生成时间:2026-03-17 08:41 UTC | 当前只做 1 个最小 clean replication:固定复用 BTC/ETH/SOL 120d 15m cache,对比 funding_8h / UTC session 的跨市场 leader-laggard follow-through。
当前 Rank 28 更诚实的 hard verdict 是 park / evidence pool,不进入 paper candidate pool。
BTC / ETH / SOL | Binance 120d | 15mutc_day 与 funding_8h2 根 15m bar 内,绝对 lead move 超过各自分位阈值且幅度最大的那条腿2 根 15m bar 跟随 leader 方向| variant | assets_tested | positive_assets | positive_asset_ratio | mean_total_return | median_total_return | mean_false_follow_ratio | mean_trades | min_trades |
|---|---|---|---|---|---|---|---|---|
| utc_day_q70 | 3.00 | 0.00 | 0.00% | -5.28% | -4.84% | 68.46% | 32.7 | 26 |
| utc_day_q60 | 3.00 | 0.00 | 0.00% | -6.73% | -6.01% | 69.16% | 41.3 | 33 |
| utc_day_q50 | 3.00 | 0.00 | 0.00% | -8.60% | -8.64% | 68.98% | 48.3 | 38 |
| funding_8h_q70 | 3.00 | 0.00 | 0.00% | -12.59% | -16.02% | 66.07% | 97.7 | 71 |
| funding_8h_q60 | 3.00 | 0.00 | 0.00% | -16.58% | -19.35% | 66.42% | 124.0 | 86 |
| funding_8h_q50 | 3.00 | 0.00 | 0.00% | -18.66% | -22.34% | 67.43% | 148.0 | 103 |
artifact:reports/artifacts/scout_rank28_crossmarket_intraday_tsmom_15m/variant_aggregate.csv
| asset | trades | total_return | win_rate | false_follow_ratio | avg_net_ret |
|---|---|---|---|---|---|
| BTC-USD | 166 | -22.89% | 32.53% | 67.47% | -0.16% |
| ETH-USD | 120 | -7.51% | 33.33% | 66.67% | -0.06% |
| SOL-USD | 86 | -19.35% | 34.88% | 65.12% | -0.25% |
| cost_bps_per_side | mean_total_return | positive_asset_ratio | mean_false_follow_ratio | mean_trades | min_trades |
|---|---|---|---|---|---|
| 6 | -16.58% | 0.00% | 66.42% | 124.0 | 86 |
| 10 | -24.40% | 0.00% | 72.65% | 124.0 | 86 |
| 15 | -33.08% | 0.00% | 81.47% | 124.0 | 86 |
| 20 | -40.72% | 0.00% | 85.00% | 124.0 | 86 |
| gate | status | actual | threshold | why_it_matters |
|---|---|---|---|---|
| positive_bucket_floor | fail | 0/3 positive buckets | >= 2 positive buckets | 不能只靠一个时间 pocket 存活。 |
| bucket_trade_floor | pass | min bucket trades = 124 | >= 4 trades per bucket | 时间切片至少要有最小样本。 |
| worst_bucket_watch | watch | worst bucket = -8.50% | ideally > -1.00% | 最差时间 pocket 明显翻负时,不应写成稳定 alpha。 |
| gate | status | actual | threshold | why_it_matters |
|---|---|---|---|---|
| neighbor_positive_floor | fail | 0/3 funding-threshold neighbors positive | >= 2 positive neighbors | 小参数邻域别一碰就碎。 |
| neighbor_trade_floor | pass | min trades across neighbors = 71 | >= 4 per asset | 参数口袋太薄时,正收益也不可靠。 |
| worst_neighbor_watch | watch | worst neighbor = -18.66% | ideally > -1.00% | 最差邻域若明显翻负,说明样本绑定很重。 |
| gate | status | actual | threshold | why_it_matters |
|---|---|---|---|---|
| positive_asset_floor | fail | 0/3 assets positive | >= 2 positive assets | 不能只靠单腿运气。 |
| min_trade_floor | pass | min trades = 86 | >= 4 per asset | 跨标的判断也要最小样本。 |
| worst_asset_watch | watch | BTC-USD total_return=-22.89% | ideally > -1.00% | 把最弱腿单独写明,避免均值掩盖。 |
| gate | status | actual | threshold | why_it_matters |
|---|---|---|---|---|
| cost_survival_floor | fail | 0/4 cost levels positive | >= 2 positive cost levels | 轻量 friction 后不该立刻归零。 |
| trade_count_floor | pass | min trades across cost ladder = 86 | >= 4 per asset | 交易数太薄时没法拿来晋级。 |
| 20bps_watch | watch | -40.72% | ideally > 0% @ 20bps | 20bps 不是硬门槛,但能看出是否只在极轻摩擦下存活。 |