I built ParlayGenIQ. It started as a personal modeling project — I'd been burned on too many gut-feeling bets and wanted to see if I could build something that beat the market with data instead of hunches — and over the past 9 months grew into the daily picks platform you're looking at now. I work full time as a Data Architect at Energy Recovery Inc.; ParlayGenIQ is a side project I build evenings and weekends, fueled by a long-running interest in AI and quantitative methods. It's been a lot of work, and I've enjoyed the grind.
The reason this exists in its current form is that I got tired of sports prediction services that hide their losers, change their criteria after the fact, and never explain why a model made the pick it made. Most of the industry is built on selection bias — show the wins, delete the losses, hope nobody checks. I built ParlayGenIQ as the version of a picks service I would actually trust myself: every pick published before games start, every result graded against reality, every model factor inspectable for the picks where the methodology is mature enough to show.
I think the right way to ship prediction products is one league at a time, with honest disclosure of what works and what's still being figured out. MLB has full factor explanations on every pick today. NHL, NBA, and WNBA explanations roll out as each model's calibration is validated. When I find I've been wrong about my own numbers — like the May 2026 MLB moneyline miscalibration I caught and corrected — I publish the correction on the Track Record page rather than quietly patching it. Trust is the product. The picks are just how I prove it.
Technically: daily predictions from XGBoost models trained on multi-season historical data. Per-pick factor attribution uses TreeSHAP. Calibration is per-league and per-bet-type. Routing decisions (BET vs. TRACK) are snapshotted at pick-generation time and preserved historically so old picks display under the rules they were routed under — not retroactively rewritten. All running on a sub-$100/month infrastructure footprint because this is a side project, not a venture-backed operation. Source data: ESPN, The Odds API, basketball-reference for WNBA stats.
If you found a bug, have a question, want to argue about a pick, or want to write about the project — please reach out. I'll respond.