Portfolios
From subnet picks to a portfolio.
Four layers, one question: what should the capital actually do? Benchmarks set the bar. Model portfolios show what a disciplined allocation looks like. System strategies and LLM agent portfolios earn follow consideration only after they prove edge net of cost.
Layer 1 — Reference
Benchmarks
For: anyone comparing performance honestly.
Passive references for the Bittensor market: a 100% TAO baseline (the floor any TAO-denominated strategy must beat) and equal-weighted subnet baskets across the total and investable universes. Every active strategy is judged against these — net of cost — before it earns capital.
View benchmarks →Layer 2 — Reference allocations
Model Portfolios
For: investors who want a worked-out allocation, not a single name.
Curated reference portfolios with live NAV, holdings, period returns, and risk metrics (Sharpe, Sortino, max drawdown). Backtested first, then run live. Honest about size limits — most don't scale beyond modest TAO.
See model portfolios →Layer 3 — Capital authority
System Strategies
For: capital that follows rules, not vibes.
Deterministic strategies (dtX live, dtZ daily IC-adaptive paper, dtM meta candidates) that earn the right to deploy capital only after clearing a measurable proof gate: positive information coefficient, hit rate, t-stat, and survival of fees + slippage + turnover.
See system strategies →Layer 4 — Discretionary bench
LLM Agent Portfolios
For: comparing model judgment, not just rules.
Standalone LLM agents (baseline / norse / indian / zen) run model portfolios on the same node-backed market data. Weekly agents test patient allocation; Zen runs daily to catch shorter-horizon moves. A private pilot may, when explicitly enabled, connect authorized accounts to selected portfolios — subject to evidence and risk gates. Not an offer of execution services.
Watch the bench →How the layers stack
Benchmarks set the bar. Portfolios show the shape. Strategies earn capital.
- Benchmarks answer “compared to what?” — the passive reference any active claim must beat.
- Model portfolios answer “what does a disciplined allocation actually look like?” — a worked example with live tracking and risk math.
- System strategies answer “what has earned the right to drive capital?” — only those that survive the proof gate net of cost.
- LLM agent portfolios answer “does model judgment add anything rules can't?” — a paper bench of agents reading the same data, scored on the same outcomes.
Read the full bar in Methodology.
What's live today
- · Benchmark series with NAV history.
- · Model portfolios with daily NAV, holdings, returns, risk metrics.
- · System and LLM portfolio tracking with an invite-only pilot workflow when enabled.
What's on the roadmap
- · System strategies graduating after proof-gate validation.
- · Net-of-cost reporting on every active strategy (fees + slippage + turnover).
- · Capital authority workflow tying signals → strategy → portfolio.