cilipep/ai-agent-with-tslab/multi-instrument-test

Multi-Instrument / Multi-Timeframe Testing

When to use

• User wants to test the same strategy on different instruments (e.g., DOGE vs SOL vs AVAX)
• User wants to compare the same strategy across timeframes (e.g., 15H vs 1H vs 4H)
• User wants to find the best instrument for an existing strategy

Workflow

Phase 1 — Discover instruments

GET /api/agent-manager/securities?dataSourceName={datasource}&query=PERP&limit=100

Filter to _PERP contracts only (exclude .FundingRate, .MarkPrice).

Phase 2 — Clone and configure each variant

For each target instrument/timeframe:

1. Copy the source script:

`` POST /api/scripts/{sourceName}/copy body: {"name":"{newName}"} ``

1. Set source mapping (if changing instrument):

``json POST /api/scripts/{newName}/ops body: {"ops":[{"op":"SetSourceMapping","blockId":"Src","dataSourceName":"{datasource}","securityId":"{instrumentId}"}]} ``

1. Set interval (if changing timeframe):

`` POST /api/scripts/{newName}/lab-options body: {"intervalBase":1,"interval":{minutes}} `` Common intervals: 15 (15H), 60 (1H), 240 (4H), 1440 (D).

Phase 3 — Lifecycle each variant

For each configured script, use ./run-lifecycle.ps1 "{scriptName}" or the manual chain:

POST /api/scripts/{name}/validate
POST /api/scripts/{name}/build
POST /api/scripts/{name}/load
POST /api/scripts/{name}/run
GET  /api/scripts/{name}/metrics-summary

If validate returns errors, fix before build. If run hangs, check status and logs.

Phase 4 — Compare results

Collect from each metrics-summary:

• netProfit, netProfitPct
• allTrades
• profitFactor
• maxDrawdownPct
• winTradesPct
• fixedRecoveryFactor

Present as a ranked comparison table sorted by netProfitPct or fixedRecoveryFactor.

Phase 5 — Cleanup (optional)

Delete test variants if user doesn't need them:

DELETE /api/scripts/{name}

Naming convention

Use {sourceName}_{suffix} for clones:

• Instruments: BB_RSI_Strategy_DOGE, BB_RSI_Strategy_SOL
• Timeframes: BB_RSI_Strategy_1H, BB_RSI_Strategy_4H

Pitfalls

• SetSourceMapping requires both dataSourceName and securityId
• lab-options POST expects intervalBase as integer (1=MINUTE), not string
• Some instruments have sparse data — check barsCount in run results
• A script with 1-5 trades is not statistically significant
• PowerShell curl.exe is preferred over Invoke-RestMethod for long timeouts