Skip to content

Troubleshooting-Hub.md

Latest commit

 

History

History
83 lines (58 loc) · 3.75 KB

Troubleshooting-Hub.md

File metadata and controls

83 lines (58 loc) · 3.75 KB
Raw

Troubleshooting Hub

This hub centralizes common ta4j issues for production-minded users.

Fast triage matrix

Symptom Likely cause First place to check
Indicator output does not match external charting tool Warmup or initialization differences; data-length dependence FAQ, Technical Indicators
Strategy behaves differently in live vs backtest Execution assumptions differ (fill timing, partial fills, costs) Backtesting, Live Trading
Strategy keeps buying (or selling) repeatedly on one candle Live-candle evaluation without tradingRecord context or fill-sync/de-dup guards Live Candle vs Closed Candle Evaluation, Live Trading
Unrealized/open-position numbers look wrong Trade recording mismatch or lot policy misunderstanding Live Trading, Live Trading Runbook
Results change after enabling moving windows Historical bars evicted by setMaximumBarCount Bar Series and Bars
Noisy or unstable leaderboard across reruns Overfitting, weak validation geometry, or non-deterministic data pipeline Walk-Forward Research, Backtesting Realism Checklist
Charting fails in CI/headless environment GUI rendering unavailable Charting

Deep-dive troubleshooting paths

Data and series issues

  • Validate symbol, interval, and timezone alignment
  • Check for gaps, duplicated bars, and late bars
  • Confirm deterministic reconstruction of the same bar range

Read:

Indicator and strategy issues

  • Validate unstable/warmup bars
  • Confirm no accidental future-data usage in custom indicators/rules
  • Compare component values step by step before testing composed strategy behavior

Read:

Execution and trading-record issues

  • Distinguish order intent from confirmed fills
  • Decide explicitly between closed-candle and live-candle evaluation
  • Use shouldEnter(index, tradingRecord) / shouldExit(index, tradingRecord) in live loops
  • Verify ExecutionMatchPolicy behavior for partial exits
  • Reconcile local record state with broker/account state after failures

Read:

Evaluation and model-quality issues

  • Confirm metric set includes both return and risk
  • Audit train/test split and walk-forward design
  • Validate strategy ranking logic and weighting assumptions

Read:

Escalation checklist

Before opening an issue:

  1. Capture exact dataset source, symbol, and date range
  2. Capture strategy parameters and execution model
  3. Include reproducible command and expected vs actual behavior
  4. Include logs or chart snapshots that isolate the discrepancy

Escalation flow:

  1. Check Execution Decision Matrix for model-assumption mismatch
  2. Verify canonical run behavior with Examples Expected Outputs
  3. Apply operational recovery steps in Live Trading Runbook
  4. Open an issue through Found a Bug?