BACKTESTING
✕Manual post-hoc on a chart
→Purged k-fold cross-validation
// 01 · regimes as hidden states
// argmax P(states | observations)
Backtest like a quant. Not a guru.
// manifesto
If your strategy can't survive bootstrapping, it can't survive Monday.
// the toolkit, in motion
Three views of the same idea: that strategy quality is a distribution, not a number.
// 02 · forward distribution of equity
// 03 · parameter surface across regimes
// the wrong way vs the right way
BACKTESTING
✕Manual post-hoc on a chart
→Purged k-fold cross-validation
PERFORMANCE
✕89% win rate looks great
→Sharpe CI excludes zero
VALIDATION
✕Equity curve goes up and to the right
→Deflated Sharpe ratio