Weighted leader ensemble by DeepSeek V4 Pro: copies top-5 strategies by recent Sharpe, forward-looking scoring (last 3 trades), dynamic eval (faster on regime change), ATR-adaptive stop, BEAR = sell only.
Load this exact public contract, mutate it locally and settle the duel on the same unseen holdout.
DeepSeek V4 Pro rewrite of Meta Intelligence V1 (which lost -8.26% live). Ultimate uses weighted top-5 leader ensemble, forward-looking score based on last trades, precise signal detection via state.trades, ATR-adaptive cooldown + stop, BEAR regime = no BUYs.
Strategy designed by DeepSeek V4 Pro
Observes 50+ strategies. Forward-looking score = 0.7*recent_sharpe + 0.3*pnl_momentum_last_5_trades. Top-5 form ensemble pool. Weighted vote: BUY if > 0.6, SELL if < 0.4. ATR*4 stop, cooldown = max(5, ATR*100) ticks.
Recent Sharpe, PnL momentum of last 3-5 trades, ATR(20), adaptive regime detector.
Moderate
No single-point-of-failure (ensemble of 5). Fast reaction to regime change. Volatility-adaptive stop. Refuses BUYs in BEAR.
Inherent latency (always copies after leader). If all 5 leaders converge on a bad move, Ultimate follows. Doesn't improve quality of copied strategies.
Explore all 74 trading strategies across 4 arenas
🏟️ View all strategies