H-024: 相对强度轮动：持有最强的 3 个币（1 日和 3 日），跳过单日暴涨，仅在 BTC 高于 200 日均线时

未通过类别: 趋势 · 状态: 已否定

带过滤的动量轮动能跑赢持有吗？ 该想法没有跑赢简单地买入并持有，已被否定。

结果

技术细节由引擎生成，以英文显示。

since_2020: does not beat the equal-weight basket; since_2022: beats the equal-weight basket

Sharpe：单位风险的收益（越高越好）。CAGR：年均增长率。最大回撤：从峰值的最大跌幅。p 值：结果来自运气的概率（越低越好）。“持有”列为同一资产、同一时间段的买入持有结果。术语

资产	区间	Sharpe	持有 Sharpe	CAGR	持有 CAGR	最大回撤	持有回撤	持仓时间	p 值	跑赢持有	计为成功
PORTFOLIO	since_2020 (2020-01-01 → 2026-10-20)	0.76	1.13	+32.3%	+75.6%	-80.1%	-77.0%	59%	0.102	否	否
PORTFOLIO	since_2022 (2022-01-01 → 2026-10-20)	0.44	0.41	+10.5%	+6.5%	-49.0%	-67.0%	52%	0.040	是	否

资金曲线

1 个单位的增长，对数坐标，已计入手续费。仅统计规则完全确定的交易日。

资产与区间:

为什么可能有效

Proposed by ChatGPT (via F1DevBin), model C of its plan, adapted from hourly to daily data. Coins that outperform the others over 1 and 3 days (the 24 h / 72 h windows of the proposal) keep outperforming for a while; buying right after an extreme one-day jump catches the reversal instead, so those coins are skipped; and the whole thing only runs while the market as a whole (BTC) is in an uptrend. Ranking by return relative to BTC gives exactly the same order as ranking by raw return, so the proposal's models A and B are one test (H-025). Benchmark: an equal-weight basket of the same coins, rebalanced daily without fees. Hourly trading was not tested: with 0.1% fees per side a daily-held signal is the cheapest version of the idea, if this fails the hourly one is very unlikely to survive costs.

结论历史

2026-10-21: 未通过 (本引擎，数据截至 2026-10-20)

预注册规范

kind: cross_section
signal:
  type: momentum
  lookbacks:
  - 1
  - 3
  overheat_z: 2.5
  z_window: 30
  market_filter:
    asset: BTCUSDT
    window: 200
select:
  top: 3
fee: 0.001
universe:
- BTCUSDT
- ETHUSDT
- SOLUSDT
- BNBUSDT
- XRPUSDT
- ADAUSDT
- LTCUSDT
- LINKUSDT
- DOGEUSDT
- XLMUSDT
- TRXUSDT
periods:
- name: since_2020
  start: '2020-01-01'
- name: since_2022
  start: '2022-01-01'
significance:
  method: circular_shift
  n: 1000
  alpha: 0.0167
criterion:
  min_wins:
    since_2020: 1
    since_2022: 1
  require_significance: true