Blog

Working notes on testing strategies with data, with the code included.

• August 11, 2026

  The lookahead bug that makes every backtest look brilliant

  Shifting a signal by one bar is the difference between a strategy that works and one that only works because it knew the future.

• August 4, 2026

  Free market data, and what it quietly costs you

  Free price data is good enough to learn on and wrong in specific ways. Know which ways before you trust a backtest built on it.

• August 17, 2026

  Costs turn a good backtest into a bad one

  A strategy that trades often can look excellent gross and be worthless net. Here is how to put commission, spread and slippage into a pandas backtest.

• August 14, 2026

  A trading bot is four parts, and only one of them is the strategy

  Data, signal, risk and execution. Most people write the signal, wire it straight to an order, and find out the hard way what the other three were for.