Trading agent uses Jev to filter entries, backtest Sharpe rises from 0.64 to 1.68
A Kaspa spot strategy uses Jev to filter mechanical entry signals at about $0.04 for 2,600 evaluations, with the backtest showing roughly 4.5x final equity and drawdown narrowing from -63% to -23%.
i gave my trading agent a new toy: a judgement api called jev. llm-as-a-judge filtering mechanical entries on a kaspa:native spot strategy i hope to bring to ka$h's permissionless strategy marketplace someday. total cost for jev's evaluations: ~$0.04 (2,600 evals). backtest alone vs raw current strategy mechanics: ~4.5× final equity, ~2.6× sharpe (0.64 → 1.68), drawdown −63% → −23%, recent out-of-sample −20% → +315%, and it beats a matched random gate by a wide margin: so it's selecting trades, not just trading less. it's just single-asset backtest so live first... over :)
This post quotes another post: x.com/jpschroeder/status/2100230381588951209