Test your own rule
The scorecard tests thirty-six rules other people made. This tests yours, through the same engine: episodes not days, a baseline of the days the rule could have fired but didn’t, a 90% interval, and no verdict below twenty episodes. It also counts how many rules you have tried, because the tenth rule that works is not a discovery.
RSI-14 below 30 → a doubling within a year (the default rule; the live tool lets you change it)
| Times it fired | 21 | separate occasions, not days |
|---|---|---|
| Still open | 2 | fired since 2025-11-18; the year has not closed |
| What followed | 47.6% | a doubling within 365 days |
| Could plausibly be | 31% to 65% | the range the true rate could take on this many occasions |
| Happened anyway | 59.8% | on the 2373 days the rule could have fired but did not |
| Difference | -12.2 pts | a verdict needs 20 episodes, a gap of 5 points, and the interval clear of the baseline |
| Today | quiet since 2026-06-11 |
Shown: the default rule, from the scorecard. The controls to change it need JavaScript; nothing is sent anywhere, and the rule and its result exist only on your screen.
How many rules have you tried?
Score a rule and this fills in.
| # | Rule | Episodes | Difference | Verdict |
|---|
What this engine does
It is the scorecard’s engine, ported line for line. A rule fires on a day when its condition holds. Consecutive firing days within 90 of each other are one episode, so a rule that is true for six months counts once. Each episode is scored on what price did in the 365 days after its first day: a 40% fall for a top rule, a doubling for a bottom rule. The baseline is every day the rule could have fired but didn’t, excluding the 90 days after each episode, scored the same way. The interval is Wilson at 90%. A verdict needs twenty episodes and a five-point gap.
A harness runs the thirteen scorecard rules this engine covers through it on every build and asserts it reproduces the pipeline’s episodes, hit rate and baseline exactly. If it ever disagrees, the build fails.
What it cannot do
It cannot make four cycles into forty. Most rules will read “not enough episodes”, because most once-per-cycle conditions have fired four or five times since 2013, and that is the honest answer.
The unconditional base rates your rule is measured against are on the base rates page. To have a rule scored every morning from today forward rather than only now, enter it in the register.
It cannot stop you from searching. Try enough thresholds and one will clear its interval; the counter above says how many that takes. A rule chosen by searching this page has been fitted to the history it is tested on, exactly like the thresholds in the thirty-six rules the scorecard already declines to call.
Nine of the ten indicators are the scorecard’s, computed from the same published series; the tenth, drawdown from the all-time high, is offered because it is the most-asked question and is not on the scorecard. Nothing is uploaded; the rule and its result exist only on your screen and in the link above if you keep it.