Feature guides
Running Prove it
How to run Prove it on a result, what each attack does, and how to read its report, with a stored report as the example.
Run it
- Finish a test in the Strategy Agent, or open a saved one on My Strategies.
- Find the Prove it section on the result and press its button.
- The price shows before anything runs. Confirm to start.
- When it finishes, the report shows one line for each attack, and a top line that counts the ones it survived.
The attacks
| Attack | What it tries |
|---|---|
| Random-entry baseline | Buying and selling on random days, as long in the market. |
| Placebo (shuffled prices) | The rule on prices shuffled into a random order. |
| Signal delay / look-ahead check | Every trade a day late. |
| Triple trading costs | Three times the trading costs. |
| Parameter neighbourhood | Slightly different settings. |
| Walk-forward, fold by fold | Settings fitted on earlier years, then run on the years after them. |
| Regime and sub-period stability | Each part of the test on its own. |
A stored report
This is the report the engine produced for a golden cross on SPY, tested from 1 Oct 2021 to 24 Sep 2026. It made 328 test runs and the rule survived 2 of 6 attacks.
| Attack | Result |
|---|---|
| Random-entry baseline | Failed |
| Placebo (shuffled prices) | Failed |
| Signal delay / look-ahead check | Passed |
| Triple trading costs | Failed |
| Parameter neighbourhood | Passed |
| Regime and sub-period stability | Failed |
In the random-entry attack, the rule's Sharpe ratio was higher than in 72% of 200 runs that bought and sold on random days. The pass mark is 90%, so it did not pass: random days often did as well.
Reading it
- Passed: the result held up against that attack.
- Failed: the attack did as well, or the result fell apart under it.
- Could not tell: there was not enough to judge.
- Surviving every attack is still a statement about the past, not a promise.
Settings chosen on earlier years
One attack chooses the rule's settings on earlier years and runs them on the years after, fold by fold, then joins the folds together. It runs on the years before any sealed months.
Related
- Core conceptsProve itProve it runs a finished result through harder tests, to see whether luck alone could explain it.
- Use caseCheck whether a result is luckPut a finished result through harder tests, and keep its latest months sealed, so a lucky past is easier to spot.
- Learn pathIs it luck?At the end you can say why a test that looks good can still be wrong, and set a rule to be watched from today on.
- Core conceptsThe sealed testWhen you first test an idea, Quantis sets its latest months aside. You open them once, when you have finished changing the rule.
These pages describe how Quantis Trade works and what its tests show about the past. They are not forecasts and not advice.