{"id":4910,"date":"2026-09-18T07:43:00","date_gmt":"2026-09-17T23:43:00","guid":{"rendered":"https:\/\/laoyulaoyu.com\/?p=4910"},"modified":"2026-09-11T17:44:09","modified_gmt":"2026-09-11T09:44:09","slug":"%e5%bc%80%e6%ba%90800%e4%b8%aa%e9%87%8f%e5%8c%96%e5%9b%a0%e5%ad%90%ef%bc%8c%e4%bd%86%e6%b2%a1%e4%ba%ba%e5%91%8a%e8%af%89%e4%bd%a0pass%e6%98%af%e4%bb%80%e4%b9%88%e6%84%8f%e6%80%9d","status":"publish","type":"post","link":"https:\/\/laoyulaoyu.com\/index.php\/2026\/09\/18\/%e5%bc%80%e6%ba%90800%e4%b8%aa%e9%87%8f%e5%8c%96%e5%9b%a0%e5%ad%90%ef%bc%8c%e4%bd%86%e6%b2%a1%e4%ba%ba%e5%91%8a%e8%af%89%e4%bd%a0pass%e6%98%af%e4%bb%80%e4%b9%88%e6%84%8f%e6%80%9d\/","title":{"rendered":"\u5f00\u6e90800\u4e2a\u91cf\u5316\u56e0\u5b50\uff0c\u4f46\u6ca1\u4eba\u544a\u8bc9\u4f60pass\u662f\u4ec0\u4e48\u610f\u601d"},"content":{"rendered":"\n

\u4f5c \u8005\uff1a\u8001\u4f59\u635e\u9c7c<\/a><\/p>\n\n\n\n

\u539f\u521b\u4e0d\u6613\uff0c\u8f6c\u8f7d\u8bf7\u6807\u660e\u51fa\u5904\u53ca\u539f\u4f5c\u8005\u3002<\/mark><\/strong><\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n
\n

\u6587\u4e2d\u793a\u4f8b\u4ec5\u7528\u4e8e\u6280\u672f\u8ba8\u8bba\uff0c\u4e0d\u6784\u6210\u4efb\u4f55\u64cd\u4f5c\u5efa\u8bae\u3002
\u91cf\u5316\u7b56\u7565\u5f00\u53d1\u5e94\u4ee5\u5b66\u4e60\u548c\u6280\u672f\u4ea4\u6d41\u4e3a\u76ee\u7684\u3002
\u672c\u53f7\u4e0d\u8350\u80a1\u3001\u4e0d\u5356\u8bfe\u3001\u4e0d\u627f\u8bfa\u6536\u76ca\u3002
\u5e02\u573a\u6709\u98ce\u9669\uff0c\u8bf7\u5408\u6cd5\u5408\u89c4\u6295\u8d44\u3002<\/p>\n<\/blockquote>\n\n\n\n

\u672c\u6587\u7ea6 3400 \u5b57\uff0c\u53e6\u6709 6 \u6bb5\u53ef\u8fd0\u884c\u4ee3\u7801 | \u9884\u8ba1\u9605\u8bfb 12 \u5206\u949f\u3002\u524d\u534a\u662f\u4ecb\u7ecd\u3001\u5b89\u88c5\u548c\u7528\u6cd5\uff0c\u540e\u534a\u6709\u5b8c\u6574\u590d\u7b97\u8fc7\u7a0b\u548c\u4ee3\u7801\uff0c\u5efa\u8bae\u5148\u6536\u85cf\u3002<\/p>\n\n\n\n

\u00b7 \u00b7 \u00b7<\/p>\n\n\n\n

GitHub \u4e0a\u6709\u4e2a\u5f00\u6e90\u91cf\u5316\u7ec4\u7ec7\uff0c\u53eb QuantSkills\uff0c\u5185\u5bb9\u53ef\u4ee5\u8bf4\u662f\u5305\u7f57\u4e07\u8c61\u3002<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n
\"\"<\/figure>\n\n\n\n

\u6211\u5728\u8fd9\u4e2a\u793e\u533a\u6316\u6398\u4e86\u4e00\u4e9b\u9ad8\u4ef7\u503c\u7684\u5185\u5bb9\uff0c\u8fd9\u6b21\u7ed9\u5927\u5bb6\u5206\u4eab\u7684\u5c31\u662f\u3010\u56e0\u5b50\u8ba1\u7b97\u3011\u7684\u76f8\u5173\u5185\u5bb9\u3002\u5b83\u628a 800 \u4e2a\u9009\u80a1\u56e0\u5b50<\/strong>\u6253\u5305\u6210\u4e86 AI \u52a9\u624b\u80fd\u76f4\u63a5\u8c03\u7528\u7684\u6280\u80fd\u5305\uff0c\u6bcf\u4e2a\u56e0\u5b50\u4e00\u4e2a\u6587\u4ef6\u5939\uff0c\u516c\u5f0f\u3001\u4ee3\u7801\u3001\u9a8c\u8bc1\u62a5\u544a\u914d\u9f50\uff0cGPLv3 \u5f00\u6e90\uff0c\u4e0d\u6536\u94b1\u3002<\/p>\n\n\n\n

\u6700\u7279\u522b\u7684\u5c31\u662f\u90a3\u4efd\u9a8c\u8bc1\u62a5\u544a\u3002IC\u3001ICIR\u3001\u5206\u7ec4\u6536\u76ca\u3001\u6362\u624b\u7387\uff0c\u4e00\u6837\u4e0d\u7f3a\uff0c\u89c4\u683c\u770b\u4e0a\u53bb\u76f4\u903c\u5238\u5546\u91d1\u5de5\u7814\u62a5\u3002\u5f00\u6e90\u5708\u91cc\u80af\u628a\u9a8c\u8bc1\u505a\u6210\u6807\u51c6\u4ef6\u7684\uff0c\u6211\u89c1\u5f97\u4e0d\u591a\u3002<\/p>\n\n\n\n

\u6211\u628a\u5176\u4e2d\u4e00\u4e2a\u5e93\u7684 216 \u4efd\u62a5\u544a<\/strong>\u5168\u90e8\u4e0b\u8f7d\u4e0b\u6765\u6c47\u603b\u4e86\u4e00\u904d\uff0c\u53c8\u7528 793 \u53ea A \u80a1\u628a\u56db\u4e2a\u4ee3\u8868\u56e0\u5b50\u91cd\u65b0\u7b97\u4e86\u4e00\u904d\u3002\u8fd9\u7bc7\u6587\u7ae0\u5206\u4e24\u534a\u3002\u524d\u534a\u8bb2\u8fd9\u5957\u4e1c\u897f\u662f\u4ec0\u4e48\u3001\u600e\u4e48\u88c5\u3001\u600e\u4e48\u7528\uff0c\u4e0d\u52a8\u4ee3\u7801\u4e5f\u80fd\u641e\u5b8c\u3002\u540e\u534a\u662f\u6211\u590d\u7b97\u7684\u5168\u8fc7\u7a0b\u548c\u4ee3\u7801\uff0c\u60f3\u81ea\u5df1\u9a8c\u8bc1\u63a5\u7740\u5f80\u4e0b\u770b\u3002<\/p>\n\n\n\n

\u4e00\u3001\u8fd9\u662f\u4e2a\u4ec0\u4e48\u4e1c\u897f<\/strong><\/h2>\n\n\n\n

\u4e00\u53e5\u8bdd\uff0c\u5b83\u53ea\u7b97\u56e0\u5b50\u503c\uff0c\u522b\u7684\u4e00\u6982\u4e0d\u7ba1\u3002<\/p>\n\n\n\n

\u8fd9\u4e09\u4e2a\u56e0\u5b50\u5e93\u4e0d\u8c03\u7528\u5927\u6a21\u578b\uff0c\u4e0d\u8054\u7f51\uff0c\u4e0d\u8bfb\u4efb\u4f55 API key\u3002\u8f93\u5165\u53ea\u8981\u516d\u5217\u884c\u60c5\u6570\u636e\u3002<\/p>\n\n\n\n

date, symbol, open, high, low, close, volume<\/code><\/pre>\n\n\n\n
\u65e5\u671f\u3001\u80a1\u7968\u4ee3\u7801\u3001\u5f00\u9ad8\u4f4e\u6536\u3001\u6210\u4ea4\u91cf\u3002\u4f60\u628a\u81ea\u5df1\u7684\u884c\u60c5\u6574\u7406\u6210\u957f\u8868\u5582\u8fdb\u53bb\uff0c\u5b83\u8fd4\u56de\u4e00\u5217\u56e0\u5b50\u503c\u3002\u6570\u636e\u5b8c\u5168\u7531\u4f7f\u7528\u8005\u8d1f\u8d23\uff0c\u4ed3\u5e93\u91cc\u8fde\u4e00\u4efd\u771f\u5b9e\u884c\u60c5\u90fd\u4e0d\u5e26\uff0c\u53ea\u653e\u4e86 toy data\u3002<\/p>\n\n\n\n
\u7ec4\u7ec7\u4e0b\u6700\u91cd\u7684\u8d44\u4ea7\u662f\u4e09\u4e2a\u56e0\u5b50\u5e93\u3002<\/p>\n\n\n\n
\u4ed3\u5e93<\/th>\u56e0\u5b50\u6570<\/th>\u8986\u76d6\u65b9\u5411<\/th><\/tr>
directional-alpha<\/td>296<\/td>\u8d8b\u52bf 148\u3001\u52a8\u91cf 50\u3001\u7a81\u7834 48\u3001\u53cd\u8f6c 25\u3001\u901a\u9053 25<\/td><\/tr>
risk-pattern-alpha<\/td>288<\/td>\u6ce2\u52a8\u7387\u3001K \u7ebf\u5f62\u6001\u3001\u9707\u8361\u3001\u56de\u64a4<\/td><\/tr>
volume-stat-alpha<\/td>216<\/td>\u6210\u4ea4\u91cf 48\u3001\u91cf\u4ef7 48\u3001\u65f6\u5e8f\u6392\u540d 48\u3001\u6536\u76ca\u5206\u5e03 48\u3001\u6d41\u52a8\u6027 24<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
\u4e09\u4e2a\u52a0\u8d77\u6765 800\u3002\u6240\u4ee5\u6211\u6807\u9898”800 \u4e2a\u56e0\u5b50”\u8fd9\u4e2a\u8bf4\u6cd5\uff0c\u6570\u5b57\u662f\u771f\u7684\u3002<\/p>\n\n\n\n
\u6bcf\u4e2a\u56e0\u5b50\u662f\u4e00\u4e2a\u72ec\u7acb\u6587\u4ef6\u5939\uff0c\u957f\u8fd9\u6837\u3002<\/p>\n\n\n\n
factors\/\n\u00a0 R020-5d-z-scored-volume-expansion\/\n\u00a0 \u00a0 SKILL.md \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u00a0 \u7ed9 AI \u52a9\u624b\u8bfb\u7684\u8bf4\u660e\n\u00a0 \u00a0 scripts\/factor.py \u00a0 \u00a0 \u00a0\u56e0\u5b50\u516c\u5f0f\u7684\u5b9e\u73b0\n\u00a0 \u00a0 references\/formula.md \u00a0\u6570\u5b66\u5b9a\u4e49\n\u00a0 \u00a0 scripts\/validate.py \u00a0 \u00a0\u81ea\u68c0\u811a\u672c\n\u00a0 \u00a0 validation_real\/ \u00a0 \u00a0 \u00a0 \u9a8c\u8bc1\u62a5\u544a<\/code><\/pre>\n\n\n\n
SKILL.md<\/code> \u662f\u7ed9 AI \u52a9\u624b\u8bfb\u7684\u8bf4\u660e\uff0cfactor.py<\/code> \u662f\u516c\u5f0f\u5b9e\u73b0\uff0cformula.md<\/code> \u662f\u6570\u5b66\u5b9a\u4e49\uff0cvalidate.py<\/code> \u662f\u81ea\u68c0\u811a\u672c\uff0cvalidation_real\/<\/code> \u653e\u9a8c\u8bc1\u62a5\u544a\u3002\u7ed3\u6784\u5f88\u5e72\u51c0\u3002<\/p>\n\n\n\n
\u7b2c\u4e00\u6b21\u7528\u7684\u65f6\u5019\u6211\u662f\u975e\u5e38\u8ba4\u53ef\u8fd9\u79cd\u65b9\u5f0f\u7684\uff0c\u89c9\u5f97\u7ec8\u4e8e\u6709\u4eba\u628a\u9a8c\u8bc1\u505a\u6210\u4e86\u6807\u51c6\u4ef6\uff0c\u800c\u4e0d\u662f\u7529\u4e00\u53e5”\u56de\u6d4b\u901a\u8fc7”\u5c31\u5b8c\u4e8b\u3002<\/p>\n\n\n\n
\u6709\u4e2a\u7ec6\u8282\u503c\u5f97\u7559\u610f\u3002\u76ee\u5f55\u9875\u4ed3\u5e93\u62ff\u4e86 1975 \u4e2a\u661f<\/strong>\uff0c\u771f\u6b63\u88c5\u56e0\u5b50\u7684\u4e09\u4e2a\u5e93\u52a0\u8d77\u6765 28 \u4e2a\u3002\u6536\u85cf\u7684\u95e8\u69db\u662f\u4e09\u5341\u79d2\uff0c\u4f7f\u7528\u7684\u95e8\u69db\u662f\u534a\u5929\uff0c\u661f\u6807\u91cf\u7684\u6c38\u8fdc\u662f\u524d\u8005\u3002\u8fd9\u4e2a\u53cd\u5dee\uff0c\u540e\u9762\u8fd8\u4f1a\u7528\u5230\u3002<\/p>\n\n\n\n
\u4e8c\u3001\u6211\u5b9e\u6d4b\u7684\u7ed3\u679c\uff0c\u5148\u7ed9\u7b54\u6848<\/strong><\/h2>\n\n\n\n
\u4e24\u4ef6\u4e8b\u3002\u7b2c\u4e00\u4ef6\uff0c\u628a volume-stat \u5e93 216 \u4e2a\u56e0\u5b50\u7684\u9a8c\u8bc1\u62a5\u544a\u5168\u4e0b\u8f7d\uff0c\u6c47\u603b\u3002\u7b2c\u4e8c\u4ef6\uff0c\u7528\u6211\u81ea\u5df1\u7684\u6570\u636e\u628a\u5176\u4e2d\u56db\u4e2a\u91cd\u65b0\u7b97\u4e00\u904d\u3002<\/p>\n\n\n\n
\u9009\u7684\u56db\u4e2a\u662fR020\u3001216 \u4e2a\u91cc\u516c\u5e03 IC \u6700\u9ad8\u7684 R684\u3001\u4e2d\u4f4d\u9644\u8fd1\u7684 R459\u3001\u6700\u4f4e\u7684 R617\u3002<\/p>\n\n\n\n
\u56e0\u5b50<\/th>\u516c\u5e03IC<\/th>\u6211\u7684IC<\/th>\u516c\u5e03\u591a\u7a7a<\/th>\u6211\u7684\u591a\u7a7a<\/th>\u516c\u5e03\u6362\u624b<\/th>\u6211\u7684\u6362\u624b<\/th><\/tr>
R020<\/td>+0.0037<\/td>-0.0055<\/td>+0.073%<\/td>-0.066%<\/td>0.704<\/td>0.692<\/td><\/tr>
R684<\/td>+0.0166<\/td>-0.0032<\/td>+0.211%<\/td>+0.070%<\/td>0.205<\/td>0.175<\/td><\/tr>
R459<\/td>+0.0018<\/td>-0.0108<\/td>+0.059%<\/td>-0.014%<\/td>0.514<\/td>0.505<\/td><\/tr>
R617<\/td>-0.0278<\/td>-0.0220<\/td>+0.006%<\/td>-0.008%<\/td>0.172<\/td>0.141<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
\"\"<\/figure>\n\n\n\n
\u56db\u4e2a\u56e0\u5b50\uff0c\u516c\u5e03\u503c\u4e0e\u6211\u590d\u7b97\u503c\u7684\u5bf9\u7167<\/p>\n\n\n\n
\u770b\u8868\u4e4b\u524d\u5148\u82b1\u4e8c\u5341\u79d2\u8bb2\u6e05 Rank IC<\/strong>\uff1a\u6bcf\u5929\u628a\u6240\u6709\u80a1\u7968\u6309\u56e0\u5b50\u503c\u6392\u4e00\u6b21\u961f\uff0c\u518d\u6309\u5b83\u4eec\u672a\u6765\u4e94\u5929\u7684\u5b9e\u9645\u6da8\u5e45\u6392\u4e00\u6b21\u961f\uff0c\u770b\u4e24\u6b21\u6392\u961f\u50cf\u4e0d\u50cf\u3002\u5b8c\u5168\u4e00\u6837\u662f 1\uff0c\u6beb\u65e0\u5173\u7cfb\u662f 0\u3002\u4e1a\u5185\u7ecf\u9a8c\u95e8\u69db\u5728 0.02 \u5230 0.03\uff0c\u80fd\u7a33\u5b9a\u505a\u5230 0.05 \u7684\u5df2\u7ecf\u662f\u751f\u4ea7\u7ea7\u6c34\u5e73\u3002<\/p>\n\n\n\n
\u7ed3\u679c\u4e09\u6761\u3002<\/p>\n\n\n\n
\u7b2c\u4e00\u6761\uff0c\u6362\u624b\u7387\u90a3\u4e00\u5217\u51e0\u4e4e\u5b8c\u7f8e\u5bf9\u4e0a\u30020.704 \u5bf9 0.692\uff0c0.205 \u5bf9 0.175\uff0c0.514 \u5bf9 0.505\u3002\u6362\u624b\u7387\u7531\u56e0\u5b50\u503c\u548c\u80a1\u7968\u6c60\u5171\u540c\u51b3\u5b9a\uff0c\u5b83\u8fd9\u4e48\u63a5\u8fd1\uff0c\u8bf4\u660e\u6211\u7684\u6570\u636e\u7ba1\u9053\u548c\u5b83\u7684\u516c\u5f0f\u53e3\u5f84\u4e00\u81f4\uff0c\u6211\u7b97\u7684\u4e0d\u662f\u53e6\u4e00\u5957\u4e1c\u897f\u3002<\/p>\n\n\n\n
\u7b2c\u4e8c\u6761\uff0cIC \u90a3\u4e00\u5217\u56db\u4e2a\u91cc\u4e09\u4e2a\u7b26\u53f7\u7ffb\u8f6c\u3002\u516c\u5e03\u503c\u6700\u9ad8\u7684 R684\uff0c\u5728\u6211\u8fd9\u4efd\u6570\u636e\u4e0a\u8dd1\u51fa\u6765\u662f\u8d1f\u7684\u3002<\/p>\n\n\n\n
\u7b2c\u4e09\u6761\uff0c\u6211\u62ff\u4e00\u4e32\u7eaf\u968f\u673a\u6570\u5f53\u56e0\u5b50\u8dd1\u4e86\u4e00\u904d\uff0cIC 0.0005\uff0ct \u503c 0.48\uff0c\u843d\u5728 216 \u4e2a\u56e0\u5b50\u7684\u4e2d\u4f4d\u9644\u8fd1\u3002\u8fd9\u6279\u56e0\u5b50\u4f5c\u4e3a\u4e00\u4e2a\u6574\u4f53\uff0c\u5bf9\u540e\u7eed\u6da8\u8dcc\u7684\u533a\u5206\u80fd\u529b\u548c\u968f\u673a\u6570\u5728\u540c\u4e00\u4e2a\u91cf\u7ea7\u3002<\/p>\n\n\n\n
\u800c 216 \u4e2a\u56e0\u5b50\uff0c\u5168\u90e8\u6807\u6ce8 pass<\/strong>\u3002<\/p>\n\n\n\n
pass \u5230\u5e95\u9a8c\u4e86\u4ec0\u4e48\uff0c\u540e\u534a\u8bb2\u3002\u5148\u628a\u5b89\u88c5\u548c\u7528\u6cd5\u8bf4\u5b8c\u3002<\/p>\n\n\n\n
\u4e09\u3001\u600e\u4e48\u88c5<\/strong><\/h2>\n\n\n\n
\u4e09\u6b65\uff0c\u4e94\u5206\u949f\u3002<\/p>\n\n\n\n
\u7b2c\u4e00\u6b65\u62ff\u4ee3\u7801\u3002git clone --depth 1 https:\/\/github.com\/quantskills\/skill-quant-factor-volume-stat-alpha.git<\/code>\u3002\u4ed3\u5e93\u4e0d\u5927\uff0c\u4e24\u767e\u591a\u4e2a\u56e0\u5b50\u6587\u4ef6\u5939\uff0c\u51e0\u79d2\u949f\u4e0b\u5b8c\u3002<\/p>\n\n\n\n
\u7b2c\u4e8c\u6b65\u88c5\u4f9d\u8d56\u3002\u6574\u4e2a\u4ed3\u5e93\u53ea\u8981\u4e09\u4e2a\u5305\uff0cpip install numpy pandas pyarrow<\/code>\u3002\u6ca1\u6709 API key\uff0c\u6ca1\u6709\u6570\u636e\u5e73\u53f0 SDK\uff0c\u8fd9\u4e00\u6761\u662f\u771f\u5fc3\u7701\u4e8b\u3002<\/p>\n\n\n\n
\u7b2c\u4e09\u6b65\u5907\u6570\u636e\u3002\u51c6\u5907\u4e00\u4e2a\u957f\u8868 CSV\uff0c\u6bcf\u884c\u662f\u4e00\u53ea\u80a1\u7968\u5728\u4e00\u4e2a\u4ea4\u6613\u65e5\u7684\u8bb0\u5f55\uff0c\u516d\u5217\uff0c\u5217\u540d\u7167\u7b2c\u4e00\u8282\u7684\u6765\u3002\u6570\u636e\u4ece\u54ea\u6765\u90fd\u884c\uff0cakshare \u7684\u514d\u8d39\u63a5\u53e3\u5c31\u80fd\u62c9\uff0c\u62c9\u53d6\u811a\u672c\u5728\u6587\u672b\u3002<\/p>\n\n\n\n
\u56db\u3001\u88c5\u597d\u4e4b\u540e\u600e\u4e48\u7528<\/strong><\/h2>\n\n\n\n
\u6bcf\u4e2a\u56e0\u5b50\u7684 compute_factor<\/code> \u662f\u72ec\u7acb\u51fd\u6570\uff0c\u52a8\u6001\u5bfc\u5165\u5c31\u80fd\u8dd1\uff0c\u4e0d\u7528\u88c5\u6210\u5305\u3002<\/p>\n\n\n\n
import importlib.util\nimport sys\nfrom pathlib import Path\n\n\ndef load_factor(factor_dir):\n    \"\"\"\u4ece\u56e0\u5b50\u76ee\u5f55\u91cc\u52a8\u6001\u8f7d\u5165 factor.py\u3002\"\"\"\n    path = Path(factor_dir) \/ \"scripts\" \/ \"factor.py\"\n    spec = importlib.util.spec_from_file_location(\"factor_module\", path)\n    module = importlib.util.module_from_spec(spec)\n    sys.modules[spec.name] = module\n    spec.loader.exec_module(module)\n    return module\n\n\nmod = load_factor(\"skill-quant-factor-volume-stat-alpha\/factors\/R020-5d-z-scored-volume-expansion\")\npanel = mod.compute_factor(panel)\ncol = mod.FACTOR_COLUMN<\/code><\/pre>\n\n\n\n
\u8dd1\u5b8c panel<\/code> \u91cc\u4f1a\u591a\u4e00\u5217\uff0c\u5217\u540d\u5c31\u662f\u56e0\u5b50\u77ed\u540d\uff0c\u6bd4\u5982 5d-z-scored-volume-expansion<\/code>\u3002\u8fd9\u4e00\u5217\u5c31\u662f\u6bcf\u5929\u7684\u56e0\u5b50\u503c\uff0c\u62ff\u53bb\u6392\u5e8f\u3001\u5206\u7ec4\u3001\u7b5b\u80a1\u90fd\u884c\u3002<\/p>\n\n\n\n
\u4e00\u53e5\u8bdd\u5b9a\u4f4d\u3002\u516c\u5f0f\u5b83\u5199\u597d\u4e86\uff0c\u7ed3\u6784\u5b83\u642d\u597d\u4e86\uff0c\u6570\u636e\u8981\u4f60\u81ea\u5df1\u63a5\uff0c\u5224\u65ad\u8981\u4f60\u81ea\u5df1\u505a\u3002<\/p>\n\n\n\n
\u4ee5\u4e0a\u662f\u666e\u901a\u8bfb\u8005\u7684\u90e8\u5206\uff0c\u88c5\u4e0a\u5c31\u80fd\u7528\u3002\u5f80\u4e0b\u662f\u52a8\u624b\u90e8\u5206\uff0c\u5199\u7ed9\u60f3\u81ea\u5df1\u7b97 IC\u3001\u60f3\u9a8c\u8bc1\u90a3\u4e9b\u62a5\u544a\u7684\u4eba\u3002<\/p>\n\n\n\n
\u4e94\u3001216 \u4efd\u62a5\u544a\u7684\u771f\u9762\u76ee<\/strong><\/h2>\n\n\n\n
\u5148\u628a\u6c47\u603b\u8868\u653e\u51fa\u6765\u3002<\/p>\n\n\n\n
\u6307\u6807<\/th>\u6570\u503c<\/th><\/tr>
5 \u65e5 Rank IC \u4e2d\u4f4d\u6570<\/td>0.001765<\/td><\/tr>
\u7edd\u5bf9\u503c\u5c0f\u4e8e 0.01 \u7684\u56e0\u5b50<\/td>189 \/ 216<\/td><\/tr>
\u7edd\u5bf9\u503c\u5927\u4e8e\u7b49\u4e8e 0.03 \u7684\u56e0\u5b50<\/td>0<\/td><\/tr>
5 \u65e5 ICIR \u4e2d\u4f4d\u6570<\/td>0.0851<\/td><\/tr>
\u591a\u7a7a\u7ec4\u5408 5 \u65e5\u4ef7\u5dee\u4e2d\u4f4d\u6570<\/td>0.058%<\/td><\/tr>
\u591a\u7a7a\u4ef7\u5dee\u5c0f\u4e8e\u7b49\u4e8e 0 \u7684\u56e0\u5b50<\/td>35 \/ 216<\/td><\/tr>
Top \u7ec4 5 \u65e5\u6362\u624b\u7387\u4e2d\u4f4d\u6570<\/td>53.6%<\/td><\/tr>
\u6807\u6ce8 pass \u7684\u56e0\u5b50<\/td>216 \/ 216<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
\"\"<\/figure>\n\n\n\n
216 \u4e2a\u56e0\u5b50\u7684 5 \u65e5 Rank IC \u5206\u5e03\uff0c189 \u4e2a\u843d\u5728\u6b63\u8d1f 0.01 \u4e4b\u95f4<\/p>\n\n\n\n
Rank IC \u4e2d\u4f4d\u6570 0.0018\u3002189 \u4e2a\u56e0\u5b50\u7684\u7edd\u5bf9\u503c\u4e0d\u5230 0.01\uff0c\u6ca1\u6709\u4e00\u4e2a\u6478\u5230 0.03\u3002\u800c\u5168\u8868 216 \u4e2a\uff0c\u5168\u90e8\u5199\u7740 pass\u3002<\/p>\n\n\n\n
\u8fd9\u4e2a pass \u662f\u600e\u4e48\u9a8c\u51fa\u6765\u7684\u3002\u6211\u6253\u5f00 validate.py<\/code>\uff0c\u5168\u6587 53 \u884c\u3002\u5b83\u505a\u56db\u4ef6\u4e8b\uff0c\u7528 np.random.default_rng(11)<\/code> \u751f\u6210 12 \u53ea\u80a1\u7968\u3001180 \u4e2a\u4ea4\u6613\u65e5\u7684\u968f\u673a\u6e38\u8d70\u6570\u636e\uff0c\u8dd1\u4e00\u904d\u516c\u5f0f\uff0c\u65ad\u8a00\u7ed3\u679c\u91cc\u5b58\u5728\u6709\u9650\u503c\uff0c\u6253\u5370\u4e00\u884c status ok<\/code>\u3002<\/p>\n\n\n\n
\u6ca1\u6709 IC\uff0c\u6ca1\u6709\u5206\u7ec4\u6536\u76ca\u3002\u5b83\u7684\u5224\u5b9a\u6761\u4ef6\u662f\u4ee3\u7801\u8dd1\u8d77\u6765\u4e86\uff0c\u6ca1\u5d29\uff0c\u4e5f\u6ca1\u5410\u51fa NaN\u3002<\/p>\n\n\n\n
\u90a3\u62a5\u544a\u91cc\u90a3\u4e9b IC \u6570\u5b57\u4ece\u54ea\u6765\u3002report.md<\/code> \u6700\u540e\u4e00\u884c\u5199\u4e86\uff0c\u7531 real_market_data\/<\/code> \u76ee\u5f55\u91cc\u7684\u7f13\u5b58\u771f\u5b9e\u884c\u60c5\u8ba1\u7b97\u5f97\u5230\u3002\u6211\u628a\u6574\u4e2a\u4ed3\u5e93\u7ffb\u4e86\u4e00\u904d\uff0c868 \u4e2a Markdown\u3001432 \u4e2a Python\u3001219 \u4e2a JSON\uff0c\u6ca1\u6709\u8fd9\u4e2a\u76ee\u5f55\u3002\u5b83\u5728\u81ea\u5df1\u7684 .gitignore<\/code> \u91cc\u3002<\/p>\n\n\n\n
\u6240\u4ee5\u73b0\u72b6\u662f\uff0c\u6570\u5b57\u516c\u5e03\u4e86\uff0c\u4f46\u590d\u7b97\u4e0d\u4e86\u3002<\/p>\n\n\n\n
\u516d\u3001\u52a8\u624b\u533a\uff0c\u81ea\u5df1\u7b97 IC \u7684\u4e8c\u5341\u884c\u4ee3\u7801<\/strong><\/h2>\n\n\n\n
\u60f3\u77e5\u9053\u4e00\u4e2a\u56e0\u5b50\u6709\u6ca1\u6709\u7528\uff0c\u522b\u770b\u6807\u7b7e\uff0c\u81ea\u5df1\u8dd1\u3002\u4e24\u6bb5\u4ee3\u7801\u3002<\/p>\n\n\n\n
\u7b2c\u4e00\u6bb5\u62c9\u6570\u636e\u3002\u514d\u8d39\u5efa\u8bae\u7528akshare \u7684\u65b0\u6d6a\u6e90\uff0c\u4e1c\u8d22\u6e90\u5728\u6279\u91cf\u8bf7\u6c42\u4e0b\u4f1a\u88ab\u9650\u6d41\uff0c\u5b9e\u6d4b\u5341\u6b21\u5168\u90e8\u5931\u8d25\uff0c\u65b0\u6d6a\u6e90\u5341\u6b21\u5168\u8fc7\u3002<\/p>\n\n\n\n
import akshare as ak\nimport pandas as pd\n\n\ndef load_panel(codes, start=\"20211201\", end=\"20260911\"):\n    \"\"\"\u628a\u80a1\u7968\u4ee3\u7801\u5217\u8868\u62c9\u6210\u957f\u8868\u3002\u65b0\u6d6a\u6e90\u6bd4\u4e1c\u8d22\u6e90\u7a33\uff0c\u6279\u91cf\u8bf7\u6c42\u4e0b\u4e1c\u8d22\u4f1a\u9650\u6d41\u3002\"\"\"\n    frames = []\n    for code in codes:\n        prefix = \"sh\" if code.startswith((\"6\", \"9\")) else \"sz\"\n        df = ak.stock_zh_a_daily(\n            symbol=prefix + code, start_date=start, end_date=end, adjust=\"qfq\"\n        )\n        if df is None or len(df) < 250:\n            continue\n        df = df[[\"date\", \"open\", \"high\", \"low\", \"close\", \"volume\"]].copy()\n        df[\"symbol\"] = code\n        frames.append(df)\n\n    panel = pd.concat(frames, ignore_index=True)\n    panel[\"date\"] = pd.to_datetime(panel[\"date\"])\n    return panel<\/code><\/pre>\n\n\n\n
\u7b2c\u4e8c\u6bb5\u7b97\u6307\u6807\u3002\u6a2a\u622a\u9762 Spearman \u76f8\u5173\u52a0\u4e94\u5206\u7ec4\uff0c\u548c\u4ed3\u5e93\u62a5\u544a\u7684\u53e3\u5f84\u5bf9\u9f50\u3002<\/p>\n\n\n\n
import numpy as np\n\n\ndef evaluate(panel, col, horizon=5, cost=0.003):\n    \"\"\"\u7b97 5 \u65e5 Rank IC\u3001ICIR\u3001\u4e94\u5206\u7ec4\u591a\u7a7a\u4ef7\u5dee\u3001Top \u7ec4\u6362\u624b\u7387\uff0c\u518d\u6263\u4e00\u6b21\u6210\u672c\u3002\"\"\"\n    d = panel.sort_values([\"symbol\", \"date\"]).copy()\n    d[\"fwd\"] = d.groupby(\"symbol\")[\"close\"].shift(-horizon) \/ d[\"close\"] - 1\n\n    def one_day(g):\n        g = g.dropna(subset=[col, \"fwd\"])\n        if len(g) < 30 or g[col].nunique() < 5:\n            return np.nan\n        return g[col].corr(g[\"fwd\"], method=\"spearman\")\n\n    ic = d.groupby(\"date\").apply(one_day, include_groups=False).dropna()\n\n    spreads, turnovers, prev = [], [], set()\n    for _, g in d.groupby(\"date\"):\n        g = g.dropna(subset=[col, \"fwd\"])\n        if len(g) < 30:\n            continue\n\n        q = pd.qcut(g[col].rank(method=\"first\"), 5, labels=False)\n        top, bot = g.loc[q == 4], g.loc[q == 0]\n        spreads.append(top[\"fwd\"].mean() - bot[\"fwd\"].mean())\n\n        cur = set(top[\"symbol\"])\n        if prev:\n            turnovers.append(1 - len(prev & cur) \/ len(prev))\n        prev = cur\n\n    ls = float(np.mean(spreads))\n    to = float(np.mean(turnovers))\n\n    return {\n        \"ic\": float(ic.mean()),\n        \"icir\": float(ic.mean() \/ ic.std(ddof=0)),\n        \"t\": float(ic.mean() \/ ic.std(ddof=0) * np.sqrt(len(ic))),\n        \"observations\": int(len(ic)),\n        \"long_short\": ls,\n        \"turnover\": to,\n        \"net\": ls - to * cost,\n    }<\/code><\/pre>\n\n\n\n
\u8dd1\u51fa\u6765\u516d\u4e2a\u6570\u5b57\uff0cic<\/code>\u3001icir<\/code>\u3001t<\/code>\u3001long_short<\/code>\u3001turnover<\/code>\u3001net<\/code>\u3002\u4efb\u4f55\u4e00\u4efd\u56e0\u5b50\u62a5\u544a\uff0c\u53ea\u8981\u5b83\u6ca1\u544a\u8bc9\u4f60\u6837\u672c\u533a\u95f4\u548c\u80a1\u7968\u6c60\uff0c\u4f60\u62ff\u8fd9\u6bb5\u4ee3\u7801\u8dd1\u4e00\u904d\uff0c\u5c31\u80fd\u628a\u8bdd\u95ee\u6e05\u695a\u3002<\/p>\n\n\n\n
\u4e03\u3001\u6362\u624b\u7387\u5bf9\u4e0a\u4e86\uff0cIC \u7ffb\u4e86\u5411<\/strong><\/h2>\n\n\n\n
\u7b2c\u4e8c\u8282\u90a3\u5f20\u8868\uff0c\u503c\u5f97\u5c55\u5f00\u8bb2\u7684\u662f\u4e24\u5c42\u3002<\/p>\n\n\n\n
\u6362\u624b\u7387\u90a3\u5c42\uff0c\u51e0\u4e4e\u9010\u4f4d\u590d\u73b0\u3002\u8fd9\u8bc1\u660e\u7684\u4e0d\u662f\u56e0\u5b50\u597d\uff0c\u662f\u6211\u7684\u7ba1\u7ebf\u548c\u5b83\u7684\u53e3\u5f84\u4e00\u81f4\u3002\u6240\u4ee5\u540e\u9762 IC \u4e0a\u7684\u5dee\u5f02\uff0c\u4e0d\u80fd\u63a8\u7ed9”\u6211\u7b97\u9519\u4e86”\u3002<\/p>\n\n\n\n
IC \u90a3\u5c42\uff0c\u56db\u4e2a\u91cc\u4e09\u4e2a\u7b26\u53f7\u7ffb\u8f6c\u3002R020 \u7531\u6b63\u7684 0.0037 \u8f6c\u4e3a\u8d1f\u7684 0.0055\u3002R459 \u7531\u6b63\u7684 0.0018 \u8f6c\u4e3a\u8d1f\u7684 0.0108\u3002R684 \u66f4\u5f7b\u5e95\uff0c\u5b83\u662f 216 \u4e2a\u91cc IC \u6700\u9ad8\u7684\uff0c\u516c\u5e03 0.0166\uff0c\u6211\u8fd9\u4efd\u6570\u636e\u4e0a\u8dd1\u51fa\u6765\u662f\u8d1f\u7684 0.0032\u3002<\/p>\n\n\n\n
\u4e3a\u4e86\u8ba9\u8fd9\u4e2a\u7ed3\u8bba\u7ad9\u5f97\u4f4f\uff0c\u6211\u5148\u505a\u4e86\u65b9\u5411\u6027\u81ea\u68c0\u30025 \u65e5\u53cd\u8f6c\u56e0\u5b50 IC \u6b63\u7684 0.0128\uff0ct \u503c 2.43\u300220 \u65e5\u52a8\u91cf IC \u8d1f\u7684 0.0287\uff0ct \u503c -4.96\u3002\u4e24\u4e2a\u65b9\u5411\u90fd\u7b26\u5408 A \u80a1\u77ed\u5468\u671f\u7279\u5f81\uff0c\u8bf4\u660e\u6211\u7684\u7ba1\u7ebf\u6ca1\u641e\u53cd\u3002\u968f\u673a\u6570\u90a3\u7ec4\u5728\u7b2c\u4e8c\u8282\u8bf4\u8fc7\u4e86\uff0cIC 0.0005\uff0c\u843d\u5728\u56e0\u5b50\u4eec\u7684\u4e2d\u4f4d\u9644\u8fd1\u3002<\/p>\n\n\n\n
\u56db\u4e2a\u91cc\u552f\u4e00\u65b9\u5411\u4e00\u81f4\u7684\u662f R617\uff0c\u516c\u5e03 -0.0278\uff0c\u6211\u7b97\u51fa -0.0220\uff0ct \u503c -6.09\uff0c\u662f\u552f\u4e00\u7edf\u8ba1\u663e\u8457\u7684\u3002\u4f46\u8fd9\u4e2a\u663e\u8457\u6027\u5403\u4e0d\u5230\u94b1\u3002\u5b83\u7684\u591a\u7a7a\u4ef7\u5dee\u53ea\u6709 -0.008%\uff0c\u53cd\u8fc7\u6765\u7528\u4e5f\u53ea\u6709 0.008%\uff0c\u4e0d\u5230\u4ea4\u6613\u6210\u672c\u7684\u56db\u5206\u4e4b\u4e00\u3002<\/p>\n\n\n\n
IC \u663e\u8457\u548c\u80fd\u8d5a\u94b1\uff0c\u4e2d\u95f4\u9694\u7740\u597d\u51e0\u9053\u5899\u3002<\/p>\n\n\n\n
\u518d\u7b97\u4e00\u6b65\u5b83\u6ca1\u7b97\u7684<\/h3>\n\n\n\n
\u4ed3\u5e93\u62a5\u544a\u91cc\u6ca1\u6709\u6210\u672c\u8fd9\u4e00\u9879\u3002\u6211\u8865\u4e0a\u3002\u6309\u6bcf\u5355\u4f4d\u6362\u624b 0.30% \u4f30\u7b97\u5f80\u8fd4\u6210\u672c\uff0c\u4f63\u91d1\u53cc\u8fb9\u3001\u5370\u82b1\u7a0e\u5355\u8fb9\u3001\u518d\u52a0\u6ed1\u70b9\uff0c\u8fd9\u4e2a\u6570\u5bf9\u5927\u4e2d\u76d8\u80a1\u7b97\u5ba2\u6c14\u3002<\/p>\n\n\n\n
\u56e0\u5b50<\/th>\u591a\u7a7a\u4ef7\u5dee<\/th>\u4f30\u7b97\u6210\u672c<\/th>\u51c0\u989d<\/th>\u6298\u5e74\u5316<\/th><\/tr>
R020<\/td>-0.066%<\/td>0.208%<\/td>-0.274%<\/td>\u7ea6 -13.3%<\/td><\/tr>
R684<\/td>+0.070%<\/td>0.053%<\/td>+0.017%<\/td>\u7ea6 +0.8%<\/td><\/tr>
R459<\/td>-0.014%<\/td>0.151%<\/td>-0.166%<\/td>\u7ea6 -8.1%<\/td><\/tr>
R617<\/td>-0.008%<\/td>0.042%<\/td>-0.050%<\/td>\u7ea6 -2.4%<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n
\"\"<\/figure>\n\n\n\n
\u6263\u6389\u6362\u624b\u6210\u672c\u524d\u540e\u7684\u591a\u7a7a\u4ef7\u5dee\uff0c\u4e09\u8272\u5bf9\u7167<\/p>\n\n\n\n
\u56db\u4e2a\u91cc\u4e09\u4e2a\u6263\u5b8c\u6210\u672c\u662f\u8d1f\u7684\u3002R020 \u6700\u5178\u578b\uff0c\u591a\u7a7a\u4ef7\u5dee\u672c\u8eab\u5df2\u7ecf\u662f\u8d1f\u7684\uff0c\u8fd8\u8981\u518d\u4ed8 0.208% \u7684\u6362\u624b\u6210\u672c\uff0c\u56e0\u4e3a\u5b83\u6bcf\u4e94\u5929\u6362\u6389\u5c06\u8fd1\u4e03\u6210\u4ed3\u4f4d\u3002\u552f\u4e00\u4e3a\u6b63\u7684 R684\uff0c\u5e74\u5316 0.8%\uff0c\u800c\u5b83\u7684 t \u503c -0.78\uff0c\u843d\u5728\u566a\u97f3\u91cc\u3002<\/p>\n\n\n\n
\u5230\u8fd9\u4e00\u6b65\uff0c216 \u4efd\u62a5\u544a\u91cc pass \u8fd9\u4e2a\u8bcd\u7684\u5206\u91cf\uff0c\u5927\u6982\u80fd\u6382\u51fa\u6765\u4e86\u3002\u7167\u5b83\u7684\u6807\u7b7e\u6311\u56e0\u5b50\uff0c\u4f60\u4f1a\u6311\u5230 R684\uff0c216 \u4e2a\u91cc IC \u6700\u9ad8\u7684\u90a3\u4e2a\u3002\u5b83\u5728\u6211\u8fd9\u4efd\u6570\u636e\u4e0a IC \u662f\u8d1f\u7684\uff0c\u6263\u5b8c\u6210\u672c\u5e74\u5316 0.8%\u3002<\/p>\n\n\n\n
800 \u7684\u6784\u6210\uff0c\u5176\u5b9e\u662f 9 \u4e58 24<\/h3>\n\n\n\n
\u7ffb\u5230\u4ed3\u5e93\u6839\u76ee\u5f55\uff0c\u6709\u4e2a\u6587\u4ef6\u53eb repo_summary.json<\/code>\uff0c\u91cc\u9762\u6709\u4e00\u884c bases \u5206\u5e03\uff0c\u628a 216 \u4e2a\u56e0\u5b50\u6309\u57fa\u7840\u516c\u5f0f\u5f52\u7c7b\u3002volume_ratio\u3001volume_z\u3001dollar_volume\u3001price_volume_corr\u3001obv_slope\u3001ts_rank_close\u3001ts_rank_volume\u3001ret_skew\u3001ret_kurt\uff0c9 \u4e2a\u516c\u5f0f\uff0c\u6bcf\u4e2a 24 \u4e2a\u53d8\u4f53\uff0c\u6b63\u597d 216\u3002<\/p>\n\n\n\n
\"\"<\/figure>\n\n\n\n
800 \u4e2a\u56e0\u5b50\u7684\u6784\u6210\uff0c9 \u4e2a\u57fa\u7840\u516c\u5f0f \u00d7 24 \u7ec4\u53c2\u6570<\/p>\n\n\n\n
24 \u4e2a\u53d8\u4f53\u662f\u540c\u4e00\u4e2a\u516c\u5f0f\u6362\u7a97\u53e3\u548c\u6362\u53d8\u6362\u5f97\u6765\u7684\u3002\u4ee5 R020 \u4e3a\u4f8b\uff0c\u5b9a\u4e49\u662f z(volume_ratio, window=5)<\/code>\uff0c\u4e94\u65e5\u91cf\u6bd4\u518d\u505a\u6eda\u52a8\u6807\u51c6\u5316\u3002\u7a97\u53e3\u6362\u6210 7\u3001\u6362\u6210 10\uff0c\u6807\u51c6\u5316\u6362\u6210\u5e73\u6ed1\u3001\u6362\u6210\u65f6\u5e8f\u6392\u540d\u3001\u6362\u6210\u6ce2\u52a8\u7f29\u653e\uff0c\u5c31\u662f\u53e6\u5916\u5341\u51e0\u4e2a\u56e0\u5b50\u3002<\/p>\n\n\n\n
\u6240\u4ee5 800 \u7684\u51c6\u786e\u8bfb\u6cd5\u662f\uff0c\u5341\u51e0\u4e2a\u60f3\u6cd5\u4e58\u4e0a\u53c2\u6570\u7f51\u683c\u3002<\/p>\n\n\n\n
\u53c2\u6570\u626b\u63cf\u672c\u8eab\u662f\u6807\u51c6\u52a8\u4f5c\uff0c\u4e0d\u7b97\u7f6a\u72b6\u3002\u4f46\u5b83\u6539\u53d8\u4e86\u4e00\u4ef6\u4e8b\u7684\u8bfb\u6cd5\u3002\u540c\u4e00\u65cf\u91cc\u7684 24 \u4e2a\u53d8\u4f53\u8868\u73b0\u5fc5\u7136\u9ad8\u5ea6\u76f8\u5173\uff0c\u4e00\u4e2a\u901a\u8fc7\uff0c\u4e00\u6574\u7247\u8ddf\u7740\u901a\u8fc7\uff0c216 \u4e2a\u5168 pass \u51e0\u4e4e\u662f\u88ab\u7ed3\u6784\u51b3\u5b9a\u7684\u3002\u66f4\u5173\u952e\u7684\u662f\uff0c\u8bd5\u4e86 216 \u4e2a\u7ec4\u5408\u603b\u6709\u51e0\u4e2a\u770b\u8d77\u6765\u4e0d\u9519\uff0c\u8fd9\u4efd\u62a5\u544a\u91cc\u6ca1\u6709 DSR\uff0c\u6ca1\u6709 PBO\uff0c\u4e5f\u6ca1\u6709\u4efb\u4f55\u4e00\u5904\u63d0\u5230\u8bd5\u4e86\u591a\u5c11\u4e2a\u624d\u6311\u51fa\u8fd9\u4e9b\u3002<\/p>\n\n\n\n
\u6709\u70b9\u8bbd\u523a\u3002\u8fd9\u4e2a\u7ec4\u7ec7\u7684\u8d44\u4ea7\u76ee\u5f55\u91cc\u5c31\u8eba\u7740\u5e72\u8fd9\u4e2a\u7684 skill\uff0cskill-backtest-overfit<\/code> \u7528 DSR \u548c PBO \u8bc4\u4f30\u56de\u6d4b\u8fc7\u62df\u5408\u3002\u5de5\u5177\u90fd\u5728\uff0c\u56e0\u5b50\u5e93\u81ea\u5df1\u6ca1\u7528\u3002<\/p>\n\n\n\n
\u516b\u3001\u90a3\u8fd9\u5957\u4e1c\u897f\u8fd8\u503c\u4e0d\u503c\u5f97\u82b1\u65f6\u95f4\u770b<\/strong><\/h2>\n\n\n\n
\u503c\u5f97\uff0c\u4f46\u8981\u770b\u5bf9\u5730\u65b9\u3002<\/p>\n\n\n\n
\u503c\u5f97\u6284\u7684\u4e09\u6837\u3002<\/p>\n\n\n\n
\u4e00\u662f\u56e0\u5b50\u76ee\u5f55\u7684\u7ed3\u6784\u3002<\/strong>\u4e94\u4e2a\u4f4d\u7f6e\u628a”\u4e00\u4e2a\u56e0\u5b50\u8be5\u4ea4\u4ee3\u4ec0\u4e48”\u56fa\u5b9a\u4e0b\u6765\uff0c\u6bd4\u7edd\u5927\u591a\u6570\u4e2a\u4eba\u7814\u7a76\u8005\u7684\u6587\u4ef6\u5939\u89c4\u8303\uff0c\u53ef\u4ee5\u76f4\u63a5\u642c\u56de\u81ea\u5df1\u7684\u56e0\u5b50\u5e93\u3002<\/p>\n\n\n\n
\u4e8c\u662f\u9a8c\u8bc1\u6307\u6807\u7684\u6e05\u5355\u3002<\/strong>\u8986\u76d6\u7387\u30015 \u65e5 Rank IC\u30015 \u65e5 ICIR\u3001\u4e94\u5206\u7ec4\u4ef7\u5dee\u3001Top \u7ec4\u6362\u624b\u7387\u3001\u65e0\u672a\u6765\u51fd\u6570\u68c0\u67e5\uff0c\u516d\u9879\u9009\u5f97\u8d34\u5730\u6c14\uff0c\u5c24\u5176\u6362\u624b\u7387\uff0c\u5f88\u591a\u62a5\u544a\u5e72\u8106\u4e0d\u5199\u3002<\/p>\n\n\n\n
\u4e09\u662f”\u6570\u636e\u7531\u4f7f\u7528\u8005\u63d0\u4f9b”\u8fd9\u4e2a\u8fb9\u754c\u3002<\/strong>\u5b83\u903c\u4f60\u628a\u6570\u636e\u7ba1\u9053\u63a5\u4e0a\u53bb\uff0c\u800c\u4e0d\u662f\u5582\u4f60\u4e00\u4efd\u4e0d\u77e5\u9053\u6765\u6e90\u7684\u6210\u54c1\u3002<\/p>\n\n\n\n
\u4e0d\u8be5\u6284\u7684\u4e5f\u6709\u4e09\u6837\u3002<\/p>\n\n\n\n
\u6807\u7b7e\u4e0d\u8981\u6284\u3002pass \u5728\u8fd9\u6279\u8d44\u4ea7\u91cc\u7684\u542b\u4e49\u662f\u4ee3\u7801\u8dd1\u901a\uff0c\u4e0d\u662f\u56e0\u5b50\u6709\u6548\uff0c\u4e24\u8005\u9694\u7740\u4e00\u6574\u5957\u9a8c\u8bc1\u5de5\u4f5c\u3002<\/p>\n\n\n\n
\u7ed3\u8bba\u4e0d\u8981\u6284\u3002216 \u4e2a\u56e0\u5b50\u7684 IC \u4e2d\u4f4d\u6570 0.0018\uff0c\u8fd9\u4e2a\u6570\u5b57\u662f\u516c\u5f00\u7684\uff0c\u53ea\u662f\u6ca1\u4eba\u6c47\u603b\u8fc7\u3002\u6c47\u603b\u5b8c\u5c31\u4e0d\u8be5\u6307\u671b\u4ece\u91cc\u9762\u76f4\u63a5\u6311\u51fa\u80fd\u7528\u7684\u56e0\u5b50\u3002<\/p>\n\n\n\n
\u9a8c\u8bc1\u7684\u53ef\u4fe1\u5ea6\u4e0d\u8981\u9ed8\u8ba4\u63a5\u53d7\u3002\u4e00\u4efd\u62a5\u544a\u91cc\u51fa\u73b0\u7684\u6570\u5b57\uff0c\u548c\u4e00\u4efd\u80fd\u88ab\u590d\u7b97\u7684\u6570\u5b57\uff0c\u662f\u4e24\u79cd\u4e1c\u897f\u3002\u524d\u8005\u53ea\u9700\u8981\u4e00\u4e2a print \u8bed\u53e5\uff0c\u540e\u8005\u8981\u628a\u6570\u636e\u3001\u80a1\u7968\u6c60\u3001\u533a\u95f4\u3001\u6210\u672c\u8bbe\u5b9a\u4e00\u5e76\u4ea4\u51fa\u6765\u3002<\/p>\n\n\n\n
\u56de\u5230\u90a3\u4e2a\u661f\u6807\u6570\u5b57\u3002\u76ee\u5f55\u9875 1975 \u661f\uff0c\u4e09\u4e2a\u56e0\u5b50\u5e93\u52a0\u8d77\u6765 28 \u661f\u3002\u8fd9\u6279\u8d44\u4ea7\u91cc\u6700\u6709\u4ef7\u503c\u7684\u90e8\u5206\u4e0d\u662f\u90a3 800 \u4e2a\u56e0\u5b50\uff0c\u662f\u90a3\u5957”\u6bcf\u4e2a\u56e0\u5b50\u90fd\u8be5\u6709\u4e00\u4efd\u4ea4\u4ee3”\u7684\u5f62\u5f0f\u3002\u800c\u5b83\u6700\u7ecf\u4e0d\u8d77\u63a8\u6572\u7684\u90e8\u5206\uff0c\u4e5f\u662f\u8fd9\u4efd\u4ea4\u4ee3\u672c\u8eab\u3002<\/p>\n\n\n\n
\u53ef\u5ba1\u8ba1\u6027\u7684\u610f\u601d\u4e0d\u662f\u628a\u6570\u5b57\u516c\u5e03\u51fa\u6765\uff0c\u662f\u522b\u4eba\u80fd\u62ff\u7740\u4f60\u7684\u6570\u636e\u628a\u6570\u5b57\u91cd\u65b0\u7b97\u4e00\u904d\u3002\u505a\u4e0d\u5230\u540e\u534a\u53e5\uff0c\u524d\u534a\u53e5\u5c31\u53ea\u662f\u6392\u7248\u3002<\/p>\n\n\n\n
\u6587\u4e2d\u63d0\u5230\u7684\u4e09\u4e2a\u56e0\u5b50\u5e93\u5747\u4e3a GPLv3 \u5f00\u6e90\u9879\u76ee\u3002\u590d\u73b0\u811a\u672c\u548c\u6570\u636e\u533a\u95f4\u5df2\u5728\u6b63\u6587\u6ce8\u660e\uff0c\u4f60\u53ef\u4ee5\u7528\u540c\u6837\u65b9\u5f0f\u5728\u4f60\u81ea\u5df1\u7684\u80a1\u7968\u6c60\u4e0a\u91cd\u8dd1\u4e00\u904d\u3002<\/p>\n\n\n\n
\u9644\uff0c\u6587\u672b\u5b8c\u6574\u811a\u672c<\/strong><\/h2>\n\n\n\n
\u5b58\u6210 repro.py<\/code>\uff0c\u8dd1\u4e4b\u524d\u5148 pip install akshare numpy pandas<\/code>\u3002\u9996\u6b21\u8fd0\u884c\u4f1a\u8054\u7f51\u62c9\u53d6\u65e5\u7ebf\uff0c\u7ea6 15 \u5206\u949f\uff0c\u4e4b\u540e\u8d70\u672c\u5730\u7f13\u5b58\u3002\u5b83\u4f1a\u4f9d\u6b21\u590d\u7b97\u6587\u7ae0\u91cc\u90a3\u56db\u4e2a\u56e0\u5b50\uff0c\u76f4\u63a5\u6253\u51fa\u5bf9\u7167\u8868\u3002<\/p>\n\n\n\n
# -*- coding: utf-8 -*-\n\n\"\"\"\nQuantSkills \u56e0\u5b50\u5e93\u590d\u7b97\u811a\u672c\u3002\n\u7528\u6cd5\uff1a\n    python repro.py\n\n\u9ed8\u8ba4\u590d\u7b97\u6587\u7ae0\u91cc\u90a3\u56db\u4e2a\u56e0\u5b50\uff0c\u80a1\u7968\u6c60\u4e3a\u6caa\u6df1300 + \u4e2d\u8bc1500 \u6210\u5206\u80a1\u3002\n\u9996\u6b21\u8fd0\u884c\u4f1a\u8054\u7f51\u62c9\u53d6\u65e5\u7ebf\u6570\u636e\uff0c\u7ea6 15 \u5206\u949f\uff0c\u4e4b\u540e\u8d70\u672c\u5730\u7f13\u5b58\u3002\n\"\"\"\n\nimport importlib.util\nimport os\nimport sys\nimport time\n\nimport akshare as ak\nimport numpy as np\nimport pandas as pd\n\nWORK = os.path.dirname(os.path.abspath(__file__))\nDATA = os.path.join(WORK, \"data\")\nPRICES = os.path.join(DATA, \"prices\")\nFACTOR_ROOT = os.path.join(WORK, \"skill-quant-factor-volume-stat-alpha\", \"factors\")\n\nSTART = \"20211201\"\nEND = \"20260911\"\nHORIZON = 5\nCOST_PER_TURNOVER = 0.0030\nMIN_NAMES = 30\n\nTARGETS = [\n    \"R020-5d-z-scored-volume-expansion\",\n    \"R684-10d-volatility-scaled-obv-slope\",\n    \"R459-7d-stability-scaled-close-ts-rank\",\n    \"R617-10d-smoothed-price-volume-correlation\",\n]\n\nINDEXES = [(\"000300\", \"\u6caa\u6df1300\"), (\"000905\", \"\u4e2d\u8bc1500\")]\n\n\ndef get_universe():\n    \"\"\"\u53d6\u6caa\u6df1300\u548c\u4e2d\u8bc1500\u6210\u5206\u80a1\uff0c\u8fd4\u56de {\u4ee3\u7801: \u4ea4\u6613\u6240}\u3002\"\"\"\n    out = {}\n    for code in [i[0] for i in INDEXES]:\n        df = ak.index_stock_cons_csindex(symbol=code)\n        for _, row in df.iterrows():\n            c = str(row[\"\u6210\u5206\u5238\u4ee3\u7801\"]).zfill(6)\n            out.setdefault(c, str(row.get(\"\u4ea4\u6613\u6240\", \"\") or \"\"))\n    return out\n\n\ndef fetch_one(code, exch):\n    \"\"\"\u62c9\u5355\u53ea\u80a1\u7968\u7684\u524d\u590d\u6743\u65e5\u7ebf\uff0c\u843d\u5730\u6210 CSV\u3002\u5df2\u5b58\u5728\u5219\u8df3\u8fc7\u3002\"\"\"\n    path = os.path.join(PRICES, f\"{code}.csv\")\n    if os.path.exists(path) and os.path.getsize(path) > 500:\n        return\n\n    prefix = \"sh\" if (\"\u4e0a\u6d77\" in exch or code.startswith((\"6\", \"9\"))) else \"sz\"\n    for attempt in range(3):\n        try:\n            df = ak.stock_zh_a_daily(\n                symbol=prefix + code, start_date=START, end_date=END, adjust=\"qfq\"\n            )\n            if df is not None and len(df) >= 250:\n                df = df[[\"date\", \"open\", \"high\", \"low\", \"close\", \"volume\"]].copy()\n                df[\"symbol\"] = code\n                df.to_csv(path, index=False, encoding=\"utf-8-sig\")\n                return\n        except Exception:\n            time.sleep(0.5 * (attempt + 1))\n    print(f\"[warn] {code} \u62c9\u53d6\u5931\u8d25\uff0c\u8df3\u8fc7\")\n\n\ndef prepare_data():\n    os.makedirs(PRICES, exist_ok=True)\n    universe = get_universe()\n    print(f\"[data] \u80a1\u7968\u6c60 {len(universe)} \u53ea\uff0c\u5f00\u59cb\u62c9\u53d6...\")\n    for i, (code, exch) in enumerate(sorted(universe.items()), 1):\n        fetch_one(code, exch)\n        time.sleep(0.15)\n        if i % 100 == 0:\n            print(f\"[data] {i}\/{len(universe)}\")\n\n\ndef load_panel():\n    \"\"\"\u628a\u843d\u5730\u597d\u7684 CSV \u8bfb\u6210\u957f\u8868\u3002\"\"\"\n    frames = []\n    for name in sorted(os.listdir(PRICES)):\n        if not name.endswith(\".csv\"):\n            continue\n        code = name[:-4]\n        df = pd.read_csv(os.path.join(PRICES, name))\n        if len(df) < 250:\n            continue\n        df[\"symbol\"] = code\n        frames.append(df[[\"date\", \"open\", \"high\", \"low\", \"close\", \"volume\", \"symbol\"]])\n\n    panel = pd.concat(frames, ignore_index=True)\n    panel[\"date\"] = pd.to_datetime(panel[\"date\"])\n    for c in (\"open\", \"high\", \"low\", \"close\", \"volume\"):\n        panel[c] = pd.to_numeric(panel[c], errors=\"coerce\")\n    panel = panel.dropna(subset=[\"close\", \"volume\"])\n\n    print(f\"[panel] {panel['symbol'].nunique()} \u53ea\u80a1\u7968\uff0c{len(panel)} \u884c\uff0c\"\n          f\"{panel['date'].min().date()} \u81f3 {panel['date'].max().date()}\")\n    return panel\n\n\ndef load_factor(factor_dir):\n    \"\"\"\u52a8\u6001\u8f7d\u5165\u56e0\u5b50\u76ee\u5f55\u91cc\u7684 factor.py\u3002\"\"\"\n    path = os.path.join(FACTOR_ROOT, factor_dir, \"scripts\", \"factor.py\")\n    spec = importlib.util.spec_from_file_location(\"m_\" + factor_dir.split(\"-\")[0], path)\n    module = importlib.util.module_from_spec(spec)\n    sys.modules[spec.name] = module\n    spec.loader.exec_module(module)\n    return module\n\n\ndef evaluate(panel, col, horizon=HORIZON, cost=COST_PER_TURNOVER):\n    \"\"\"\u7b97 5 \u65e5 Rank IC\u3001ICIR\u3001t \u503c\u3001\u4e94\u5206\u7ec4\u591a\u7a7a\u4ef7\u5dee\u3001Top \u7ec4\u6362\u624b\u7387\u3001\u6263\u6210\u672c\u51c0\u989d\u3002\"\"\"\n    d = panel.sort_values([\"symbol\", \"date\"]).copy()\n    d[\"fwd\"] = d.groupby(\"symbol\")[\"close\"].shift(-horizon) \/ d[\"close\"] - 1\n\n    def one_day(g):\n        g = g.dropna(subset=[col, \"fwd\"])\n        if len(g) < MIN_NAMES or g[col].nunique() < 5:\n            return np.nan\n        return g[col].corr(g[\"fwd\"], method=\"spearman\")\n\n    ic = d.groupby(\"date\").apply(one_day, include_groups=False).dropna()\n\n    spreads, turnovers, prev = [], [], set()\n    for _, g in d.groupby(\"date\"):\n        g = g.dropna(subset=[col, \"fwd\"])\n        if len(g) < MIN_NAMES:\n            continue\n        q = pd.qcut(g[col].rank(method=\"first\"), 5, labels=False)\n        top, bot = g.loc[q == 4], g.loc[q == 0]\n        spreads.append(top[\"fwd\"].mean() - bot[\"fwd\"].mean())\n\n        cur = set(top[\"symbol\"])\n        if prev:\n            turnovers.append(1 - len(prev & cur) \/ len(prev))\n        prev = cur\n\n    ls = float(np.mean(spreads))\n    to = float(np.mean(turnovers))\n    std = float(ic.std(ddof=0))\n\n    return {\n        \"ic\": float(ic.mean()),\n        \"icir\": float(ic.mean() \/ std),\n        \"t\": float(ic.mean() \/ std * np.sqrt(len(ic))),\n        \"n\": int(len(ic)),\n        \"long_short\": ls,\n        \"turnover\": to,\n        \"net\": ls - to * cost,\n    }\n\n\ndef main():\n    prepare_data()\n    panel = load_panel()\n\n    rows = []\n    for folder in TARGETS:\n        mod = load_factor(folder)\n        out = mod.compute_factor(panel.copy())\n        col = mod.FACTOR_COLUMN\n        r = evaluate(out, col)\n        rows.append((mod.SPEC[\"factor_id\"], mod.SPEC[\"name_cn\"], r))\n        print(f\"{mod.SPEC['factor_id']}  IC={r['ic']:+.4f}  ICIR={r['icir']:+.3f}  \"\n              f\"t={r['t']:+.2f}  \u591a\u7a7a={r['long_short'] * 100:+.4f}%  \"\n              f\"\u6362\u624b={r['turnover']:.3f}  \u51c0\u989d={r['net'] * 100:+.4f}%\")\n\n    print(\"\\n| \u56e0\u5b50 | IC | ICIR | t | \u591a\u7a7a | \u6362\u624b | \u51c0\u989d |\")\n    print(\"|-|-|-|-|-|-|-|\")\n    for fid, name, r in rows:\n        print(f\"| {fid} {name} | {r['ic']:+.4f} | {r['icir']:+.3f} | {r['t']:+.2f} | \"\n              f\"{r['long_short'] * 100:+.4f}% | {r['turnover']:.3f} | {r['net'] * 100:+.4f}% |\")\n\n\nif __name__ == \"__main__\":\n    main()<\/code><\/pre>\n\n\n\n
\u8dd1\u5b8c\u8fd9\u5f20\u8868\uff0c\u5c31\u662f\u7b2c\u4e8c\u8282\u90a3\u5f20\u5bf9\u7167\u8868\u7684\u590d\u7b97\u4fa7\u3002\u628a TARGETS<\/code> \u6362\u6210\u4efb\u610f\u4e00\u4e2a\u56e0\u5b50\u76ee\u5f55\u540d\uff0c\u5c31\u80fd\u590d\u7b97\u90a3\u4e00\u4e2a\u3002<\/p>\n\n\n\n
\u4e00\u4e2a\u63d0\u9192<\/strong><\/h3>\n\n\n\n
\u4e0a\u9762\u811a\u672c\u91cc TARGETS<\/code> \u7528\u7684\u662f\u5b8c\u6574\u6587\u4ef6\u5939\u540d\u3002\u4ed3\u5e93\u91cc\u8fd9\u4e24\u767e\u591a\u4e2a\u6587\u4ef6\u5939\u7684\u547d\u540d\u662f”\u7f16\u53f7 + \u82f1\u6587\u77ed\u540d”\uff0c\u540c\u4e00\u4e2a\u57fa\u7840\u516c\u5f0f\u6362\u7a97\u53e3\u548c\u53d8\u6362\u4e4b\u540e\u77ed\u540d\u4f1a\u5f88\u50cf\uff0cR020<\/code> \u548c R021<\/code> \u53ea\u5dee\u4e00\u4e2a\u8bcd\u3002\u6284\u76ee\u5f55\u540d\u4e4b\u524d\u5148 ls factors\/<\/code> \u5bf9\u4e00\u904d\uff0c\u5199\u9519\u4e00\u4e2a\u8bcd\u5c31\u662f FileNotFoundError<\/code>\uff0c\u8fd9\u4e2a\u5751\u6211\u8e29\u8fc7\u3002<\/p>\n\n\n\n
\u5199\u5728\u6700\u540e<\/strong><\/h2>\n\n\n\n
\u6211\u60f3\u628a\u95ee\u9898\u8fd8\u7ed9\u505a\u56e0\u5b50\u7684\u4eba\uff0c\u4e5f\u5305\u62ec\u6211\u81ea\u5df1\u3002<\/p>\n\n\n\n
\u4f60\u7684\u56e0\u5b50\u5e93\u91cc\uff0c\u6709\u591a\u5c11\u4e2a\u662f\u4ee3\u7801\u4e00\u8dd1\u901a\u5c31\u5f53\u80fd\u7528\u4e86\u3002\u8fd9\u4e2d\u95f4\u90a3\u4e00\u6b65\uff0c\u4f60\u662f\u81ea\u5df1\u7b97\u8fc7 IC\uff0c\u8fd8\u662f\u770b\u4e86\u4e00\u773c\u522b\u4eba\u7ed9\u7684\u6570\u5b57\u3002<\/p>\n\n\n\n
\u8fd9\u4e2a\u95ee\u9898\u7684\u7b54\u6848\uff0c\u6bd4 800 \u8fd9\u4e2a\u6570\u5b57\u91cd\u8981\u5f97\u591a\u3002\u89c9\u5f97\u6709\u7528\u7684\u8bdd\uff0c\u628a\u8fd9\u7bc7\u8f6c\u7ed9\u540c\u6837\u5728\u641e\u91cf\u5316\u7684\u670b\u53cb\u3002\u6587\u672b\u5b8c\u6574\u811a\u672c\u53ef\u4ee5\u76f4\u63a5\u590d\u5236\u8fd0\u884c\uff0c\u9ed8\u8ba4\u8dd1\u6587\u7ae0\u91cc\u90a3\u56db\u4e2a\u56e0\u5b50\u3002<\/p>\n\n\n\n
\n\n\n\n
#\u91cf\u5316\u56e0\u5b50<\/a> #\u56e0\u5b50\u9a8c\u8bc1<\/a> #RankIC<\/a> #\u5f00\u6e90\u91cf\u5316<\/a> #QuantSkills<\/a> #AI\u91cf\u5316<\/a> #\u56e0\u5b50\u6316\u6398<\/a> #ICIR<\/a> #\u4ea4\u6613\u6210\u672c<\/a> #\u91cf\u5316\u56de\u6d4b<\/a> #\u91cf\u5316\u7814\u7a76<\/a> #A\u80a1<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"
\u4f5c \u8005\uff1a\u8001\u4f59\u635e\u9c7c \u539f\u521b\u4e0d\u6613\uff0c\u8f6c\u8f7d\u8bf7\u6807\u660e\u51fa\u5904\u53ca\u539f\u4f5c\u8005…<\/p>\n
Continue reading\u5f00\u6e90800\u4e2a\u91cf\u5316\u56e0\u5b50\uff0c\u4f46\u6ca1\u4eba\u544a\u8bc9\u4f60pass\u662f\u4ec0\u4e48\u610f\u601d<\/span><\/a><\/div>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[17,4,5,7],"class_list":["post-4910","post","type-post","status-publish","format-standard","hentry","category-aiinvest","tag-ai","tag-7","entry"],"_links":{"self":[{"href":"https:\/\/laoyulaoyu.com\/index.php\/wp-json\/wp\/v2\/posts\/4910","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/laoyulaoyu.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/laoyulaoyu.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/laoyulaoyu.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/laoyulaoyu.com\/index.php\/wp-json\/wp\/v2\/comments?post=4910"}],"version-history":[{"count":1,"href":"https:\/\/laoyulaoyu.com\/index.php\/wp-json\/wp\/v2\/posts\/4910\/revisions"}],"predecessor-version":[{"id":4918,"href":"https:\/\/laoyulaoyu.com\/index.php\/wp-json\/wp\/v2\/posts\/4910\/revisions\/4918"}],"wp:attachment":[{"href":"https:\/\/laoyulaoyu.com\/index.php\/wp-json\/wp\/v2\/media?parent=4910"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/laoyulaoyu.com\/index.php\/wp-json\/wp\/v2\/categories?post=4910"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/laoyulaoyu.com\/index.php\/wp-json\/wp\/v2\/tags?post=4910"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}