01 / HOW IT WORKS
Daily SWING reporting. This page presents end-of-day closing results for the experimental SWING (1D/3D) strategy generated by NIPPER AI_AGENT systems. The figures are research snapshots, not live trade confirmations.
Multi-agent selection. Eleven principal agents evaluate candidate stocks, supported by more than 40 auxiliary agents. Their outputs contribute to the selection process; scoring weights should not be interpreted as independently validated predictive accuracy.
Two HIT methods. CONF (MARKOV_2) and LIMIT independently assess qualifying price conditions. A stock enters the main tracking list when the configured HIT criteria are met; the signal time and level are recorded.
Simulation sizing. The agent may assign illustrative position values to test calculations. These are not actual orders, executions or holdings.
Review period. Results are usually assessed after the US regular trading session. Users can examine 1-day, 3-day or longer observation windows. Prior observations suggest some patterns become clearer over approximately 10 days, without guaranteeing future outcomes.
02 / SYSTEM ARCHITECTURE
Hybrid GAS / JavaScript design. The AI_AGENT systems run in Google Apps Script (GAS), combining JavaScript-based rule-driven processing with AI-assisted evaluation.
Research environment. NIPPER combines deterministic JavaScript / Google Apps Script workflows with AI-assisted evaluation to test multi-agent decisions against changing market data.
Processing scale. The architecture is designed around a target capacity of approximately 200,000 request-response operations per day. This is a design target, not a verified daily usage measurement.
Historical comparisons. Previous tables are retained for experimental comparisons. Displayed percentages depend on their calculation methodology and should not be treated as independently audited portfolio returns.
Data access. Some components and source details are not publicly disclosed for security or access-control reasons.
Percentage aggregation. Position sizes and individual share prices differ across stocks. Therefore, adding displayed stock-level percentages may not reproduce a portfolio-level percentage; any aggregate return requires a consistent position-weighted calculation.
