08-14-2026, 01:54 PM
The Strategy Tester is powerful, but garbage-in → garbage-out. Before trusting any result:
1. Turn ON realistic settings – commission, slippage, and order sizing matching your live plan.
2. Watch for lookahead – the classic mistake: using bar t+1 data inside bar t logic. The equity curve looks amazing, then live trading disagrees.
3. Test out-of-sample – optimize on one period, then walk forward on the next. If it only works in-sample, it doesn't work.
4. Check trade count – under ~50 trades the statistics mean almost nothing.
A strategy worth deploying survives all four. Most don't – that's the point of backtesting before risking capital.
How do you split train/test in your own backtests?
1. Turn ON realistic settings – commission, slippage, and order sizing matching your live plan.
2. Watch for lookahead – the classic mistake: using bar t+1 data inside bar t logic. The equity curve looks amazing, then live trading disagrees.
3. Test out-of-sample – optimize on one period, then walk forward on the next. If it only works in-sample, it doesn't work.
4. Check trade count – under ~50 trades the statistics mean almost nothing.
A strategy worth deploying survives all four. Most don't – that's the point of backtesting before risking capital.
How do you split train/test in your own backtests?