Footgold

Thomas Augustus Grice

Footgold

Visit Footgold →
  • Footgold's dark green and cream dashboard showing upcoming matches, summary metrics, filters, model version, and the largest model-to-market edge.
    The dashboard turns 11 competitions into one scan-friendly decision queue.
  • Expanded Seattle versus Portland analysis comparing football probabilities with executable YES and NO market prices, fees, edges, and expected returns.
    Independent football probabilities sit beside executable YES/NO prices and fee-adjusted returns.
  • Footgold research board showing a saved match brief, rest differential, forecast baseline, evidence, and final-analysis status.
    Research, lineup questions, schedule context, and analysis status stay attached to the same frozen forecast.
  • Footgold model audit showing candidate performance, realized return, model comparison controls, experimental status, and a cautionary recommendation.
    Each major release carries an immutable audit, including failed or inconclusive experiments.
  • Footgold model guide explaining domestic form, cross-league translation, penalty modeling, and market-independent release gates.
    A visual explainer makes the model's inputs, fallbacks, and release gates inspectable.
  • Expanded Seattle versus Portland match card showing the proposed contract, executable price, model edge, expected return, and recent form.
    A compact fixture expands into an explicit action state without hiding uncertainty or execution costs.

Footgold is an allowlisted football decision lab I built to make forecasting inspectable rather than oracular. It covers 11 competitions and combines point-in-time probabilities, executable Kalshi prices, structured research, a personal decision ledger, and model audits in one operational interface.

The central design constraint is separation. Forecasts use only information available before kickoff; market prices are compared only after the probability is frozen; and qualitative research can change the decision context without silently editing the football model.

I designed and built the product end to end with Next.js, React, TypeScript, Drizzle ORM, Neon Postgres, Clerk, Vercel, OpenAI, Sportmonks, American Soccer Analysis, and Kalshi.

The main interface challenge was turning uncertain evidence into explicit states—action, pass, wait for a lineup, experimental, or fallback—while still exposing probabilities, both sides of a market, fees, expected return, data freshness, and model version.

Each major forecast release ships with an immutable historical audit. Failed or inconclusive experiments stay visible rather than being erased; the current audit explicitly warns when an apparent edge has not translated into a validated wagering signal.

Footgold is experimental decision-support software, not a promise of profitable betting. Its value is the evidence trail: a workflow for screening matches, recording decisions, and learning honestly from later results. The screenshots above use synthetic local data and contain no private account information.