Backtesting Honestly

Building a simple backtester, then making it honest — realistic fills and costs, no look-ahead, and metrics that describe risk as well as return.

What you will be able to do

Build a vectorised and an event-driven backtest of the same rules, and reconcile them.
Model commissions, spread and slippage.
Remove look-ahead bias with correct signal timing.
Report expectancy, drawdown, Sharpe ratio and exposure from a trade list and an equity curve.

1
A vectorised backtest in pandas
Positions, returns and an equity curve computed from a signal column in a few lines.
Coming next
2
An event-driven backtest
Processing bars one at a time, the way a live system sees them.
Coming next
3
Fills, costs and slippage
Modelling the gap between the price that triggered a signal and the price you actually get.
Coming next
4
Timing and look-ahead bias
Signal on bar t, fill on bar t + 1 — and the other timing rules that keep a test honest.
Coming next
5
Reading a backtest report
Expectancy, drawdown, Sharpe ratio, exposure and trade count, read together rather than one at a time.
Coming next