Skip to content

2026 - 01

Goal: First profitable month in live trading. | Phase: 1 (Validate)

Strategy: Launch Profitable Strategy — 🔶

  • OKX account verified and live
  • Cross-venue arb on Binance + OKX deployed
  • Single-venue strategy deployed on any exchange
  • Positive P&L for the month

Cross-venue arb on Binance + OKX deployed for ETHUSDT and BTCUSDT. Single-venue strategies live: SMA Trend Following, SMA Mean Reversion. ML Classification strategy deployed to paper.

Paper Trading Results (as of 2026-02-01):

Strategy P&L Win Rate Sharpe Trades
CrossVenueArb ETHUSDT +$4.00 87.5% 1.19 8
SMATrendFollowing ETHUSDT +$0.33 55.6% 0.13 9
SMATrendFollowing BTCUSDT -$0.76 44.4% -0.25 9
SMAMeanReversion ETHUSDT -$1.27 0% -5.12 9
SMAMeanReversion BTCUSDT -$0.13 33.3% -0.16 3

Net paper P&L: +$2.17. Mixed results across strategies. Paper vs live can differ significantly — CrossVenueArb assumes optimistic fills that live may not deliver due to latency; arb strategies operate at order-book scale vs others at min order size; HFT execution complexity is higher in live. All strategies need to launch live for proper calibration, not just paper "winners."

ML: Produce Usable Signal — 🔶

QR delivers a model that generates actionable trading signals this month. Not research — a deployable model.

  • Deploy ML strategy to paper trading: real-time inference pipeline operational
  • Achieve >55% directional accuracy on validation and test set
  • Evaluate: signals profitable or not?

Classification strategy deployed to paper, real-time inference pipeline operational. Accuracy metrics on validation/test not yet measured. Strategy has 0 trades in paper — needs investigation (signal threshold? market conditions?). ML infrastructure works; signal quality assessment deferred — focus on trading strategies with proven paper results first.

Frontend: Alpha Release — ✅

Release dashboard to team at company domain. Enable iteration based on real-time live data.

  • Configure company domain for dashboard access
  • Create auth accounts for all team members
  • Distribute access and confirm everyone can view live data

Three environments deployed: test, paper, staging. Company domains configured (test/paper/stag.martianmobile.com). Auth accounts created for all team members, 1Password credential distribution ready. Team can now observe live trading and provide feedback.

What's NOT in January

  • ❌ Backtest validation (deferred until strategy profitable)
  • ❌ New exchange integrations beyond OKX
  • ❌ Team hiring
  • ❌ Infrastructure improvements (unless blocking profitability)
  • ❌ Complex ML architecture changes

Other Progress

  • Disaster Recovery: RDS Postgres portable backup/restore IaC complete — RDS snapshot limitation with Copilot worked around
  • Backtest Deployment: Paper mode with historical data access deployed; OOM limitation on long time ranges (needs incremental loading or memory optimization)
  • OKX Fee Negotiation: Institutional account approved with -1/2 bps maker/taker rate (effective 2/2)

Monthly Summary

Goal Target Result Status
Strategy Deployed Cross-venue + single-venue live All deployed to paper
Positive P&L At least one strategy profitable +$2.17 net paper P&L 🔶
ML Signal Deployable model with >55% accuracy Deployed, accuracy TBD 🔶
Alpha Release Team dashboard access Complete

Phase Assessment

Still in Phase 1 (Validate). Paper results positive but not yet validated in live.

Learning

Paper results ≠ live results. Mixed paper P&L (+$2.17 net) provides a starting point, but paper vs live can differ significantly. Arb strategies assume optimistic fills that live may not deliver due to latency. Different strategies operate at different scales (order book liquidity vs min order size). HFT execution complexity is higher in live. Simpler strategies may outperform in live despite worse paper results. Conclusion: launch all strategies to live for calibration. Use actual live results, not paper, to decide what to scale. Don't prematurely deprecate based on paper alone.

Fee negotiation matters. OKX fee reduction from standard rate to -1/2 bps maker/taker directly improves strategy profitability. For high-frequency strategies, fee structure can be the difference between profit and loss.

Strategy diversification > single winner. Goal should be multiple profitable strategies with different characteristics (arb, trend, mean reversion, ML). Diversification provides resilience to market regime changes, different capital requirements and risk profiles, and faster learning across strategy types.