NFL · Research roadmap
Research roadmap, not a current capability
NFL is the most heavily modelled betting market in existence, which makes it simultaneously the most attractive and the least likely place for a small experiment to find something. Nothing described here is running.
Current status
Nothing here is running
- Data collection
- None
- Model
- None
- Published record
- None
- Timeline
- Not committed
Roadmap means roadmap
VigPulse does not run NFL. There is no model, no data pipeline, no graded entry, and no committed date. The only sport currently covered is MLB, that coverage is moneyline only, and it is paper-only.
The record
What a published record would have to contain
| Field | MLB | NFL |
|---|---|---|
| Graded entries | 33 | 0 |
| Record | 9W-24L | None |
| Paper units | +2,417.5u | None |
| Pre-game CLV sample | 6 | 0 |
| Stake type | Paper only | Nothing staked |
MLB figures are read from the published ledger. NFL columns are empty because no football entry has ever been logged, not because the data is withheld.
Market structure
How NFL differs
A weekly cadence changes everything
Sixteen games a week against fifteen a day means roughly a thirtieth of the sample rate. A method that would take three months to evaluate in baseball would take several seasons in football, which is a fundamental obstacle for an experiment that is trying to learn quickly.
The spread is the market
Football pricing is organised around a moving point spread rather than a moneyline, so movement analysis has to track two dimensions at once: the number and the price attached to it. Half-point moves around key numbers matter disproportionately.
Key numbers dominate
Margins of victory cluster on 3 and 7, so the value of a half point is wildly non-linear. Any movement model that treats a spread as a continuous quantity will misprice exactly the moves that matter most.
Long information windows
A week of injury reports, practice participation, and weather forecasts means information arrives gradually rather than in a single scheduled release. That spreads repricing across days instead of concentrating it.
Method
What would have to change in the method
The sequence on the methodology page was written against a two-way moneyline. Below is what each step would have to become for a spread market. It is a description of unsolved work, written in the conditional on purpose.
Six steps, rewritten for a spread
not implementedMarket
A football market is a pair, not a price: a point spread and the price attached to it. Both move, sometimes in opposite directions, and a capture that stores only one of them throws away most of the information.
Consensus
De-vigging a two-sided price is the same arithmetic as in baseball, but books can sit on different numbers at the same moment. A consensus across books therefore has to reconcile spreads before it can average prices, which baseball never requires.
Movement
Information arrives across a week of practice reports and forecasts rather than in one scheduled lineup release. Repricing is spread thin, so the movement signal is lower amplitude and easier to mistake for drift.
Model
Value depends on the probability mass sitting exactly on 3 and 7, so a model has to produce a distribution of margins rather than a point estimate. A mean-only model is systematically wrong about the half-points that matter most.
Opportunity
The comparison is against cover probability at a specific number, not against a moneyline. Moving off the number to get a better price is a trade-off the model has to price explicitly.
Risk
Sixteen games a week gives almost no diversification. Correlated exposure across a single Sunday can be a large fraction of a season's variance, which changes staking before it changes modelling.
The sample-rate problem
StructuralBaseball produces roughly fifteen gradeable events a day. Football produces roughly sixteen a week. A method that would take a few months to evaluate on the MLB cadence would take several seasons here, and no amount of engineering compresses that.
For a project whose main constraint is evidence per unit of calendar time, that is the single strongest argument against starting with football.
The efficiency problem
StructuralThe NFL market is the most heavily modelled betting market that exists. Disagreement between books is smaller and shorter-lived than in thinner markets, so the raw material this kind of analysis works on is scarcer exactly where the most attention is aimed at it.
Difficulties
What makes football hard
Known problems
4 listed- The market is enormous and extremely efficient. Finding disagreement worth acting on is much harder than in a thinner market.
- Low event volume makes statistical evaluation of any method take years rather than months.
- Key-number structure requires modelling the distribution of margins, not just the mean.
- Public money is heavily concentrated, which produces line shading that can look like a signal.
What building this would require
3 listed- A spread-aware movement model that handles the number and the price together.
- Explicit handling of key numbers rather than treating the spread as continuous.
- Acceptance that evaluation would take multiple seasons, not weeks.
Readiness
Gates before NFL would be anything but a research note
| Gate | Status | Notes |
|---|---|---|
| Capture spread and price together, across books, over time | Not started | No pipeline exists. Nothing is being collected. |
| Model the distribution of margins rather than the mean | Not started | No model exists. This is the substantive research problem, not a configuration step. |
| Price half-points around key numbers explicitly | Not started | Not started. Treating the spread as continuous would misprice the moves that matter. |
| Accept a multi-season evaluation window before claiming anything | Not started | Not started. At roughly a thirtieth of baseball's event rate, there is no fast version of this. |
| Publish a graded record before describing NFL as a capability | Not started | Not started. No NFL entry has ever been logged or graded. |
Comparison
Where football sits against the rest
| Sport | Status | Cadence | Primary market | Liquidity |
|---|---|---|---|---|
| MLB | Current · paper | Daily, roughly 15 games | Moneyline | High on major markets |
| NFL | Roadmap | Weekly, roughly 16 games | Point spread | Highest of any US sport |
| NBA | Roadmap | Daily, roughly 8 games | Point spread and total | High |
| WNBA | Roadmap | Seasonal, several games per week | Point spread and total | Lower than NBA |
| Tennis | Roadmap | Near-continuous, year-round | Match winner | Varies enormously by event tier |