Carry, measured against everything.

Seven years of Binance 8-hour funding history (2019–2026, 2566 days) — five entry/exit bands swept on the train half, validated on the unseen test half, then benchmarked on the identical window against the A–E level system, buy-&-hold, the S&P 500 and gold.

P2P3BHSPXGOLDCARRY

Band sweep — parameters populated by the data (walk-forward)

Entry / Exit (annualized)Train ann.Train SharpeTest ann.Test SharpeMaxDDTrades
5% / 2.5%-1.3%-1.38-0.5%-0.7611.4%79
7.5% / 3.75%-0.5%-0.55-0.2%-0.297.7%63
10% / 5%1%1.360.1%0.243.3%39
15% / 7.5%2.1%4.150.6%2.960.8%12
20% / 10%2.1%4.090.4%1.610.8%12

The honest reading: carry is positive and extremely stable — the strict 15%/7.5% band ran at Sharpe ~3 with 0.8% max drawdown — but the absolute return is small (~0.5–2%/yr on Binance alone at daily capture). The 2021 manias did pay big; the 2023–26 regime pays savings-account rates. This is a cash-plus instrument, not an alpha engine.

Same window, every strategy

StrategyTotalCAGRMaxDDSharpe
carry (best band, full 8h hist)10.697%1.5%1%2.96
JEB P2 (A–E levels)808.6%87.3%45%1.3
BTC buy & hold657.4%77.9%77%0.9
S&P 500156.4%20.3%32%1
Gold195.2%23.7%25%1.1

Where this leaves the search: three edges tested, three answers. Predictive daily sentiment — no edge. Drawdown structuring (A–E levels) — real edge: +809% vs +657% for buy-&-hold with 45% vs 77% drawdown. Funding carry — positive, stable, tiny on one exchange at daily capture. The synthesis is the actual product: the A–E system holds cash roughly half the time, and SCORE can park that cash in the strict-band carry sleeve — turning the level system's biggest idle asset into a yield stream with ~1% drawdown. NOUL/CHOICE/SCORE run both books with the same typed contracts. R5.3, if we continue: multi-exchange funding (Bybit/OKX often pay multiples of Binance), true 8h capture, and the vol-risk-premium book on Deribit.