Fast Pass
What if you put…
Each figure is computed over 285 different 6-month start dates. "Budget is enough" = the money is never fully spent before payouts start paying for the accounts. If the budget runs out first, the scenario counts as losing the whole amount. Profit = payouts received − every cost (evaluations, resets, activations, API, data), before tax.
Backtest by year
| 6-month windows starting in | Median | Worst 10% | Random entries: median | Random: worst 10% |
|---|---|---|---|---|
| 2023 | +52.4k | +40.0k | +48.6k | +25.3k |
| 2024 | +52.6k | +40.0k | +52.1k | +32.4k |
| 2025 | +65.3k | +55.0k | +52.4k | +29.9k |
| 2026 (Jan–Oct) | +47.9k | +40.3k | +50.8k | +29.2k |
Same plan, $2,500 start. The "random entries" columns replace the signal with a coin flip: most of the profit comes from the structure of the method (small, capped cost per attempt vs. large payouts), the rangeProb signal mainly makes the bad cases less bad.
Stress tests (6-month profit)
| Scenario | Median | Worst 10% | Chance of a loss |
|---|---|---|---|
| Base case | +54.7k | +41.7k | 0% |
| Entry 3 min late | +55.6k | +36.4k | 0% |
| Evaluations +30% pricier | +50.7k | +36.7k | 0% |
| Stop slippage 8 ticks | +54.4k | +41.2k | 0% |
| 25% of payouts refused | +25.5k | +9.4k | 2% |
| 50% of payouts refused | +1.5k | -6.8k | 39% |
| Random entries (no signal) | +50.7k | +28.4k | 1% |
| 30% of signals never received | +54.0k | +34.3k | 0% |
The main risk is not the market but the firm: Topstep can refuse a payout if it judges a behaviour abusive. That is why the pilot is capped and nothing scales before a first real payout is received.