Strategy Research Lab
Explore the original backtests, corrected ৳1 lakh replays, trade fills and the neural study. Historical profit is not a live trading result.
Validity verdict
The 4,465 published strategy outputs reproduced exactly, but the old 98.4 lakh portfolio chart was hardcoded. The actual original portfolio replay lost 23.57% with ৳5 lakh initial cash and 12.23% with ৳10 lakh. Optimistic liquidity and same-day event ordering can inflate single-stock returns. The rankings below are retrospective, including their validation-period selection; they do not prove future profit.
All-stock buy signal rankings
See qualifying strategies across every stock, ranked by audited evidence with clear signal dates.
New experimentGP adaptive strategy search →
Compare search batches, profit against trade count, stress failures and every finalist’s trade ledger.
One-lakh replay
Original, conservative, 2022–23 validation and recent stress NAV are compared below for every visible stock.
51 rule families · four variants
125,800 training samples, eight epochs and 181,083 later-period trade rows. Exploratory research, with no verified live edge.
What changed in the audit
Next-session orders, stop-first ambiguity, 1% volume capacity, turnover gate, modeled costs and T+3 settlement.
Choose a DSE stock
399 visible audited symbols. Search for GP, BATBC or another listed stock.
What the model found for UNILEVERCL
Trained on 125,800 samples over 8 epochs. These are stored research outputs, not refreshed predictions.
Explore all 51 tested rule families
The later period was already exposed during earlier research, so it is not a pristine holdout. The model’s percentages do not establish a deployable strategy.
What the audit changed · UNILEVERCL
NAV from ৳100,000 per strategy · same scale for all bars
Unlock Systematic Strategy Signals & Executed Trade Ledger
Audited 1-lakh replay execution logs, trade fills, and systematic buy/sell signals require Tier 4 (Institutional AI - ৳3,999/mo).