样本：310 个因子，来自 hm100-hm105；每个因子 2424 个交易日（中位数），共 742135 个（因子,日）样本。IC>0 的因子占 44.2%。
|IC| 分位 [50,75,90,100] = [0.0361 0.0553 0.0694 0.0856]
分解残差 |IC - sum(c_d)|：均值 0.000423，最大 0.001336；相对 |IC| 均值的比例 1.08%

=== A 镜像对称性 ===
全局最大 |IC(-f) + IC(f)| = 0.0
逐日偏差 > 1e-12 的（因子,日）样本数 = 0 / 742135
每因子最大偏差取值集合 = [0.]
空组被跳过的因子 40 个，跳过天数合计 9056

=== B 前 50（按 |IC|）===
share: mean=0.4810 median=0.4577 std=0.0599 t=-2.25 p=2.92e-02 >0.5 的个数=21/50 25/75 分位=[0.4334 0.5453]
share_ic: mean=0.5565 median=0.5556 std=0.0264 t=15.15 p=4.04e-20 >0.5 的个数=49/50 25/75 分位=[0.5458 0.5711]
全样本 share_ic: mean=0.4173 median=0.5457 25/75=[0.4816 0.5862] >0.5 的个数=215/310
top50 中 IC<0 的个数 = 28
top50 归正后 lambda_short 中位数=1.0000 均值=0.9829
top50 归正后 lambda_long  中位数=1.0000 均值=0.8693
top50 归正后 lsh>llo 的个数 = 26/50
top50 raw lambda_short 中位数=-0.5243 raw lambda_long 中位数=-1.0000
全样本归正后 lambda_short 中位数=1.0000 lambda_long 中位数=1.0000 lsh>llo 个数=138/310

=== C 相关性 ===
corr(ic,lambda) pearson=0.9087 spearman=0.8728 n=310
corr(ic,ssm) pearson=0.8225 spearman=0.8334 n=310
corr(ic,gamma) pearson=0.9170 spearman=0.8765 n=310
corr(ic,c_s3) pearson=-0.1509 spearman=-0.1891 n=310
corr(ic,c_l3) pearson=0.0274 spearman=-0.0152 n=310
corr(absic,lam_or) pearson=0.5613 spearman=0.5183 n=310
corr(absic,ssm_or) pearson=0.5381 spearman=0.4892 n=310
corr(absic,c_s3) pearson=0.9373 spearman=0.9306 n=310
corr(absic,c_l3) pearson=0.8985 spearman=0.9013 n=310
corr(ic,share_ic) pearson=-0.0130 spearman=-0.2459 n=310
corr(absic,share_ic) pearson=0.0896 spearman=0.1366 n=310
corr(absic,ladder_ratio) pearson=0.1079 spearman=-0.0405 n=310

=== B' 阶梯占比（lead 补充）===
ladder_ratio 分位 [5,25,50,75,95] = [0.9782 0.9861 0.9888 0.9935 0.9994]
top50(|IC|) ladder_ratio: 中位数=0.9893 均值=0.9892 最小=0.9837 >0.9 的个数=50/50
top50 中 ladder_ratio>0.9 的子集（50 个）share_ic: 中位数=0.5556 均值=0.5565 >0.5 个数=49/50
top50 中 ladder_ratio<=0.9 的（0 个）share_ic: 中位数=nan 均值=nan
ssm3_or 分位 [5,25,50,75,95] = [-1.    -0.19   0.664  1.     1.   ]; top50 ssm3_or 中位数=1.0000 均值=0.8542
ssm_or 与 ssm3_or 差异：|ssm3-ssm| 中位数=0.0000 最大=0.0000; ssm3<ssm 的因子数=0
全样本 ladder_ratio<=0.5 的因子数=1；这些因子 share_ic 分位[5,50,95]=[-44.931 -44.931 -44.931]
median |c_short3| = 0.01910  median |c_long3| = 0.01578  median 比值(|c_s3|/|c_l3|) = 1.2755
median |c1|=0.01270 ... 各组 |c_d| 均值 = [0.0127  0.00588 0.00245 0.00082 0.00019 0.00029 0.00131 0.00297 0.00531
 0.00862]

=== D 组内离散度 ===
归正后各组 r5 组内 std 均值(bp) = [666.1 604.7 583.9 568.1 554.9 543.5 531.7 520.3 509.3 514.9]
归正后 空头组(第1组) - 多头组(第10组): 均值=151.2bp 中位数=142.3bp 空头更离散的因子数=285/310
配对 t = 24.23, p = 1.983e-73; 比值中位数 = 1.2722
未归正(raw)各组 std 均值(bp) = [583.6 551.7 549.1 548.3 548.8 549.6 551.4 555.1 562.3 597.4]; raw gsd1-gsd10 均值=-13.9 空头更离散个数=146
组内残差占比: mean(|resid|)/mean(|IC|) = 1.08%

=== E 例子 ===
-- IC 前 20%（|IC|>=0.0579）里 SSM 最差的 8 个 --
                                                                  factor      ic  ssm_or  lam_or  lsh_or  llo_or       r1       r2      r3      r4      r5      r6      r7      r8       r9      r10
              hm100/microcapm_price_log_return_3s_rolling_alpha_mean_p95 -0.0644 -1.0000  0.6826     1.0 -1.0000  15.3363  25.2225 29.1393 29.7332 30.2315 28.9380 24.6809 17.4328   3.6252 -48.7371
                                     hm101/bid_post15_obs_vs_mkt_mean_p5  0.0618 -0.6166  0.8093     1.0 -0.6166 -36.7089  -7.0953  7.2351 17.7444 24.8388 30.3113 31.5874 32.0298  30.4184  24.7835
 hm100/microcapm_book_imbalance10_innovation_capm_abs_residual_mean_mean -0.0669 -0.1574  0.8241     1.0 -0.1574  25.5564  31.8025 30.4983 30.2437 27.2554 24.0218 17.4844  7.8535  -6.1974 -32.9499
                      hm104/hotpool_ext_ind_rel_x15y10_ba_hot_rk_buy_p75  0.0598 -0.1516  0.6009     1.0 -0.1516 -31.1240  -9.3484  3.8524 12.6968 20.5427 25.5023 30.2371 34.9274  38.2403  20.9489
                           hm103/SEG3_T180_BIN16_TARGET_UNIQUE_RATIO_RAW  0.0620  0.1232  0.8955     1.0  0.1232 -52.0824 -10.3653  4.8958 16.5099 23.7435 29.5256 33.0565 35.3579  35.7262  30.8855
                                            hm105/x15y10_ba_hot_f28_mean -0.0597  0.1440  0.7961     1.0  0.1440  25.5660  31.7973 30.0513 25.6670 23.4696 17.7145 13.3163  4.1808  -6.1903 -23.0852
                             hm103/SEG1_T180_LOGVOL_TARGET_MAX_SHARE_RAW  0.0592  0.2894  0.9052     1.0  0.2894 -46.3004  -8.0569  6.6414 14.7242 21.4628 28.0282 30.9831 34.0064  35.4029  31.3383
hm100/microcapm_book_imbalance10_innovation_capm_cooks_distance_mean_p25 -0.0799  0.4623  0.9272     1.0  0.4623  33.2908  36.2067 34.7451 32.5164 28.2758 23.1352 15.8575  3.4173 -10.9445 -40.9482
-- IC 中等（40-60 分位，|IC| 在 0.0314~0.0422）里 SSM 最好的 6 个 --
                                                                                                    factor      ic  ssm_or  lam_or  lsh_or  llo_or       r1      r2      r3      r4      r5      r6      r7      r8      r9      r10
hm100/microcapm_order_gap_signed_vw_rolling_beta_mean_corr_microcapm_order_gap_magnitude_rolling_beta_mean  0.0314     1.0     1.0     1.0     1.0  -5.1297  2.1217  8.4842 13.8083 17.8956 21.1937 21.9205 23.7990 25.3980  29.4531
                                           hm100/microcapm_spread_bps_capm_residual_zscore_mean_entropy_1d  0.0320     1.0     1.0     1.0     1.0  -1.4305  2.7968  9.9667 13.9976 16.3225 19.2240 20.8995 23.6191 27.0758  29.8779
                                                                      hm101/ask_obs_pre_rank_autocorr1_abs -0.0359     1.0     1.0     1.0     1.0  31.8646 28.4781 26.5020 23.0439 19.5856 15.2933 10.1700  4.8527 -1.0283  -8.1313
                                                                   hm101/ask_post15_vol_rank_autocorr1_abs -0.0377     1.0     1.0     1.0     1.0  28.6553 28.2222 27.5936 24.2172 22.8597 19.3465 14.2660  9.4810  0.3913 -24.4223
                                                                   hm101/bid_post15_vol_rank_lz_complexity  0.0393     1.0     1.0     1.0     1.0 -16.1698 -0.7619  7.4693 13.7793 18.3475 21.7614 24.5296 26.5932 29.2188  31.2398
                                                      hm101/bid_post15_vol_vs_mkt_mean_corr_pre15_vol_rank -0.0418     1.0     1.0     1.0     1.0  31.6459 29.4785 28.6551 25.6315 24.9625 21.2172 16.3516 10.2291 -4.4520 -28.3947
ssm_or 分位 [5,25,50,75,95] = [-1.    -0.19   0.664  1.     1.   ]
lam_or 分位 [5,25,50,75,95] = [0.017 0.562 0.904 1.    1.   ]
gamma 分位 [5,25,50,75,95] = [-1.    -0.988 -0.388  0.964  1.   ]
