15 Services from money-mind-extract.onrender.com

Corrects inflated t-statistics in backtests by clustering trades by date to account for non-independence, returning both naive and date-adjusted t-values.

Applies family-wise multiple testing correction to a backtest winner's t-statistic to determine if it survives statistical scrutiny after searching many strategies.

Runs three statistical validity gates (selection correction, deflated Sharpe, date-clustered t-test) on a return series to verdict whether it is a genuine edge or a lucky artifact of multiple testing.

Runs a single-gate selection correction (family-wise p-value) on a return series to check whether a backtest winner is real or just the luckiest of N tested strategies, for $0.01 per call.

Cross-checks a claim against 2-4 user-supplied source URLs and returns cited passages plus a verdict of consistent, conflicting, or insufficient_evidence based on source agreement.

Detects spurious correlation between two time series, distinguishing real statistical relationships from autocorrelation artifacts in trading research data.

Runs an OHLC-bar backtest and signal series through 11 rigorous statistical gates to determine whether a trading strategy result is genuine or a statistical artefact.

Analyzes whether a trading strategy's drawdown reflects a genuinely broken strategy or just a normal bad patch, using statistical comparison against shuffled returns.

Corrects A/B test results for multiple-comparison bias and determines whether variant B genuinely beat variant A in a statistically valid way.

Adjusts a backtest's Sharpe ratio downward to correct for selection bias introduced by testing multiple strategies or parameter sets.

Computes the minimum number of observations required for a trading strategy test to have statistical power before running a backtest.

Adjudicates whether a trading strategy's recent performance shift represents a genuine regime change or is within normal statistical variation.

Extracts web content from JavaScript-rendered or static pages using three independent cross-checked methods, returning confidence-scored results with tamper-evident attestation

Triages up to 1,000 backtests or trading strategies in a single call using selection-bias correction, returning each as killed (NO), survived (PROVISIONAL), or malformed (ERROR), ranked by family-wise p-value.

Runs a trading strategy or backtest through eleven validation gates to detect data snooping, inflated t-stats, price data fabrication, and duplicate edges across instruments, returning a verdict and diagnostic reading.