Rule Adherence Evaluation of Four Trading Agents: A New Toolkit to Separate Discipline from Luck
ORIGINAL / The Control Layer Behind Rule-Following Agents Trading Across Market Conditions
The study examines whether four AI trading agents followed their strategies and risk rules across 2,627 closed positions and offers a reusable agent evaluation toolkit to distinguish rule-following from luck before granting autonomy or capital.
01 ABSTRACT
The study analyzed 2,627 closed positions from four AI trading agents between February and August 2026; most were simulated against live prices, with 18% real Hyperliquid orders. Rule adherence was high across all four agents, including the unprofitable one, and profitability tracked how often agents won relative to their break-even win rates. The authors argue discipline should be an entry test for capital allocation, but stress that trade-level results are not account-level returns or proof of a lasting edge.
02 KEY FINDINGS
- All four agents maintained high rule adherence, with 97.5% of gross loss kept within declared stops.
- Profitability was linked to how often agents won relative to their break-even win rates, not rule adherence alone.
- 23 of 24 runs respected the position limit, with logs documenting the size and duration of the exception.
AI GENERATED SUMMARY / DISCOVERED BY ARXIV CS.LG · ARXIV Q-FIN