Research · 2026-05-05
Point-in-time data, or nothing
A backtest inherits every sin of its data. We prefer to inherit fewer.
Survivorship, restated OHLC, and spreads as they are now rather than as they were quoted: these are how polite backtests are born.
We insist on point-in-time series, venue-native funding and swap, spreads as they were shown. The claim is written down, dated and filed before a single parameter is fit, so the data cannot be asked to invent a reason after the fact.
If we cannot name who is on the other side of the trade and why they are systematically willing to pay, the idea is filed, not fitted.