Funding Harvest (Cash-and-Carry)
Status: Manual two‑leg strategy · Library entry #6 · Found 2026‑08‑08 Instruments: BTC, ETH (BNB passes only at the strict gate) · spot and USDT‑M perp Cadence: decisions at the 8h funding settlements · Direction: market‑neutral Deployment: episodic — in the market ~10–30% of the time, in cash otherwise
What this is — and what it is not
This entry contains no price prediction. You buy the asset on spot and short the same notional on the perpetual: price moves cancel, and what remains is the funding payment the crowded side pays you every 8 hours. When longs are paying (positive funding), the short perp leg collects it. It is how professional desks monetize crypto's sub‑daily positioning pressure — the thing actually underneath "low‑timeframe opportunity."
It exists in this library because six research sweeps established that no sub‑1h price‑shape strategy on candles clears the bar (research log). The profitable structure below daily bars is positioning, not patterns. Entries #1/#5 trade its directional echo; this entry collects the payments themselves.
The app cannot automate it yet — the strategy DSL is single‑leg. Execution is manual (any spot + futures account), and the how‑to is below. It is validated by a dedicated two‑leg simulator on real spot, perp, and funding data — not the builder engine — so this page has no interactive chart.
The rules (pinned before the run; nothing was tuned)
At each 8h settlement (00/08/16 UTC), using settled rates only: Enter (both legs, equal notional) when the mean of the last 3 rates > 0.01%/8h. Exit (both legs) when the mean of the last 3 rates < 0.003%/8h. Fully collateralized short (1× — no practical liquidation risk). Capital = 2× notional. Costs: spot 0.1%/side + perp 0.05%/side, both legs, in and out (0.3% per episode); basis P&L taken from real perp‑vs‑spot prices at entry/exit.
Validation evidence (returns on total CAPITAL, not notional)
| Asset · window | Episodes | Net | While‑deployed APR | Deployed | Max DD | Worst episode | Strip‑best? |
|---|---|---|---|---|---|---|---|
| BTC W1 2019‑09→2022‑08 | 7 | +9.5% | ~11%/yr | 29% | 13.4% | +0.1% | ✅ (+6.0%) |
| BTC W2 2022‑08→2026‑08 | 3 | +1.8% | ~6%/yr | 8% | 4.4% | +0.0% | ✅ (+0.7%) |
| ETH W1 | 4 | +12.1% | ~14%/yr | 30% | 18.1% | +1.4% | ✅ (+7.3%) |
| ETH W2 | 3 | +6.5% | ~9%/yr | 19% | 0.4% | +0.4% | ✅ (+1.9%) |
| BNB W1 | 21 | +6.5% | ~7%/yr | 9% | 2.3% | −0.2% | ✅ (+3.6%) |
| BNB W2 | 28 | +0.2% | ~0%/yr | 10% | 0.9% | −0.2% | ❌ (−0.5%) |
Pre‑registered neighbors (the only variants run): a looser 0.005% entry gate fails W2 on all three assets (weak funding churns the 0.3% fee hurdle); a stricter 0.02% gate passes all six cells including BNB. The pattern is monotone and mechanically expected: the fee hurdle is fixed, so the edge lives only in genuinely hot funding. Selectivity is the strategy.
The honest characterization — read this twice
- The payments are real and the episode risk is tiny. Worst episode across every run: −0.2% of capital. You are collecting transfers, not predicting anything.
- The headline APR is thin in quiet regimes. W2's 0.4–1.6%/yr on total capital is BELOW cash rates — because the trade got crowded after 2022 and you're idle 80–90% of the time. This is not a buy‑and‑forget yield product.
- The right mental model: cash sits in T‑bills; when funding runs hot, a few weeks of ~6–14% annualized harvest switch on, at near‑zero episode risk. In mania regimes (2020–2021, funding >40% annualized for stretches) it prints; in quiet ones it sleeps.
- Drawdown is basis marks, not losses: the W1 13–18% marks are mid‑episode perp‑vs‑spot dislocations (2021‑style). Episodes still closed positive — but you must be able to sit through marks without panic‑unwinding one leg.
Guardrails (read before trading)
- Both legs together, always. Enter and exit spot and perp in the same minute, equal notional. A single leg is a naked directional position — the opposite of this strategy.
- Fully collateralize the short (collateral ≈ notional). Thin collateral turns a market‑neutral trade into a liquidation lottery on wicks.
- Exchange/counterparty risk is the REAL tail risk — both legs typically sit on one venue; an FTX‑style failure loses the position, not a trade. Size accordingly.
- BTC/ETH first. BNB only clears at the strict 0.02% gate; alt perps' funding prints are noisier and their books thinner. Untested ≠ works.
- Taxes, spot‑withdrawal frictions and borrow quirks are not modeled; the 0.3% round trip assumes standard taker tiers (maker fills improve it further).
How it was found
Sweep #8, after the timeframe ladder closed: the pre‑registered angle change from predicting sub‑daily candles (dead, sweeps #1–#7) to collecting the sub‑daily transfers that make those candles hostile. Rules pinned before the run; both neighbors reported; BNB's W2 failure and the quiet‑regime thinness disclosed above rather than averaged away.