What we tested, including what failed
Most of what we test fails. The list so far, with the numbers: the structure model, the dealing range, sweeps, five timing rules and every way we tried to cut The Close's drawdown.
Most of what we test fails. Here is the list so far, with the numbers, because a record that only shows the winners isn't a record.
What the evidence says
| What we tested | What happened |
|---|---|
| A structure model that enters on the change of character | Tested out of sample on 15 markets, it averaged about zero per trade. |
| Taking entries only from the discount or premium half of the dealing range | No edge out of sample. |
| Fading sweeps mechanically (not our study: a 2026 walk-forward study of Asia-session sweeps on MNQ, Mesfin, preprint) | Lost about 2.2 points a trade after costs. |
| Five documented timing rules for owning an index, on SPY, QQQ, DIA and IWM from launch, after costs, registered before the data was fetched | One passed: The Close. Four didn't. |
| Ways to cut The Close's drawdown: a time stop, volatility sizing, a crash stop, early exits, and spreading the trade across four index micros | None adopted. The time stop was worse on all four ETFs. Volatility sizing helped on one of the four and the crash stop on two. Every early exit cut the bounce. Dips arrive together, so spreading the trade didn't cut the drawdown. |
Hypothetical results1, not actual trading. The full statement is at the end of this page.
What we do with a failure
We keep it. A rule that fails stays failed. We don't tune its settings until a backtest looks good, because that only fits the past. Settings are fixed before a test starts and never changed after it. A change is a new rule, and it needs a new test.
That is also why structure is how we place trades and never how we pick them. The words are defined in Structure in five words.
The Close has posted its levels before the open and its result after the close since October 2026. Nothing from those posts is claimed here; see the record. Those posts aren't in this list, which is about backtests.
Education only. Nothing here is advice about your account.