源文件:research/quant_digests/2026-03-26_0850_cbeth-eth-rolling-fair-basis-mr.md
15m/5m 最小快检long cheap LSD / short ETH 或反向,赌 rolling fair-basis 回归先回答 base alpha:不是“LSD 基本面解释”本身,而是 CBETH-ETH 这类 liquid staking basis 围绕慢变公平值的短周期回归。
主论文是 Scharnowski, Jahanshahloo (2024), _The Economics of Liquid Staking Derivatives: Basis Determinants and Price Discovery_, Journal of Futures Markets。论文本身更偏 basis determinants / price discovery,但对我们 desk 最值钱的旁支很清楚:既然 LSD basis 会被收益率、波动、流动性、注意力持续推来推去,而且公平值又不是固定 1:1,那么短周期最该测的不是“回 1”神话,而是“回 rolling fair basis”这条 relative-value raw alpha。
CBETH-USD / ETH-USD 15m/5m K 线做 rolling-z 最小快检,验证短周期偏离有没有 admission-level 生存性。CBETH-USD vs ETH-USD5m/15mentry / exit / timeout / cost 都能一次写清CBETH spot + ETH perp,不一定要求两边都现货可融。spread_t = log(CBETH_t) - log(ETH_t) 相对 rolling fair basis 的偏离会回归long cheap CBETH + short ETH perp 或反向;按 CBETH 深度限仓;|z| 回到 close band 或 timeout 平仓;显式计入 spot+perp 费用、滑点、funding 与残余 basis drift我直接用 Coinbase Exchange 公共 candles 对 CBETH-USD 与 ETH-USD 做最小 proxy:
15m:最近 60 天,rolling lookback 7 天,timeout 24h,pair round-trip 成本先按 20 bps;5m:最近 14 天,rolling lookback 3 天,timeout 12h,成本同样按 20 bps;z 超过 1.25 / 1.5 / 2.0 开仓,|z| <= 0.25 平仓。15m 结果CBETH/ETH premium 约 12.22%,说明绝对 peg 根本不是正确锚。1.5σ:408 笔,net_total_bps ≈ +8305,mean_trade ≈ +20.4 bps,胜率 91.4%。2.0σ:220 笔,net_total_bps ≈ +6570,mean_trade ≈ +29.9 bps,胜率 99.5%。5m 结果1.5σ:203 笔,net_total_bps ≈ +1943,mean_trade ≈ +9.6 bps,胜率 71.9%。2.0σ:110 笔,net_total_bps ≈ +1799,mean_trade ≈ +16.4 bps,胜率 94.5%。15m 更像 first-production 候选。CBETH spot + ETH perp,显式加上 perp funding,而不是只看双 spot 收盘价。10 / 20 / 30 / 40 bps pair RT,确认 edge 是不是只在过低成本假设下才成立。stETH/rETH 的可交易分钟数据,比较这条线到底是 CBETH-specific 还是 LSD basis 的普遍 pocket。5m/15m 做市或 stat-arb desk 的策略论文;这篇 digest明确抽的是更适合我们 desk 的旁支。CBETH 的可融券、借贷、托管与跨 venue 对冲现实,会决定这条线能不能从研究走到 live。10.1002/fut.22556https://doi.org/10.1002/fut.22556https://api.exchange.coinbase.com/products/ETH-USD/candleshttps://api.exchange.coinbase.com/products/CBETH-USD/candleshttps://api.exchange.coinbase.com/products/CBETH-USD/book?level=1reports/artifacts/quant_digests/cbeth_eth_basis_probe_20260326_0850_15m/summary.jsonreports/artifacts/quant_digests/cbeth_eth_basis_probe_20260326_0850_15m/summary.csvreports/artifacts/quant_digests/cbeth_eth_basis_probe_20260326_0850_15m/trade_log.csvreports/artifacts/quant_digests/cbeth_eth_basis_probe_20260326_0850_15m/pair_series.csvreports/artifacts/quant_digests/cbeth_eth_basis_probe_20260326_0850_15m/meta.jsonreports/artifacts/quant_digests/cbeth_eth_basis_probe_20260326_0850_5m/summary.jsonreports/artifacts/quant_digests/cbeth_eth_basis_probe_20260326_0850_5m/summary.csvreports/artifacts/quant_digests/cbeth_eth_basis_probe_20260326_0850_5m/trade_log.csvreports/artifacts/quant_digests/cbeth_eth_basis_probe_20260326_0850_5m/pair_series.csvreports/artifacts/quant_digests/cbeth_eth_basis_probe_20260326_0850_5m/meta.json