Betting research, measured
Data window: 29 Jun – 2 Aug 2026 (4.27M odds observations)
· Papers published Aug 2026 · Page reviewed Aug 2026
Original studies and open data on the price and behaviour of betting markets.
Everything here is free to quote, chart and build on under
CC BY 4.0 with attribution.
Most of what circulates about sports betting is opinion. These pages are measurements:
millions of recorded odds turned into open, verifiable numbers — how much the bookmaker
charges, how honest the odds are, and how prices move. Journalists and researchers use them;
you're welcome to as well.
The studies
4.78%
The price of a bet, measured across 205,491 markets and 52 bookmakers — including
the first public measurement of player-prop margins (7.93%).
Read the study →
20.8%
Why sportsbooks keep twice the quoted margin: parlays compound leg margins — a
5-leg multi of ordinary legs has a built-in 20.8% hold. Home of the quarterly
Bet Better Margin Index.
Read the study →
±2 pts
18,238 settled markets: vig-free odds are calibrated within ~2 points everywhere —
but the margin is loaded onto the longshots.
Read the study →
Daily
What a bet actually costs, measured every day across head-to-head, handicap, totals
and player-prop markets. Open CSV, CC BY 4.0.
See today’s index →
2.76 pts
11.7 million snapshots: UFC lines move 2.76 points, MLB 0.60 — and price discovery
happens 1–7 days out, not in the final hours.
Read the study →
$31.5B
What Australians actually lose, per adult and per state — the official QGSO numbers,
explained plainly.
See the statistics →
2.1%
How much taking the best available price is worth — measured across books, the one
improvement every bettor controls.
Read the report →
Open datasets
The data behind the studies, published as citable, DOI-registered datasets. No odds, prices or
bookmaker identities are redistributed — aggregates only.
| Dataset | What's in it | DOI |
|---|
| Bookmaker Margin Panel | Daily measured margins: day × league × region × market type, plus player props |
10.5281/zenodo.21755380 |
| Market Calibration Dataset | 36,476 de-vigged probabilities vs real outcomes, five sports |
10.5281/zenodo.21764674 |
| Line Movement Dataset | Open-to-close movement and distance-to-close by sport, market and time bucket |
10.5281/zenodo.21765675 |
| AFL Player Match Statistics 2000–2023 | 206,722 player-games × 81 columns |
10.5281/zenodo.21612783 |
Working papers
Journal-format versions of the research, each with a citable DOI on Zenodo:
Also from the same programme: the free model API ·
our methodology.
Using this research
Cite any page or dataset with attribution to Bet Better and a link. If you're a journalist or
researcher and need a custom cut of the data, email
edward@betbetter.world — turnaround is usually
same-day.
18+. This research explains why the average
bet loses money — none of it is betting advice. If gambling is causing you harm:
Gambling Help Online or
1800 858 858.