The transparency page — written before there is anything to brag about.

GiottoO is in its founding phase. No live track record is published yet, and nothing on this page — or anywhere on this platform — should be read as a performance claim. Tracking begins the moment research briefs are published to members: each brief is timestamped at publication and followed to resolution. Nothing is added, removed, or edited after the fact.

Six things, updated as data accrues: (1) a probability calibration curve — forecast probability vs. resolved frequency; (2) the Brier score, a standard measure of probabilistic forecast accuracy where lower is better; (3) hit rate by probability bucket, so a 60% forecast is judged against 60%, not against perfection; (4) sample sizes for every bucket, because small samples prove nothing; (5) every win and every loss — the full resolved set, not a curated subset; and (6) the slippage assumptions used to mark entries and exits.

Entries are marked at the quoted mid-price at publication time with a stated slippage haircut, and exits follow the plan printed on the brief — profit target, stop, or time stop, whichever resolves first. Briefs that become untradeable (liquidity vanishes, market closes the structure) are logged as scratches and disclosed, not deleted. The counting rules will be published alongside the first data and versioned if they ever change.

A research engine that outputs probabilities should be judged on whether those probabilities were honest. Calibration answers that directly: across all briefs marked ~70%, did roughly 70% resolve favorably? A platform can post impressive-sounding win rates while its probabilities are systematically inflated. Calibration reporting makes that failure mode visible — which is exactly why we commit to it before having any numbers to show.

It will never show backfilled results, hypothetical returns presented as achieved, annualized extrapolations from small samples, or resolved briefs quietly removed. Past model behavior — once published here — will not be a promise about future behavior, and the page will say so next to every table.

The scoring weights, probability estimation, and Monte Carlo engine behind every brief are published in full on the methodology page. Before acting on any research, read the risk disclosure. If calibration data ever stops supporting the methodology, the methodology changes — and both pages change with it.