Your own playbook

After this lesson you can take a trading idea from hunch to tested rule — the full loop, honestly run, which is what this school existed to teach.

From taste to test

Seven schools have handed you a complete method, and somewhere along the way you've probably started having ideas of your own. "Pullbacks to the 21-day seem to work better when the group broke out within the month," that kind of itch. This capstone course is the procedure for doing something honest with one, and it's the research loop our own desk runs, scaled down to one careful trader.

Step 1 — write the rule before touching data. Precisely enough that a stranger could execute it: universe, setup conditions, trigger, stop, management, regime filter, all of it in numbers rather than adjectives. This is pre-registration. A rule written after you've explored the data has already seen the answers, which is course 2's overfitting turned on yourself, and pre-registration is the thing that keeps it from happening. Date the document and keep it.

Step 2 — test it against history, honestly. Course 2's discipline is what makes a test worth reading: point-in-time universes, costs charged per School I's numbers, no lookahead, and the parameter-neighborhood check. Look at the results regime by regime rather than in aggregate (course 1). If an idea only worked in 2020–21, what you have is a description of 2020–21, and it stops working when that regime does.

Step 3 — the robustness interrogation. Move every parameter a step in each direction and see whether the result survives its neighbors or was a spike. Delete the top three trades and check whether it's still positive (course 4's outlier audit). Split the history in half and see whether both halves agree in sign. None of this is sophisticated, and it kills most fragile ideas in an afternoon, which is a cheap place for an idea to die.

Step 4 — forward it. A paper campaign under course 3's rules, written first with costs charged and no mulligans, then live pilots through the promotion gates. The idea earns size the same way you did, up the ladder and on evidence.

Step 5 — one change per loop. School V's review rule, now at playbook scale. Each pass changes a single element, so causes stay attributable. Two changes at once means learning nothing from whatever happens next.

The playbook itself

The output of years of this loop is a short document listing the setups that survived, each with its written rule, its tested history, its live sample, its known failure regimes and its process-grade record. Most professionals' playbooks are shorter than beginners expect. That document is your trading business, and it compounds in a way the account doesn't: a drawdown can't take anything out of it, and every honest loop adds to it, including the loops that end in "no." A falsified idea is knowledge you've already paid for, and it doesn't bill you again.

That was the destination the whole curriculum was walking toward: someone who can come up with a hypothesis, test it without fooling themselves, and act on what comes back at a measured size. Markets happen to be where you're practicing it.

Check yourself

1. Why must the rule predate the data exploration? (A rule written after seeing the data has already fit its accidents — pre-registration is the only cure that works on yourself.)
2. An idea backtests at +0.35R/trade, but deleting its best two trades turns it negative, and it only worked 2020–2021. Verdict? (A fragile regime artifact — two outliers in one regime. Step 3 killed it for the price of an afternoon.)
3. Why does a falsified idea belong in the playbook records? (It's knowledge you've already paid for. Once it's in the record it never costs anything again, and an idea you forget you falsified comes back later to be re-tested at real size.)

The idea this lesson installs

Write the rule before you look at the data. Change one thing per loop, and let ideas die cheaply.

This completes School VIII. Next: School IX — Context, the macro literacy that frames everything else.