Skip to content

2025 - 09

Plan

Break-even Strategy Priority

  • Achieve Break-even Performance: Iterate strategy to reach break-even or better in paper trading
  • Feature Engineering: Add better features to improve model generalization
  • Model Development: Fine-tune existing models or explore new ML models for better performance metrics

Supporting Infrastructure (Only if Needed for Break-even)

  • Fix Performance Metrics: Resolve PnL discrepancies
  • dashboard clarity: Improve strategy details dashboard clarity
  • Auto integrity checks (#1): Automated state verification of data integrity
  • Strategy Monitoring (#5): Status & metrics dashboard and Alert via Logfire

Live Deployment

  • Deploy Live Strategy: Launch break-even strategy to production
  • Validate Live Performance: Monitor production performance and safety systems

Progress

Completed

  • ✅ Feature Engineering: Added better features for model generalization
  • ✅ Dashboard Clarity: Improved strategy details dashboard
  • ✅ Hyperliquid Gateway: Integrated first DEX for cross-venue arbitrage support

In Progress

  • 🔄 Model Development: Fine-tuning ML models for better performance
  • 🔄 Achieve Break-even Performance: Still iterating to reach target

Not Started

  • ❌ Fix Performance Metrics: PnL discrepancy resolution deferred
  • ❌ Auto Integrity Checks (#1): Automated state verification deferred
  • ❌ Strategy Monitoring (#5): Logfire dashboard and alerts deferred
  • ❌ Deploy Live Strategy: Blocked by break-even target
  • ❌ Validate Live Performance: Blocked by live deployment

Learning

Strategic Pivot from ML-Only to Diversification

After completing feature engineering with reasonable generalization, we shifted focus to Hyperliquid integration instead of continuing pure ML development. This pivot came from recognizing that betting everything on ML-based strategies is risky when model performance remains uncertain. Cross-venue arbitrage worked well before and provides a proven fallback strategy while we iterate on ML models.

Integration Reveals Hidden Complexity

Hyperliquid deployment uncovered many issues in strategy/gateway/account configuration that weren't visible in paper trading. Real exchange integration exposes edge cases in API handling, order management, and state synchronization even in simulations. Each issue required fixes across multiple services.

Database Management Pain Points

Database schema changes, migrations, and managing large storage proved more difficult than expected. The difficulty suggests we need better database tooling and migration processes before live trading scales up. This validates August's database separation work as critical infrastructure.

Break-Even Target Remains Elusive

Despite feature engineering and new gateway integration, we still haven't achieved break-even performance. This shows that profitable trading requires more than just good infrastructure - the actual trading logic and risk management need significant refinement. October needs sharp focus on strategy profitability, not just adding more features.

Q4 Window of Opportunity

We have limited time in Q4 to validate our new trading system generation. Missing this window delays proving our business model for another quarter. This creates urgency to focus ruthlessly on break-even performance rather than building more infrastructure.