#002 Engineering Validation - Top Probability Framework v0
Run date: 2026-06-04 16:47:50
1. Historical universe PE availability
| cycle | stocks | anchor_positive | complete_pe | complete_price |
|---|---|---|---|---|
| 2014_2015_growth | 13 | 13 | 13 | 13 |
| 2017_2018_supply_reform | 12 | 10 | 12 | 12 |
All 2014-2015 and 2017-2018 v0 candidates have RQData pe_ratio_ttm and price rows in the cycle window. Final report should still mark this as a survivorship-biased leader reference library, not a forward alpha universe.
2. 2020-2021 snapshot backtest summary
| offset_days | samples | avg_votes | median_votes | score_ge_3 | L1_rate | L2_rate | L3_rate | L4_rate | L5_rate |
|---|---|---|---|---|---|---|---|---|---|
| -90.000 | 18.000 | 0.444 | 0.000 | 0.000 | 0.056 | 0.056 | 0.056 | 0.000 | 0.278 |
| -60.000 | 18.000 | 0.667 | 0.000 | 0.056 | 0.111 | 0.056 | 0.222 | 0.000 | 0.278 |
| -30.000 | 18.000 | 0.444 | 0.000 | 0.000 | 0.111 | 0.056 | 0.111 | 0.000 | 0.167 |
| -15.000 | 18.000 | 0.333 | 0.000 | 0.000 | 0.056 | 0.000 | 0.167 | 0.000 | 0.111 |
| -5.000 | 18.000 | 0.556 | 0.000 | 0.056 | 0.111 | 0.056 | 0.278 | 0.000 | 0.111 |
| 0.000 | 18.000 | 0.778 | 1.000 | 0.000 | 0.111 | 0.167 | 0.500 | 0.000 | 0.000 |
Interpretation: L1/L2/L3 are stock-specific and show useful cross-sectional variation near peak. L5 has limited 2020-2021 discrimination because composite option PCR history is ETF/index-level and not available before late 2019; use as market sentiment background, not stock-level evidence.
3. Current AI strict-leader main sample (~25)
| votes | stocks |
|---|---|
| 0 | 21 |
| 1 | 1 |
| 2 | 3 |
| order_book_id | name | sws_l2 | pe_ttm | pe_expansion_pct | ema20_dev | capital_flow_5d_out | composite_pcr_pct60 | votes | top_probability_pct |
|---|---|---|---|---|---|---|---|---|---|
| 300394.XSHE | 天孚通信 | 通信设备 | 178.437 | 3.978 | 0.231 | False | 0.473 | 2 | 40 |
| 300502.XSHE | 新易盛 | 通信设备 | 72.539 | 2.530 | 0.200 | False | 0.473 | 2 | 40 |
| 300308.XSHE | 中际旭创 | 通信设备 | 95.091 | 3.553 | 0.191 | False | 0.473 | 2 | 40 |
| 600183.XSHG | 生益科技 | 元件 | 87.242 | 1.447 | 0.236 | False | 0.473 | 1 | 20 |
| 300033.XSHE | 同花顺 | 软件开发 | 50.397 | -0.400 | -0.040 | False | 0.473 | 0 | 0 |
| 603629.XSHG | 利通电子 | 消费电子 | 102.040 | -0.667 | 0.112 | False | 0.473 | 0 | 0 |
| 688150.XSHG | 莱特光电 | 电子化学品Ⅱ | 99.810 | 1.048 | -0.108 | False | 0.473 | 0 | 0 |
| 300548.XSHE | 长芯博创 | 通信设备 | 192.318 | 1.159 | 0.012 | False | 0.473 | 0 | 0 |
| 300604.XSHE | 长川科技 | 半导体 | 86.415 | 0.762 | 0.020 | False | 0.473 | 0 | 0 |
| 000988.XSHE | 华工科技 | 自动化设备 | 91.352 | 1.743 | 0.038 | False | 0.473 | 0 | 0 |
| 002463.XSHE | 沪电股份 | 元件 | 58.063 | 1.415 | 0.103 | False | 0.473 | 0 | 0 |
| 603893.XSHG | 瑞芯微 | 半导体 | 63.948 | -0.302 | -0.039 | False | 0.473 | 0 | 0 |
| 688525.XSHG | 佰维存储 | 半导体 | 37.433 | nan | 0.023 | False | 0.473 | 0 | 0 |
| 688313.XSHG | 仕佳光子 | 通信设备 | 193.119 | 1.870 | 0.046 | False | 0.473 | 0 | 0 |
| 300857.XSHE | 协创数据 | 消费电子 | 67.285 | 0.582 | -0.048 | False | 0.473 | 0 | 0 |
| 688127.XSHG | 蓝特光学 | 光学光电子 | 72.721 | 0.655 | -0.021 | False | 0.473 | 0 | 0 |
| 300054.XSHE | 鼎龙股份 | 电子化学品Ⅱ | 80.398 | 0.648 | 0.017 | False | 0.473 | 0 | 0 |
| 688025.XSHG | 杰普特 | 自动化设备 | 112.243 | 1.875 | 0.036 | False | 0.473 | 0 | 0 |
| 301308.XSHE | 江波龙 | 半导体 | 41.433 | nan | 0.020 | False | 0.473 | 0 | 0 |
| 300476.XSHE | 胜宏科技 | 元件 | 74.225 | 0.848 | -0.001 | False | 0.473 | 0 | 0 |
| 001309.XSHE | 德明利 | 半导体 | 34.215 | -0.862 | -0.015 | False | 0.473 | 0 | 0 |
| 601138.XSHG | 工业富联 | 消费电子 | 39.566 | 1.298 | 0.146 | False | 0.473 | 0 | 0 |
| 002380.XSHE | 科远智慧 | IT服务Ⅱ | 60.978 | 1.700 | -0.041 | False | 0.473 | 0 | 0 |
| 603508.XSHG | 思维列控 | 计算机设备 | 14.464 | -0.178 | -0.059 | False | 0.473 | 0 | 0 |
| 300684.XSHE | 中石科技 | 电子化学品Ⅱ | 48.132 | 0.530 | -0.078 | False | 0.473 | 0 | 0 |
4. L4 production recommendation
- Capital-flow coverage: 25/25 current main-sample stocks have at least 5 rows for the current 5-trading-day check.
- Northbound coverage: 0/25 current main-sample stocks have at least 5 local stock-connect rows for the same current check.
- Recommendation: use RQData get_capital_flow active buy/sell net value as the production L4 stock-specific signal. Use northbound as a secondary annotation only when coverage exists. Reason: northbound is sparse for many STAR/ChiNext/AI-chain names and may be unavailable for current dates, while get_capital_flow is stock-level, daily, and covers the strict-leader sample consistently.
Deliverables
- 002_historical_universe_pe_availability.csv
- 002_top_probability_2020_2021_snapshot_backtest.csv
- 002_top_probability_2020_2021_snapshot_summary.csv
- 002_ai_strict_leader_main_sample_25.csv
- 002_current_ai_strict_leader_top_probability.csv
- 002_l4_capital_flow_vs_northbound_coverage.csv