Built to Close the Gap Between Data Science and Editorial
Most sports media operations have a CMS, a calendar, and a team of writers. They do not have a data engineering team. Onyx Odds was built to change that balance.
The Problem We Watched Play Out in Newsrooms
Before Onyx Odds, Leul ran analytics at a sports media company. The editorial team was sharp and fast, but their pre-game analysis workflow was a spreadsheet patched together by one analyst. When that analyst left, coverage quality dropped in two weeks.
Owen had spent years building prediction models at a sports data firm. He had watched the same spreadsheet problem from the other side: teams that could afford a data engineer got good models; teams that could not, didn't.
The gap was not about intelligence or effort. It was infrastructure. We founded Onyx Odds in 2025 to build the piece that was missing: an API that delivers what a data team would produce, without requiring a data team to operate it.
The gap was not about intelligence or effort. It was infrastructure.
The People Behind the API
Spent several years working on analytics infrastructure at a sports media company, turning game log data into prediction signals for editorial coverage. Brought that experience in-house when the infrastructure gap he watched play out daily became the problem Onyx Odds is built to solve. Based in Las Vegas.
Built prediction models for a sports analytics firm serving broadcast clients. Developed the core probability distribution approach the Prediction Engine is built on. Formerly a senior data scientist at a sports data company.
How We Think About the Work
-
Calibration over confidence
A 68% probability score is only useful if it is right 68% of the time across a large sample. We measure calibration, not just accuracy, because a miscalibrated model erodes editorial trust faster than an uncertain one.
-
Infrastructure that disappears
Good API infrastructure is infrastructure that editorial teams forget about. It shows up before the game window, returns consistent output, and does not require babysitting. We design for that, not for impressive dashboards.
-
Honest output framing
We include confidence intervals and sample sizes in every output. Talking points carry a confidence score. Low-confidence outputs are flagged for editorial review. We do not hide uncertainty behind round numbers.
-
Built for small teams
National networks have data budgets. Independent publishers, regional outlets, and community platforms do not. We price and design for the teams that need this infrastructure the most, not just the ones that could afford to build it themselves.
Talk to the Team
Questions about the API, a potential integration, or just want to see the data before committing? We are happy to do a 30-minute call.