Honest backtest capstone
The end of Track C: build a rule, test it the honest way, and keep the report. Tune it on 16 years of history as often as you like. Then run it once on 8 locked years it has never seen.
The house rules
- No look-ahead, by construction: a signal reads candles up to its close, and the trade fills at the next open.
- Costs are always on. You choose how much, never zero.
- Every different rule you run on the in-sample years counts as a variant. So far: 0.
- Each rule sees the locked years once. Changing the costs does not reopen them.
Simulated data from the course’s market engine, not real prices. The same fixed history for every learner, about 24 years of daily candles.
Your rule
In plain words
- Buy at the next open when the 10-day average closes above the 50-day average after being below it.
- No trend filter.
- Stop 2 ATR(14) from the entry price.
- Target 2R.
- Exit at the close after 20 days if neither is hit.
The locked years
When you have a rule you believe in, run it once on the locked years. The report then shows both periods side by side and saves to your logbook and record.
Run this exact rule on the in-sample years first.
Before this, the Track C lessons on backtesting and overfitting explain every check used here.