Why survivorship bias quietly breaks most stock backtests
If your backtest universe excludes the stocks that went to zero, your results are fiction. Here's how we keep delisted tickers in — and why our numbers hold up live.
Read article →Calibrated win-probability: making an ML model you can actually trust
A model that ranks well but lies about its confidence is dangerous. Here's the isotonic calibration that makes a win-probability mean what it says.
Read article →Reading the market regime before you trade
Bull/bear × calm/volatile. Why a setup that prints in a melt-up dies in chop — and how the engine switches strategy as conditions change.
Read article →Position sizing: the one thing most retail traders skip
Entries get all the attention; survivors obsess over sizing and max daily loss. The risk framework we bake into every pick.
Read article →SIP vs IEX: why the market-data feed you scan on matters
The cheap feed covers a small slice of trades. We run on the full SIP — 100% of US volume. Here's what that changes for a real scanner.
Read article →We deleted a strategy that looked great. Here's why.
Pullback-to-VWAP backtested beautifully — and failed out-of-sample. The case for cutting setups that don't survive honest testing.
Read article →Want these in your inbox?
We'll email new posts as they publish — no spam.