What is the v7 strategy?

v7 keeps the long-side machinery from v6.4 and adds short signals on SOL and BNB (ETH-short was net negative in backtests and is excluded from v7's short side).

Per-coin parameters were re-tuned via Optuna with TPE sampler (~2000 trials) then refined with coord-descent neighbor search. Backtest min score 310.41 vs v6.4's 167.68 — +142 points.

~800 closed trades over the 2020-2026 backtest window (off-set 0), distributed across 6 markets (3 long + 2 short). SOL short and BNB long carry ~80% of total profit in backtests.

Win rate ~58%. Average hold time across all trades roughly 30 hours; medians shorter.

Same TP/SL/max-hold structure as v6.4, but tighter per-coin TPs on the short side. Short SOL: TP +10%, SL -4%, max-hold 96 hours. Short BNB: TP +10%, SL -4%, max-hold 96 hours.

v7 also uses an extra 'close_opp' rule: if a position is open in one direction and a flip signal fires the opposite way after the min-hold window, close the existing position rather than wait.

v7 has higher backtest gross returns but also higher max drawdown periods. The longer historical sample (vs v6.4) gives more confidence, but past performance doesn't guarantee future results — backtests are walk-forward validated on splits we couldn't see during tuning, but live markets evolve.

Both v6.4 and v7 can run side-by-side on the same subaccount thanks to the conflictGuard cross-strategy market lock. Per-strategy toggles let you enable just one if you want.