Skip to content

2026 Plan

Objective

Build the proprietary trading platform that produces repeatable profit at controlled risk through a diversified portfolio of automated strategies, validated on live data with rapid iteration on top of solid infra and risk core.

Operating Model

The design we're building toward, applies equally to every phase:

  • Diversified portfolio of strategy philosophies (mean reversion, trend following, market making, statistical arbitrage, ML-driven). No single bet.
  • Solid infra + risk core. Multi-venue execution, state reconciliation, lifecycle management, automated kill switches, hardened verification. The moat.
  • Rapid iteration on live data. Each variant is small and fast to deploy; cross-variant convergence is the signal.
  • Capital deployed proportional to proven edge. Minimum quantity until edge is proven; ramped during stabilization on internal funds; external capital only after sustained track record.
  • Conservative venue + symbol surface. Expand only after the current set is fully validated. Each new venue/symbol adds attack surface and cognitive load.

Success Definition

Target: Phase 2 entered by year-end. Realistic stretch: Phase 2 completed.

Phase 1 exit (single clean month with 2 convergent strategies at min quantity) is the binding milestone — without it the rest doesn't unlock. Phase 2 completion (3 sustained months + capital ramped via internal funds) is the realistic upper bound for the year. Phase 3 (external capital or SaaS API) is aspirational and tied to whether Phase 2 actually closes mid-year.

If end-of-year we're still in Phase 1, the year wasn't a success and the model needs structural re-evaluation.

Phase-Based Approach

Phases are outcome-triggered, not calendar-bound. Move forward only when the criteria are actually met.

Phase 1: Validate

Find 2 strategies whose edge survives variant comparison. Build the platform that lets the right answer surface.

Phase 1 setup (the constraints that define this phase):

  • Minimum order quantity per strategy — iteration cost intentionally small
  • 2 symbols at high/mid frequency (ETH, BTC)
  • Multiple strategy types iterating concurrently across the portfolio

Criteria to Exit:

  • 2 converged + profitable strategies of distinct types, demonstrated in a single clean month at min order quantity. Convergence = multiple variants showing similar performance metrics (not one lucky variant). Profitable = positive P&L and positive profit margin. Clean = no incident-luck asterisks.
  • Backtest accurate enough for parameter selection — measured against known live results, with a defined validation gate before deploying to paper/live.
  • Risk core sufficient to survive incidents — verification rules covering vector consistency, lifecycle invariants enforced (close-on-disable), operator runbook for non-MJ on-call.

Activities:

  • Strategy iteration across the portfolio (MR, TF, MM, StatsArb, ML-driven)
  • Backtest infrastructure to make iteration cheap and trustworthy
  • ML research — the most promising new-edge source
  • Risk hardening continues reactively from incidents and proactively from monitor gaps
  • Operator-facing documentation

What's intentional, not deviation:

  • Heavy infra investment is the moat. Not "wasted on plumbing."
  • ML research running in parallel is correct, not "premature."
  • Multiple strategy types iterating concurrently is the model. Sequential single-strategy iteration was the wrong framing.

Phase 2: Stabilize

Hold the convergence steady. Ramp order quantity using internal funds.

Criteria to Exit:

  • 2 converged + profitable strategies sustained for 3 consecutive months
  • Order quantity ramped from min to material (5-10x) using internal capital only — no external funding required
  • No P1 incidents in the quarter
  • Defined drawdown rules that preserve capital during losing streaks

Activities:

  • Gradual order-quantity ramp per strategy, validated against convergence metrics at each step
  • Add a 3rd strategy type if a candidate converges
  • Operator runbook in active use; second operator (teammate or contractor) trained
  • AI-agent driven iteration: parameter sweeps, variant comparison, surface the convergent ones

Phase 3: Scale

Grow beyond internal capital. Pick a revenue mechanism: external funding or SaaS API.

By Phase 3 the platform has a track record (Phase 2 demonstrated 3+ months of sustained convergence with capital ramped). To grow further, internal capital is the binding constraint. Two distinct paths:

  • External funding: raise allocator capital or fund formation against the demonstrated track record
  • SaaS API: monetize the platform itself — sell strategy signals, execution infra, or risk-management primitives to other firms. Same moat (multi-venue execution + risk core), different customer

These are not exclusive but each requires meaningful focus to execute. Phase 3 entry forces the choice.

Criteria to Enter:

  • Phase 2 complete
  • Predictable monthly P&L distribution per strategy
  • Operational maturity: second operator real, runbook tested in production incident

Activities (path-dependent on revenue mechanism choice):

  • Capital scaling beyond internal funds (if external funding path)
  • Productize platform components for external consumption (if SaaS path)
  • Symbol expansion beyond ETH/BTC where strategies have signal
  • Production-grade venue expansion (Top 3 exchanges + 1 OTC)
  • Team expansion if revenue trajectory supports it

Guardrails

Capital preservation at min order quantity: Phase 1 capital at risk is bounded by min-qty design — typical monthly drawdown < $100. Hard kill the platform if monthly drawdown exceeds $500 (capital-preservation, not strategy-quality).

Incident response: P1 (cost > $100 or any unintended position > $1000) → immediate operator review and documented write-up within 24h. Reactive PRs ship before resuming the affected strategy class.

Execution discipline: Parallel iteration across the diversified portfolio is the model. Avoid:

  • Pivoting strategy types in response to single-variant noise
  • Treating operational success (volume, uptime) as evidence of strategy edge
  • Treating monthly P&L (single number) as evidence of strategy convergence
  • Letting documentation debt grow

Decision Checkpoints

Checkpoint Trigger Decision
Q1 End Still in Phase 1 Re-evaluate Phase 1 bar and method. Course-correct framing if needed.
Mid-Year Still in Phase 1 Re-evaluate strategy mix. Honest look at which types have any edge.
Q3 End Still in Phase 1 Evaluate whether the model is viable or needs structural rethinking.

What's Explicitly NOT Planned

  • Order-quantity ramping before Phase 2 — min order qty until Phase 1 exit (clean profitable month)
  • External capital / fund formation until Phase 3 — internal funds carry through Phase 2
  • SaaS API productization until Phase 3 — would split focus before the trading track record is real
  • Customer/partnership outreach beyond opportunistic until trading track record is real
  • Hiring until revenue trajectory supports it (Phase 3)
  • New venue integrations beyond Binance + OKX (production) until existing strategies validated
  • Strategy types beyond the current portfolio — fix the existing ones first

Tracking

Monthly review docs (plan-report/2026/2026-MM.md) track Plan vs Progress per month, Phase Assessment, and Learning. This annual plan stays stable; current-month status lives in monthly reviews. The plan is rewritten only when fundamental assumptions change.