Skip to main content

Next Steps for Mercury: A Structured Plan

· 2 min read
Max Kaido
Max Kaido
Architect

A structured plan for improving Mercury's profitability, automation, and infrastructure.

1. Immediate Focus: Enhancing Testing & Dynamic TP/SL Fixes​

1.1 Improve Market Comparison Tests​

  • Currently, tests only confirm that data is being generated but do not validate its accuracy.
  • Next steps:
    • Add verification logic to ensure market comparisons are meaningful.
    • Implement consistency checks to identify weak sorting decisions.

1.2 Debug and Validate Dynamic TP/SL​

  • Dynamic TP/SL has been implemented but likely contains a bug.
  • Next steps:
    • Develop targeted unit tests to check the logic against historical data.
    • Compare expected vs. actual TP/SL behavior in different market conditions.

2. Mid-Term: Enhancing MAAT for Process Oversight​

2.1 Implement MAAT Sampling​

  • MAAT is crucial for identifying systemic errors in market selection and TP/SL strategies.
  • Next steps:
    • Ensure MAAT samples TA data, pairwise market comparisons, and TP/SL decisions.
    • Generate a daily gist with randomized selections for human review.
    • Use O1 AI for primary analysis, flagging anomalies for deeper inspection.

3. Mid-Term: Automating Tournament Operations​

3.1 Atlas Module Enhancement​

  • The Atlas module currently lacks automation for launching tournaments and shadow portfolios.
  • Next steps:
    • Implement automatic tournament scheduling.
    • Enable auto-logging of results for further analysis.
    • Integrate with MAAT to feed data into the monitoring pipeline.

3.2 Grafana Integration​

  • Without a proper frontend, data analysis is constrained to TG/Gist, making deep insights difficult.
  • Next steps:
    • Connect tournament results to Grafana dashboards.
    • Enable real-time monitoring of win rates, TP/SL success, and tournament efficiency.

4. Long-Term: Improving Market Selection and Heuristic Filtering​

4.1 Optimize Market Sorting​

  • Currently, pairwise comparisons are applied across the entire market, making sorting slow and inefficient.
  • Next steps:
    • Introduce heuristic filtering to eliminate weak markets early.
    • Improve sample selection for MAAT to detect contradictory or misleading rankings.
    • Experiment with alternative sorting methods to increase profitability.

4.2 Integrate Regime Detection​

  • At some point, regime shifts in market conditions should be accounted for.
  • This would allow dynamic adaptation of tournament strategies to current market phases.

5. Long-Term: Choosing a Frontend Framework for Analysis​

5.1 Current MDX Limitations​

  • MDX provides interactivity but requires custom solutions for each new feature.
  • The lack of standardization slows down development.

5.2 When to Move Beyond MDX​

  • If most interactive elements become React components, Next.js + MDX is a better fit.
  • If API interactions and live data analysis are required, Next.js + TRPC is a stronger choice.
  • If UI complexity increases, ShadCN could provide a structured design system.
  • Short-term: Continue using MDX for basic analysis.
  • Medium-term: Prototype key views in Next.js before fully migrating.
  • Long-term: Move to a full frontend framework when MDX limits development speed.

Conclusion​

  • Immediate priority: Improve testing and debugging TP/SL logic.
  • Mid-term: Automate tournaments (Atlas) and improve MAAT monitoring.
  • Long-term: Develop a structured frontend for analysis and implement regime detection.

This structured approach ensures that each step contributes to Mercury’s profitability, with a clear roadmap for automation and UI improvements.