Forward test
One rule, fixed in advance, is tested only on trades that appeared after it was recorded. The rule and the stop criteria cannot be changed while the test is running. Details are in docs/FORWARD_TEST.md. This is a paper test: no real orders and no promise of a result.
Main rule: collecting, 0 of 100 closed trades until the first checkpoint.
Second rule: collecting, 0 of 100 closed trades until the first checkpoint.
Test start05.10 20:00UTC · setups confirmed since then: 2 (1h)
Main rule: closed trades0waiting for entry: 0 · in the market: 0
Average result per trade—Rtotal 0.00R after fees · share of targets —
To the checkpoint0 / 100stop threshold -12R Rules and results
| Rule | Closed | Awaiting entry | In the market | Share of targets | Average R | Total R | Max stop streak | Max drawdown |
|---|
| MAIN rule: 1h · crypto only · stop 1.5–2.5% of price · stop beyond the head · target 3R | 0 | 0 | 0 | — | — | 0.00 | 0 | 0.0R |
| SECOND rule: the main one + “coin form” (skip a coin if its last 2 closed trades at 3R were stops) | 0 | 0 | 0 | — | — | 0.00 | 0 | 0.0R |
| The same with a 2R target | 0 | 0 | 0 | — | — | 0.00 | 0 | 0.0R |
| Crypto, any stop size, 3R | 0 | 2 | 0 | — | — | 0.00 | 0 | 0.0R |
| Comparison: all coins, any stop, 2R (as in the Lab) | 0 | 2 | 0 | — | — | 0.00 | 0 | 0.0R |
Bold rows are the two fixed rules; they are tested separately by the same criteria. The others are shown for comparison and are not candidates: if “somewhere it is better”, that means nothing, they are not selected during the test.
Curve of the main rule
Fewer than two closed trades so far: the chart will appear when there are.
Forward test: one fixed rule on new data
A forward test checks a trading rule only on data that did not exist when the rule was written. Our main rule was frozen on 5 October 2026: which setups qualify, where the entry, stop and target are, and which fee is charged. From that date the rule runs unchanged, and every trade it would have taken is recorded.
Why not just backtest?
Historical tests are vulnerable to fitting: with enough filters, any past period can look profitable. Our own lab found a "candidate" that earned +21.9R on one window and lost −30.4R on the ten months before it. A forward test is the cure: no one can optimize on data that has not happened yet.
Stop criteria written in advance
Before the test started we wrote down when it will be considered failed or successful, and the minimum number of trades needed for a verdict. The page shows the rules, the running result curve and the latest trades. If the rule fails, we will publish that too. Until the sample is large enough, treat the curve as information, not as a promise.
What the curve shows
The curve is the running sum of results in R for the frozen rule, after fees. Flat or falling stretches are normal even for a working rule; what matters is the total after enough trades and whether it stays within the range expected from the history. Losing streaks are listed openly, because their length is what a trader has to survive in practice. The rule, the fee and the start date are printed on the page, so anyone can recompute the result.
Questions and answers
What does "frozen rule" mean?
The definition of setups, entries, stops, targets and fees cannot be changed during the test. Any change would start a new test.
How is this different from the journal?
The journal records all published ideas. The forward test follows one specific rule chosen in advance.
When will there be a verdict?
When the number of trades reaches the threshold written in the rules. Earlier results are shown but not interpreted as proof.
Can I follow the rule myself?
You can read every trade it takes, but the test is published as research, not as a recommendation to trade it.
What exactly is the main rule?
Quasimodo setups on 1h, crypto coins only, stop beyond the head with a width of 1.5–2.5% of the price, target 3R. The second fixed rule is the same plus a "coin form" filter: a coin is skipped while its last two closed trades were stops. Both were chosen on the year of history up to 5 October 2026 and are tested separately.
What happens if the rule fails?
The result stays published, the rule is marked as rejected, and any new idea starts its own test from zero. It is a paper test: no real orders are placed.