A systematic autopsy of every mechanism by which a backtest flatters a strategy that will lose money in production. Covers vectorbt, backtrader, and Zipline-reloaded, the full failure catalogue from look-ahead bias to p-hacking, and the validation methods that survive contact with live markets.
Algorithmic-Trading
-
Backtesting Frameworks and the Ways Backtests Lie -
Decoding Market Indicators with Code: OBV and the Golden Cross from Scratch Technical indicators look like chart-reading mysticism until you implement them — then they reveal themselves as small, deterministic functions over an OHLCV time series. This post takes two classics apart in Python: On-Balance Volume (the cumulative sum that turns volume into a running confirmation signal) and the Golden Cross (a moving-average crossover you can detect exactly with a sign change and a diff). Vectorized pandas implementations, the math behind each, the look-ahead-bias trap that makes naive backtests lie, and an honest accounting of why a lagging indicator is a description of the past, not a prediction of the future.