eagL / research
Research notes
Technical notes on the methods behind prediction-market quant work. No trend pieces and no forecasts for the year ahead; each note explains one mechanism well enough to use, whether or not you ever buy the software.
- Brier scores and calibration curves for prediction markets
How to grade a probability forecast: Brier score, log loss and ten-bin calibration curves, and why accuracy alone misleads on binary contracts. - Microprice and queue imbalance in binary contracts
The displayed price of a thin event contract is often not the tradable price. Microprice, depth imbalance, queue imbalance and liquidity elasticity measure the gap. - Why chronological holdouts matter in prediction-market backtests
Random train-test splits leak the future into the past. The mechanics of the leak, and the holdout regime that closes it. - The failure modes of prediction-market arbitrage scanners
Five ways an arbitrage scanner shows you money that is not there: screen prices, flat fee haircuts, mismatched contracts, single-leg thinking and survivorship in the alert log.
Where these come from
Each note documents a mechanism that runs inside eagL, written up in general form. The product-level specification, including the numbers and the limitations, is the methodology page.
Related reading
All of this ships inside eagL.
$499 once, for Apple Silicon Macs. No subscription.