The held-out year
The most recent year of data was locked away when you started. Now your rule sees it for the first time. If it only held up on the data it was built on, this is where you find out.
Flip it
Trade the exact opposite direction. If both versions make money, the result didn't come from your rule.
The random twin
Same stops, same targets, random entries. Your rule has to beat its random twin clearly. If it doesn't, your entry adds nothing.
The neighbors
Nudge every setting a little. A rule that holds up with a 20-point stop and falls apart at 22 was luck.
Triple the costs
Run it again with three times the slippage. Real trading is rougher than a simulation.
The verdict
You get one table and one word: keep, watch or discard. Most ideas end in discard. Write down why, and move to the next one.
Your research notebook
Every idea gets a page in your notebook: the rules, the tests, the verdict and one line on why. Discarded ideas stay in it for good.
Your assistant reads the notebook before it builds or proposes anything, so nothing gets tested twice.
Build sections 7 and 9 of the blueprint if they aren't built yet. Then run the full test protocol on my strategy: held-out year, flipped direction, random-entry twin, neighboring settings and triple slippage. Give me one table with pass or fail for each test, then a verdict: keep, watch or discard. Write one line on why, and file the rules, the table and the verdict in my research notebook.
- All five tests ran
- You have a verdict and it's in your notebook
- If it's discard: good. You just saved yourself real money.