mean for predictions?
The two markets on the same game broadly agree, but not exactly: the spread rates the favorite slightly lower (about 1.5 points). We found no evidence that the spread tells you early how moneyline odds will move, or that either market is more accurate, so neither should be treated as the “smarter money”.
What can’t we say yet?
This is a retrospective study: the data are public, so it is not blind. We only have trade prices, not order books, so a “contradiction” does not mean both sides could actually be bought at once. Converting spreads to win chances relies on the history of final margins; other methods give gaps of 1.5–4 points. “Who moves first” needs a reference model to remove artefacts; the model we chose assumes the two markets move together, but in reality they barely do, so the NFL “spread first” verdict depends on the model and is not robust. NHL spread markets are too thin and are only described. The rest of the 2026 NFL season will be checked again with the same frozen rules.
Terms used on this page
- Moneyline
- A market on which team will win.
- Point spread
- A market on whether the favorite wins by more than a set margin (e.g. 6.5 points).
- Implied win chance
- Odds turned into a probability: odds of 1.67 is about 60%; percentage points are differences between two probabilities.
- Lead-lag
- Whether one market’s price change is followed by the other’s.
- Brier score
- The squared gap between a forecast and the outcome; lower is better.
Detailed results
Gap between spread- and moneyline-implied home win chances, NFL
Median gap in percentage points, by time before kickoff and conversion method.
Pre-registered description | retrospective, not blind | trade prices only | research, not betting advice.
HypothesesFor deeper reading
- H1When both markets trade, the spread-leads and moneyline-leads directions differ (three leagues combined, 5-minute bins, net of timing artefacts) → no detectable leader. League by league, the NFL met the pre-set rule for “spread moves first”, but that verdict depends on the reference model we chose in advance; with a model closer to the real data (a post-hoc check) there is no clear leader, so it is at most a weak hintInsufficient evidence
- H2In NFL games where the two disagree by more than 3 points, which is closer to the result (Brier score) → no detectable difference (52 games, +0.001, range −0.008 to +0.011); we said in advance this test can only detect large differencesInsufficient evidence
Spread vs Moneyline: Research Methods
Research questions
- One hour before kickoff, how many percentage points apart are the home team’s win chances implied by the spread and the moneyline? What share of games differ by more than 3 points?
- When both markets are trading, does the spread move first and the moneyline follow, or the other way round?
- When they disagree, which one is closer to the result? How often, and for how long, is “win by more than the spread” priced above “win”?
Method summary
- Sample: NFL 2025 week 8 to the Super Bowl, NBA and NHL 2025-26 regular seasons; only trades from 24 hours to 5 minutes before the start.
- When a game has several spread lines, we use the one with the most money traded; for the one-hour comparison the line is chosen using only earlier trades.
- Lead-lag uses only 5-minute bins where both markets trade in a row, removes timing artefacts by simulation, and passes a shuffled-pairs placebo check (near zero).
- “Illogical” = the spread price above the win price, checked minute by minute when both markets traded within the last 5 minutes.
Data and limitations
Data we use (all public and free)
- Polymarket trades: Polymarket-v1 dataset (Boka Qin, Rui Yang, Hugging Face, CC BY 4.0)
- NFL schedules, scores, spread lines and play-by-play: nflverse (CC BY 4.0)
- NBA start times: NBA.com; NHL start times: official NHL schedule
Limitations
- Retrospective, not blind; only one NFL season of about 170 games.
- Trade prices only, no order books: a “contradiction” does not mean both sides were buyable at once.
- The lead-lag conclusion depends on a reference model; the pre-chosen model did not fit the data, so the NFL verdict is not robust.
- Converting spreads to win chances depends on the history of margins; methods range from 1.5 to 4 points.
- NHL spread markets are too thin to test; NBA and NHL use scheduled rather than actual start times.
Latest progress
Latest：Main results: a median 2.3-point gap one hour before NFL games; no detectable leader or accuracy difference; illogical prices absent in the NFL and rare in the NBA. Plan and code were timestamped first; results externally reviewed
Next step：About 165 remaining 2026 NFL games are being collected; after the season we will re-check with the frozen rules
Study timeline
- Novelty and licensing checked; plan finalized and timestamped after four external code reviews
- Analysis code timestamped, then run; results reviewed twice externally
- Collecting the rest of the 2026 NFL season (about 165 games)In progress
- Re-check on the 2026 season with the frozen rules2027-03
Update log
Version history
- v1.0
Plan finalized and timestamped (T0 = 2026-10-29 02:39 UTC); retrospective, not blind, with the rest of the 2026 NFL season as a prospective test
Study record
a698b781c9618c090fe2e868ebe5395fa08d1ea09404845f13e75b753c1c2ec92fed1dcfff36dda10f87229561aa4681a9469c7a1b737bd5ad6a61039408c068c80b776ae61767de60d8c7822ff3afd5f3bfacda155a414d35e070837da36d0aThis page is a research log for reference only and is not betting advice or a guarantee of profit.