Odds Data API for Sports Betting Models & Backtesting

Sports betting machine learning data API with retained events, no-vig fair odds labels, and results for training and backtesting models on AFL, NRL and more.

A betting model needs features (prices), labels (outcomes) and a benchmark (fair prices) in one place. SockOdds gives you all three for the Australian market: every book's price, results, and de-vigged fairOdds, on events that are retained for backtesting.

Why use SockOdds for betting models

Odds and outcomes in one schema. Records arrive pre-joined.
No-vig labels. fairOdds, fairSpread, fairOverUnder.
Normalised ids. One oddID across books, sports and leagues.
Retained events. Backtest without building an archive first.
Python and TypeScript SDKs. The SGO SDKs pointed here.

Building and backtesting a model with our data

Feature engineering

Per-book prices, consensus, fair price, book count, availability, minutes to kick-off.

Labels and ground truth

results when status.finalized.

Backtesting against the last price

Snapshot on a schedule to measure movement; the feed retains the event so the label is always there.

Example request

curl "https://api.sockodds.com/v2/events/?leagueID=AFL&finalized=true&limit=100" \
  -H "x-api-key: YOUR_API_KEY"

Finished events, prices and results, one page at a time:

{
  "eventID": "afl_2026-09-03_fremantle_vs_hawthorn",
  "leagueID": "AFL",
  "odds": {
    "points-all-game-eo-odd": {
      "oddID": "points-all-game-eo-odd",
      "opposingOddID": "points-all-game-eo-even",
      "marketName": "Odd/Even Total Points",
      "statID": "points",
      "statEntityID": "all",
      "periodID": "game",
      "betTypeID": "eo",
      "sideID": "odd",
      "fairOdds": null,
      "bookOdds": "-116",
      "byBookmaker": {
        "unibet": {
          "odds": "-115",
          "decimal": 1.87,
          "available": true
        },
        "tabtouch": {
          "odds": "-118",
          "decimal": 1.85,
          "available": true
        }
      }
    }
  }
}

SockOdds vs assembling a dataset yourself

Capability	SockOdds	assembling a dataset yourself
Odds + outcomes	Pre-joined	Two systems
No-vig fair odds	Built in	De-vig yourself
Line history	Retained events; snapshot yourself	Snapshot and store every poll
Normalised ids	One `oddID`	Reconcile per book
Cleaning effort	Minimal	High

Frequently asked questions

What data do I need to train a betting model?

Prices per book, a fair benchmark and outcomes — all on the event.

Does the API include historical odds?

Retained events from 2026-09; no open/close snapshots.

How do I get bet outcomes for labels?

results on finalized events.

What is closing line value and can I compute it?

The difference between your price and the last price before kick-off; snapshot to compute it.

Which SDK should I use?

Python via pip install sports-odds-api with the base URL pointed here.

Is there a free tier?

Yes — AFL and NRL.

Related use cases

Historical Odds Data API: Finished events are never pruned from the feed, so an odds archive of every Australian book starts the day you start pulling..
Odds API for Positive EV Betting: Positive EV betting API with no-vig fairOdds and every AU book's price side by side, so you can surface +EV bets without building your own devig pipeline..
Odds API for Bet Tracking & Settlement: A bet tracking and settlement API: grade bets from results and the started/ended/finalized flags.
Pricing: plans priced on request rate and scope, including a free Lite tier with no card.

Start building today

Free plan available. Set up in 5 minutes. Scale when you're ready.

Get your API key Explore the docs