jgrusewski df5c591441 spec(crypto): funding harvest — clean OOS validated + deployable spec ...

Clean OOS test (filter chosen on 2019-24, applied blind to 2025-26): IS-best filter holds
OOS Sharpe +5.1; 9/12 strong-IS filters also OOS-positive; naive no-filter FAILS OOS (-5.8)
= validated, not curve-fit. Wrote deployable spec: delta-neutral long-spot/short-perp harvest,
tf30>5bp regime filter, equal-weight, daily rebalance, 10bp cost budget. Counterparty/exchange
tail (-100pct) is THE risk (Sharpe-blind) -> venue selection + collateral spreading + withdrawal
discipline + deleverage kill-switch. Realistic ~2-3.6 Sharpe market-neutral (after basis-vol
haircut). Phased rollout: paper-forward -> micro-live -> small-live -> scale, gate each.
The one validated edge past the 0.7 ceiling; price of admission is crypto + counterparty risk.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

2026-06-07 01:25:23 +02:00

3.9 KiB

Raw Permalink Blame History

View Raw