Four ways to deploy. Pick the one that fits your stack.
Build on top
Raw predictions delivered via REST API. JSON endpoints for matches, predictions, ratings, and results. Use your own UI, your own brand, your own analytics layer. Bearer JWT auth.
Stream into your platform
Structured JSON prediction feed delivered directly into your platform. Built for operators, broadcasters, and aggregators. Sub-second latency, per-client rate limits, scoped endpoints.
Deploy as-is
Your brand, our engine. iOS, Android, Web, themed to your colours. Full audit trail included. We run the backend; you own the user relationship.
Long-term enterprise
Engine held in third-party escrow for business-continuity guarantees. Air-gapped deployment options available for enterprise data-residency requirements. Long-term enterprise licensing.
All routes include dedicated lifetime support, NDA package, methodology documentation and audit cooperation.
Performance & analytics teams. Pre-game opponent intelligence, ELO tracking, injury-adjusted ratings, historical audit history.
Embed audited predictions in match coverage. Drive engagement on your platform with our engine. Confidence-banded picks, head-to-head splits, methodology footnotes.
Independent risk benchmarks. Pre-game edge signals, audit-grade predictions, cross-reference our public verify page for credibility.
Multi-sport platform expansion underway. Same engine architecture. Same audit philosophy.
Six commitments from the founder
Run the backtest yourself. The SHA-256 hash never changes.
I published the SHA-256 hash before anyone asked me to. If the number was going to mean anything, it had to be verifiable by someone other than me. Run it yourself, the result will be identical.
67.58% is our floor. Two components disabled for the audit. The live engine runs higher.
I made the decision to disable two modules for the public audit. They're trained on historical data that overlaps the test window, so leaving them active would have been dishonest. The floor is real. The ceiling is higher.
We only publish the number we can prove.
There were higher numbers I could have published. I chose not to. Every figure on this site comes with a methodology you can inspect and a hash you can verify. If I can't prove it, it doesn't go on the page.
Not a script. A deployed, versioned, fully audited prediction platform.
I've seen a lot of 'AI tipsters' that are just a spreadsheet with a logo. This is a production API running on AWS Fargate, version-controlled on GitHub, with a six-stage audit trail. The difference matters.
Six independent checks. Zero future data. All pass.
The six checks weren't designed to pass, they were designed to catch me out. Chronological order, timestamp integrity, no data leakage, no duplicates. I ran them hoping they'd find nothing. They found nothing.
Above 60% every season for 16 years, including COVID, rule changes and expansion teams.
2020 was the hardest year to model: hub games, no crowds, condensed schedule. 2019 brought rule changes. GWS and Gold Coast joined mid-dataset. The engine stayed above 60% through all of it. That consistency is what I'm most proud of.
"Send one email. NDA within 24 hours. Full audit package within 48."
Request a partnership conversation