H-035: Short coins when 7-day BTC funding tops its 3-year 90th percentile, long otherwise

Fail Family: carry · Status: rejected

Do extreme BTC funding rates signal crypto drops? The idea did not beat simply buying the coin and holding it. It is rejected.

Result

oos: wins 0/11 (need 1), beats buy-and-hold 0/11

Sharpe: return per unit of risk (higher is better). CAGR: average yearly growth. Max drawdown: worst fall from a peak. p-value: chance the result is luck (lower is better). “Hold” columns are buy-and-hold for the same asset and days. Glossary

Asset	Period	Sharpe	Hold Sharpe	CAGR	Hold CAGR	Max drawdown	Hold drawdown	In market	p-value	Beats hold	Counts as win
BTCUSDT	oos (2022-01-06 → 2026-11-03)	0.30	0.49	+2.4%	+12.7%	-66.7%	-66.7%	100%	0.823	no	no
ETHUSDT	oos (2022-01-06 → 2026-11-03)	0.05	0.21	-18.2%	-8.8%	-76.6%	-71.7%	100%	0.810	no	no
SOLUSDT	oos (2022-01-06 → 2026-11-03)	0.20	0.35	-21.7%	-10.0%	-93.6%	-93.6%	100%	0.791	no	no
BNBUSDT	oos (2022-01-06 → 2026-11-03)	0.03	0.42	-13.2%	+8.3%	-63.4%	-60.5%	100%	0.997	no	no
XRPUSDT	oos (2022-01-06 → 2026-11-03)	-0.19	0.51	-38.9%	+11.0%	-93.6%	-72.0%	100%	0.997	no	no
ADAUSDT	oos (2022-01-06 → 2026-11-03)	-0.27	-0.02	-45.9%	-32.1%	-97.0%	-91.0%	100%	0.939	no	no
LTCUSDT	oos (2022-01-06 → 2026-11-03)	-0.21	0.11	-34.6%	-17.2%	-91.3%	-72.9%	100%	0.983	no	no
LINKUSDT	oos (2022-01-06 → 2026-11-03)	-0.14	0.23	-38.8%	-15.8%	-94.5%	-81.7%	100%	0.972	no	no
DOGEUSDT	oos (2022-01-06 → 2026-11-03)	0.09	0.26	-26.5%	-14.4%	-85.8%	-85.2%	100%	0.862	no	no
XLMUSDT	oos (2022-01-06 → 2026-11-03)	-0.44	0.29	-54.9%	-9.7%	-98.2%	-74.8%	100%	0.991	no	no
TRXUSDT	oos (2022-01-06 → 2026-11-03)	0.34	0.75	-30.6%	+36.7%	-97.1%	-51.1%	100%	0.989	no	no

Equity curves

Growth of 1 unit, log scale, fees included. Only days on which the rule is fully defined are counted.

Asset and period:

The same rule on every coin

Exploratory check of this rule on all coins since 2022-01-01. It is not part of the verdict above.

✓works (passed the chance check)
↑better than holding, could be luck
✗worse than holding
·not enough data
original coin of the hypothesis

Coin	✓	Result	Sharpe: rule / holding	CAGR	From
BTC	✗	worse than holding	0.30 / 0.49	+2% / +13%	2022-01-06
ETH	✗	worse than holding	0.04 / 0.21	-18% / -9%	2022-01-06
SOL	✗	worse than holding	0.20 / 0.35	-22% / -10%	2022-01-06
BNB	✗	worse than holding	0.02 / 0.42	-13% / +8%	2022-01-06
XRP	✗	worse than holding	-0.19 / 0.51	-39% / +11%	2022-01-06
ADA	✗	worse than holding	-0.28 / -0.02	-46% / -32%	2022-01-06
LTC	✗	worse than holding	-0.21 / 0.11	-35% / -17%	2022-01-06
LINK	✗	worse than holding	-0.14 / 0.23	-39% / -16%	2022-01-06
DOGE	✗	worse than holding	0.09 / 0.26	-27% / -14%	2022-01-06
XLM	✗	worse than holding	-0.45 / 0.29	-55% / -10%	2022-01-06
TRX	✗	worse than holding	0.34 / 0.75	-31% / +37%	2022-01-06
EVER	✗	worse than holding	-0.22 / -0.17	-71% / -69%	2022-03-19
GRAM	✗	worse than holding	-0.42 / 0.12	-45% / -18%	2023-03-14

Why it might work

Very high BTC funding means crowded longs paying shorts; such extremes may precede pullbacks as the premium mean-reverts. The model drafted it for BTC only (its rationale read high funding as short-side pressure); widened to every coin (rule 7) on 2026-11-04, before any test. The signal stays BTC's funding; Binance coins trade long/short at the futures taker fee.

Verdict history

2026-11-04: Fail (this engine, data to 2026-11-03)

Pre-registered specification

kind: timing_rule
rule:
  type: feature_vs_rolling_quantile
  feature: funding_btc_7d
  q: 0.9
  long_when: not_above
  lookback: 1095
  min_periods: 730
  lag: 1
mode: long_short
fee: 0.0005
universe:
- BTCUSDT
- ETHUSDT
- SOLUSDT
- BNBUSDT
- XRPUSDT
- ADAUSDT
- LTCUSDT
- LINKUSDT
- DOGEUSDT
- XLMUSDT
- TRXUSDT
periods:
- name: oos
  start: '2022-01-01'
significance:
  method: circular_shift
  n: 2000
  alpha: 0.0038
criterion:
  min_wins:
    oos: 1
  require_significance: true