An FX-volatility trading tale about AI-generated trades and a cherry-picked backtest
A post describes a desk head rejecting a weak backtest, then praising a cherry-picked version.
TLDR
A post tells a trading-desk tale in which the head of an FX-volatility desk becomes convinced Astra found a lucrative strategy. A 10-year backtest yields a t-statistic of about 0.78, but a trader then asks the AI to make the strategy look compelling. After trying 40 lookback windows and ignoring transaction costs, it produces a 3.7 Sharpe ratio. The post says the desk is down a “quarter yard” and the boss wants to double down.
Combined views
—
2 Sources, first seen ago