scope: no new research overall: risk found
variant: f64_h12_floor150_mult2p0 · sample: 2026-02-10T02:15:00Z → 2026-03-28T02:15:00Z · rows=369 · step=[180]
readiness note · ffill_impact_audit · asof_universe_long_history_review · regime_review · funding-adjusted long-history
| 检查项 | 结论 | 证据 |
|---|---|---|
| 1) 每次 rebalance 信号是否只使用当时之前已收盘的 64 根 15m bar | confirmed clean | {
"source_file": "tmp_rank213_p2_admission_check.py",
"patterns_found": true,
"timestamp_step_minutes_unique": [
180
],
"expected_step_minutes": 180,
"all_timestamps_on_15m_grid": true
} |
| 2) 是否存在任何向前填充、未来 bar 泄漏、重采样错位 | risk found | {
"forward_fill_present_in_seed_builder": true,
"resample_present_in_seed_builder": false,
"future_like_ops_detected": false,
"timestamp_monotonic": true,
"timestamp_unique": true,
"note": "发现 .ffill() 路径即按 risk found 处理（即便本样本未必触发实质填充）。"
} |
| 3) universe 冻结是否在样本外重新选择 | confirmed clean | {
"runner_uses_frozen_variant_filter": true,
"runner_has_no_universe_selection_api_calls": true,
"frozen_universe_size_from_admission_summary": 30,
"frozen_universe_symbol_count": 30
} |
| 4) funding 计提是否严格在 entry < funding_time <= exit | confirmed clean | {
"funding_window_filter_found": true,
"source_file": "scripts/build_rank213_long_history_review_with_funding.py"
} |
注：出现 .ffill() 路径即按 risk found 标注；这表示流程存在“向前填充机制”，不等于必然发生了未来信息泄漏。