TSI Shadow Books
Hardened variants of the TSI-90 crypto momentum family, run as prospective paper-trading experiments alongside the live paper book. paper onlysimulated money
Prospective scoreboard (evidence accrues from 2026-11-01)
Equity since 2019 (log scale, post-cost paper) — prospective window starts 2026-11-01
Drawdown
Backtest metrics (disclosed; selection was in-sample)
| metric | TSI-100 (hardened) | TSI-100R (return-max) | tsilt90 (live paper book) |
|---|
Full disclosures
- This is a backtest plus prospective paper accrual — not a live track record. All fills are modeled at daily open with 7 bps/side costs, no slippage; a one-day fill lag historically costs ~45% of CAGR.
- TSI-100 missed its pre-declared KEEP gates 5/7 (Sharpe by 0.002 — a statistical tie; and the execution-lag gate, which the live book also misses). It is published, not promoted.
- TSI-100R was the best of a 6-trial grid chosen on the same dev window — its +8 pp/yr edge over the live book is inside selection noise until 90 prospective days judge it.
- Performance is regime-dependent: roughly 45% of the family's dev-window growth came from the unusually strong 2026 alt-market year. Calmer regimes ran ~70–100%/yr; drawdowns of 20–30% occur regularly; any random year carries a 6.5% (TSI-100) to 17% (TSI-100R) chance of a ≥30% drawdown.
- Guardrails embedded in the shadow specs: per-name weight cap 0.25 of NAV; gross halved whenever trailing-30-day funding cost exceeds 15%/yr; TSI-100 additionally scales 0.75 while Bitcoin's 15-day trend is negative. Thresholds were frozen before results and are fixed for the whole horizon.
- Retire lines at day 90 (fixed in advance): cumulative ≤ −25.8%, or drawdown deeper than −40%, or ≥10 funding-budget trigger days.
- No real money is or will be traded by these books without a separate, explicitly authorized decision.