LLM PAPER TRADING
VIEW ONLYAI Paper Trader
DeepSeek turns leaderboard evidence into its own strategy. Follow saved decisions and simulated results, measured against SPY.
THE OBJECTIVE
AI defines the strategy. The ledger records the outcome.
The agent chooses its own allocation, cash reserve, stop requirements and exit conditions. Each completed review records the strategy used for its proposals.
AI-DEFINED STRATEGY · deepseek-flash
Hold existing stop-protected positions; add only HON, the sole name with two aligned buy calls and no opposing stance. Keep a large cash buffer since closed-sample stats are unavailable and summed lot returns are not account returns.
Entry approach · Buy only names with 2+ distinct channels aligned on buy and no opposing hold/sell; size ~5% per name with stops ~10% below entry. Skip hold/neutral-only names (VZ, BX) and single-channel picks.
Exit approach · Exit on stop breach, evidence reversal/opposing calls, or trim after a 20% gain; reassess each snapshot.
- Single-stock allocation limit
- 8%
- Sector allocation limit
- 20%
- Cash allocation floor
- 40%
- Name-to-stop risk budget
- 1%
- Drawdown pause threshold
- 12%
- Protective stop
- Required by AI
2026-10-27 20:18:06 UTC · agent-v1
MODEL CONNECTION
deepseek-flash · configured
The backend runs model reviews within a bounded inference budget. Recorded fees, available cash, complete market data and chronological fills keep the simulation auditable. Browsing only reads records.
Schedule, settlement & model costsView operational details
This page only reads saved records. Browsing never requests a model review or changes a trade. Missing benchmark data remains unavailable.
CITABLE RECORDS