How biweek works
Current version biweek 1.0
What biweek does
Every trading day, our deep-learning models scan thousands of US stocks and rank the ones with the highest modeled probability of rising +10% within 14 trading days. No human stock-picking — the models decide from data.
Two engines — pick your style
Same exit rules (trailing stop), different stock-selection models.
Both engines only work as designed if you diversify across the whole list — buying just one or two picks breaks the math.
Why it can make money while missing half the calls
It's not about hit-rate — it's about asymmetry: let winners run, cut losers fast. Winners can go +10%, +40%, sometimes +100%; losers get cut near −7%. A few big winners can cover many small losses. That's the trailing-stop principle — and why diversification across all picks is mandatory, not optional.
How we validate
- Time-series deep-learning models trained on years of US market data.
- Walk-forward validation — tested only on future periods the models never saw, retrained per fold, three consecutive years above the S&P 500 in simulation.
- Continuous re-validation with new data; versions increment as models improve.
- All performance figures are simulations assuming a broker with automated stop-limit orders.
Operating rules (fully disclosed)
- Daily ranking by modeled probability — Steady 8 / Aggressive 10, equal weight.
- Exits: −7% hard stop / 12% trailing stop from the peak / 60-trading-day cap — whichever comes first. No fixed profit target.
- Only exited slots are replaced by the next-ranked stock (no mid-hold switching).
Disclaimer
This service is not investment advice. Displayed performance is a retrospective simulation (not a real account). You may lose money.
biweek is invite-only for now. Join the waitlist to get access when a spot opens.