========================================================================================================================
SUITE4 EXPERIMENT 2 — COUNCIL OF SPECIALISTS + COMPETENCE MODEL
V2-COMPATIBLE COMPLETE OPERATIONAL EVALUATION ENABLED
========================================================================================================================

========================================================================================================================
EXP2 COMPETENCE COUNCIL | fold=1
[FORWARD EXPERTS] blocks=[53.0, 54.0, 56.0, 57.0, 59.0, 60.0]
  block=53 n=3,802 AP q/global/recent/rel=0.0533/0.0540/0.0440/0.1189
  block=54 n=3,683 AP q/global/recent/rel=0.5078/0.1549/0.2901/0.2239
  block=56 n=6,574 AP q/global/recent/rel=0.1391/0.0405/0.0269/0.1433
  block=57 n=4,785 AP q/global/recent/rel=0.6507/0.6575/0.6104/0.1071
  block=59 n=7,413 AP q/global/recent/rel=0.0686/0.0661/0.0226/0.0803
  block=60 n=5,347 AP q/global/recent/rel=0.1669/0.0322/0.1872/0.0173
[POLICY] alpha=0.75 temperature=2.00
committee                    AP/AUC=0.1112/0.7903 F2@20%=0.0772 P/R=0.0169/0.7143
competence_council           AP/AUC=0.1876/0.8842 F2@20%=0.0803 P/R=0.0176/0.7429 ΔAP=+0.0764
[COMPETENCE] target top1 accuracy=0.902 | NON_DEPLOYABLE best-expert oracle AP=0.9693
  history0     n=1,729 AP 0.0067->0.0074 weights=[0.256 0.254 0.253 0.237]
  history1_2   n=3,143 AP 0.3125->0.3615 weights=[0.253 0.251 0.25  0.246]
  history3plus n=2,521 AP 0.1726->0.1578 weights=[0.253 0.252 0.251 0.244]

========================================================================================================================
EXP2 COMPETENCE COUNCIL | fold=2
[FORWARD EXPERTS] blocks=[53.0, 54.0, 56.0, 57.0, 59.0, 60.0, 62.0]
  block=53 n=3,802 AP q/global/recent/rel=0.0564/0.0458/0.0424/0.0848
  block=54 n=3,683 AP q/global/recent/rel=0.5258/0.2911/0.1399/0.1400
  block=56 n=6,574 AP q/global/recent/rel=0.1682/0.0201/0.0552/0.0413
  block=57 n=4,785 AP q/global/recent/rel=0.6436/0.4468/0.3367/0.1151
  block=59 n=7,413 AP q/global/recent/rel=0.0716/0.0687/0.0130/0.0948
  block=60 n=5,347 AP q/global/recent/rel=0.1895/0.1315/0.1167/0.1883
  block=62 n=7,393 AP q/global/recent/rel=0.0976/0.1441/0.0780/0.1321
[POLICY] alpha=0.75 temperature=2.00
committee                    AP/AUC=0.0909/0.8927 F2@20%=0.2182 P/R=0.0555/0.8158
competence_council           AP/AUC=0.0960/0.8985 F2@20%=0.2252 P/R=0.0573/0.8421 ΔAP=+0.0051
[COMPETENCE] target top1 accuracy=0.817 | NON_DEPLOYABLE best-expert oracle AP=0.9444
  history0     n=609 AP 0.0732->0.1465 weights=[0.307 0.236 0.262 0.195]
  history1_2   n=1,663 AP 0.2588->0.2767 weights=[0.253 0.251 0.25  0.246]
  history3plus n=3,312 AP 0.0813->0.0671 weights=[0.253 0.251 0.25  0.246]

========================================================================================================================
EXP2 COMPETENCE COUNCIL | fold=3
[FORWARD EXPERTS] blocks=[53.0, 54.0, 56.0, 57.0, 59.0, 60.0, 62.0, 63.0]
  block=53 n=3,802 AP q/global/recent/rel=0.0538/0.0715/0.0533/0.0949
  block=54 n=3,683 AP q/global/recent/rel=0.5239/0.1339/0.0476/0.2136
  block=56 n=6,574 AP q/global/recent/rel=0.1222/0.0339/0.0880/0.1507
  block=57 n=4,785 AP q/global/recent/rel=0.6686/0.5688/0.6663/0.1700
  block=59 n=7,413 AP q/global/recent/rel=0.0835/0.0445/0.0832/0.0587
  block=60 n=5,347 AP q/global/recent/rel=0.1967/0.2955/0.2540/0.1041
  block=62 n=7,393 AP q/global/recent/rel=0.1087/0.1916/0.0198/0.1047
  block=63 n=5,584 AP q/global/recent/rel=0.0901/0.0784/0.0367/0.1789
[POLICY] alpha=1.00 temperature=2.00
committee                    AP/AUC=0.1204/0.8968 F2@20%=0.2561 P/R=0.0665/0.8932
competence_council           AP/AUC=0.0943/0.8696 F2@20%=0.2060 P/R=0.0535/0.7184 ΔAP=-0.0260
[COMPETENCE] target top1 accuracy=0.629 | NON_DEPLOYABLE best-expert oracle AP=0.9488
  history0     n=1,426 AP 0.0725->0.0597 weights=[0.313 0.305 0.257 0.125]
  history1_2   n=2,942 AP 0.1893->0.2154 weights=[0.257 0.254 0.248 0.241]
  history3plus n=2,552 AP 0.2562->0.2343 weights=[0.253 0.254 0.253 0.24 ]

========================================================================================================================
EXP2 COMPETENCE COUNCIL | fold=4
[FORWARD EXPERTS] blocks=[53.0, 54.0, 56.0, 57.0, 59.0, 60.0, 62.0, 63.0, 65.0]
  block=53 n=3,802 AP q/global/recent/rel=0.0594/0.0351/0.0152/0.0492
  block=54 n=3,683 AP q/global/recent/rel=0.4925/0.0469/0.1379/0.2051
  block=56 n=6,574 AP q/global/recent/rel=0.1525/0.0671/0.0380/0.0513
  block=57 n=4,785 AP q/global/recent/rel=0.6523/0.6618/0.5365/0.0922
  block=59 n=7,413 AP q/global/recent/rel=0.0614/0.0322/0.0276/0.1593
  block=60 n=5,347 AP q/global/recent/rel=0.2173/0.2706/0.1191/0.1493
  block=62 n=7,393 AP q/global/recent/rel=0.0798/0.1043/0.0443/0.0989
  block=63 n=5,584 AP q/global/recent/rel=0.0889/0.0831/0.0775/0.1806
  block=65 n=6,920 AP q/global/recent/rel=0.1096/0.0851/0.0979/0.0493
[POLICY] alpha=1.00 temperature=1.00
committee                    AP/AUC=0.4713/0.9579 F2@20%=0.3410 P/R=0.0960/0.9423
competence_council           AP/AUC=0.4447/0.9568 F2@20%=0.3410 P/R=0.0960/0.9423 ΔAP=-0.0266
[COMPETENCE] target top1 accuracy=0.535 | NON_DEPLOYABLE best-expert oracle AP=0.9777
  history0     n=168 AP 0.2860->0.2682 weights=[0.276 0.261 0.267 0.196]
  history1_2   n=1,687 AP 0.5124->0.5399 weights=[0.31  0.284 0.252 0.154]
  history3plus n=3,252 AP 0.4654->0.3470 weights=[0.272 0.271 0.266 0.192]

========================================================================================================================
EXP2 COMPETENCE COUNCIL | fold=5
[FORWARD EXPERTS] blocks=[53.0, 54.0, 56.0, 57.0, 59.0, 60.0, 62.0, 63.0, 65.0, 66.0]
  block=53 n=3,802 AP q/global/recent/rel=0.0585/0.0441/0.0324/0.1597
  block=54 n=3,683 AP q/global/recent/rel=0.5056/0.2598/0.2498/0.1359
  block=56 n=6,574 AP q/global/recent/rel=0.1363/0.0176/0.0579/0.0850
  block=57 n=4,785 AP q/global/recent/rel=0.6508/0.5601/0.5086/0.1472
  block=59 n=7,413 AP q/global/recent/rel=0.0865/0.0555/0.0219/0.2491
  block=60 n=5,347 AP q/global/recent/rel=0.2017/0.2666/0.0824/0.1236
  block=62 n=7,393 AP q/global/recent/rel=0.1000/0.1375/0.1206/0.1026
  block=63 n=5,584 AP q/global/recent/rel=0.0932/0.0909/0.0574/0.1882
  block=65 n=6,920 AP q/global/recent/rel=0.1172/0.0951/0.1165/0.0942
  block=66 n=5,107 AP q/global/recent/rel=0.4755/0.4553/0.3099/0.2220
[POLICY] alpha=0.50 temperature=2.00
committee                    AP/AUC=0.3319/0.9355 F2@20%=0.3492 P/R=0.1008/0.9097
competence_council           AP/AUC=0.3136/0.9302 F2@20%=0.3467 P/R=0.1001/0.9032 ΔAP=-0.0183
[COMPETENCE] target top1 accuracy=0.787 | NON_DEPLOYABLE best-expert oracle AP=0.9862
  history0     n=1,288 AP 0.2215->0.2552 weights=[0.273 0.266 0.263 0.197]
  history1_2   n=2,883 AP 0.4884->0.4743 weights=[0.279 0.272 0.272 0.177]
  history3plus n=2,824 AP 0.4111->0.4119 weights=[0.269 0.259 0.255 0.217]

========================================================================================================================
DEVELOPMENT-SELECTED OPERATIONAL POLICIES [beta=2]
Thresholds and warning ratios below are frozen before holdout evaluation.
========================================================================================================================
  competence_council  : AP=0.2296 | AUC=0.8996 | F2-opt tau=0.341621 | P=0.1741 R=0.4292 F1=0.2477 F2=0.3319 | flag=3.64% | ratio=2.465
  committee           : AP=0.2616 | AUC=0.9063 | F2-opt tau=0.155193 | P=0.2110 R=0.4630 F1=0.2899 F2=0.3737 | flag=3.24% | ratio=2.195

========================================================================================================================
EXP2 COMPETENCE COUNCIL | fold=6
[FORWARD EXPERTS] blocks=[53.0, 54.0, 56.0, 57.0, 59.0, 60.0, 62.0, 63.0, 65.0, 66.0, 68.0]
  block=53 n=3,802 AP q/global/recent/rel=0.0606/0.0395/0.0572/0.0282
  block=54 n=3,683 AP q/global/recent/rel=0.5096/0.1142/0.2864/0.1271
  block=56 n=6,574 AP q/global/recent/rel=0.1823/0.0352/0.0270/0.0485
  block=57 n=4,785 AP q/global/recent/rel=0.6446/0.4236/0.5010/0.1368
  block=59 n=7,413 AP q/global/recent/rel=0.0661/0.0532/0.0547/0.0301
  block=60 n=5,347 AP q/global/recent/rel=0.1921/0.1387/0.1624/0.1624
  block=62 n=7,393 AP q/global/recent/rel=0.0782/0.2216/0.0300/0.1813
  block=63 n=5,584 AP q/global/recent/rel=0.0898/0.0696/0.0664/0.1739
  block=65 n=6,920 AP q/global/recent/rel=0.1283/0.1104/0.1252/0.1010
  block=66 n=5,107 AP q/global/recent/rel=0.4763/0.4510/0.3675/0.2016
  block=68 n=6,995 AP q/global/recent/rel=0.3108/0.3385/0.2825/0.2467
[POLICY] alpha=0.25 temperature=2.00
committee                    AP/AUC=0.3394/0.8835 F2@20%=0.3992 P/R=0.1333/0.7964
competence_council           AP/AUC=0.3375/0.8901 F2@20%=0.3962 P/R=0.1323/0.7904 ΔAP=-0.0019
[COMPETENCE] target top1 accuracy=0.763 | NON_DEPLOYABLE best-expert oracle AP=0.9343
  history0     n=136 AP 0.3974->0.4009 weights=[0.292 0.245 0.245 0.217]
  history1_2   n=1,303 AP 0.4819->0.4794 weights=[0.276 0.263 0.266 0.195]
  history3plus n=3,551 AP 0.3068->0.3022 weights=[0.273 0.267 0.268 0.192]

========================================================================================================================
EXP2 COMPETENCE COUNCIL | fold=7
[FORWARD EXPERTS] blocks=[53.0, 54.0, 56.0, 57.0, 59.0, 60.0, 62.0, 63.0, 65.0, 66.0, 68.0, 69.0]
  block=53 n=3,802 AP q/global/recent/rel=0.0561/0.0843/0.0353/0.1027
  block=54 n=3,683 AP q/global/recent/rel=0.5358/0.0730/0.1052/0.1416
  block=56 n=6,574 AP q/global/recent/rel=0.1422/0.0907/0.0175/0.0563
  block=57 n=4,785 AP q/global/recent/rel=0.6627/0.4682/0.6505/0.1179
  block=59 n=7,413 AP q/global/recent/rel=0.0752/0.0156/0.1048/0.1239
  block=60 n=5,347 AP q/global/recent/rel=0.1642/0.3054/0.1814/0.2073
  block=62 n=7,393 AP q/global/recent/rel=0.0999/0.1998/0.0681/0.0818
  block=63 n=5,584 AP q/global/recent/rel=0.0867/0.0967/0.0404/0.1700
  block=65 n=6,920 AP q/global/recent/rel=0.1148/0.0987/0.1242/0.0478
  block=66 n=5,107 AP q/global/recent/rel=0.4708/0.4497/0.2637/0.1622
  block=68 n=6,995 AP q/global/recent/rel=0.3031/0.2715/0.2232/0.3152
  block=69 n=4,990 AP q/global/recent/rel=0.3263/0.3174/0.2648/0.1876
[POLICY] alpha=0.00 temperature=0.50
committee                    AP/AUC=0.1597/0.8197 F2@20%=0.3092 P/R=0.1022/0.6267
competence_council           AP/AUC=0.1597/0.8197 F2@20%=0.3092 P/R=0.1022/0.6267 ΔAP=+0.0000
[COMPETENCE] target top1 accuracy=0.692 | NON_DEPLOYABLE best-expert oracle AP=0.9911
  history0     n=1,329 AP 0.1301->0.1301 weights=[0.319 0.291 0.267 0.123]
  history1_2   n=2,552 AP 0.1942->0.1942 weights=[0.303 0.305 0.282 0.109]
  history3plus n=2,776 AP 0.2000->0.2000 weights=[0.301 0.31  0.284 0.106]

========================================================================================================================
SUITE4 ORIGINAL RANKING / EQUAL-BUDGET SUMMARY
========================================================================================================================
development NON_DEPLOYABLE_best_expert_oracle  pooled AP=0.9155 mean-fold AP=0.9653 F2@20=0.2816
development committee                          pooled AP=0.2616 mean-fold AP=0.2251 F2@20=0.2483
development competence_council                 pooled AP=0.2296 mean-fold AP=0.2273 F2@20=0.2398
holdout     NON_DEPLOYABLE_best_expert_oracle  pooled AP=0.9450 mean-fold AP=0.9627 F2@20=0.4973
holdout     committee                          pooled AP=0.2395 mean-fold AP=0.2495 F2@20=0.3542
holdout     competence_council                 pooled AP=0.2361 mean-fold AP=0.2486 F2@20=0.3527

### DEVELOPMENT [model=competence_council, beta=2]
  fold    test   base% |      AP     AUC |     P@t     R@t    F1@t    Fβ@t   rate@t |     P@r     R@r    F1@r    Fβ@r   rate@r
     1    7393   0.47% |  0.1876  0.8842 |  1.0000  0.0286  0.0556  0.0355    0.01% |  0.0632  0.4857  0.1118  0.2078    3.64%
     2    5584   1.36% |  0.0960  0.8985 |  0.0588  0.0132  0.0215  0.0156    0.30% |  0.1538  0.1316  0.1418  0.1355    1.16%
     3    6920   1.49% |  0.0943  0.8696 |  0.0983  0.3398  0.1525  0.2279    5.14% |  0.1121  0.2524  0.1552  0.2019    3.35%
     4    5107   2.04% |  0.4447  0.9568 |  0.1996  0.8846  0.3257  0.5245    9.03% |  0.3422  0.6154  0.4399  0.5307    3.66%
     5    6995   2.22% |  0.3136  0.9302 |  0.2236  0.4774  0.3045  0.3891    4.73% |  0.2108  0.4774  0.2925  0.3811    5.02%
------------------------------------------------------------------------------------------------------------------------------------
POOLED   31999   1.48% |  0.2296  0.8996 |  0.1741  0.4292  0.2477  0.3319    3.64% |  0.1730  0.4038  0.2422  0.3188    3.45%

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
HOLDOUT: pooled development threshold and warning ratio are frozen. No holdout label is used for threshold tuning.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

### HOLDOUT (test) [model=competence_council, beta=2]
  fold    test   base% |      AP     AUC |     P@t     R@t    F1@t    Fβ@t   rate@t |     P@r     R@r    F1@r    Fβ@r   rate@r
     6    4990   3.35% |  0.3375  0.8901 |  0.3143  0.3952  0.3501  0.3759    4.21% |  0.2747  0.4491  0.3409  0.3985    5.47%

  Confusion matrix — fold 6 — fixed development threshold 0.341621
  ------------------------------------------------------------------
  Actual / Predicted            Normal >=5      Failure <5
  ------------------------------------------------------------------
  Normal >=5                         4,679             144
  Failure <5                           101              66
  ------------------------------------------------------------------
  Normal correctly rejected: 97.01%
  Failures successfully detected: 39.52%

  Confusion matrix — fold 6 — drift-aware rate 5.462%
  ------------------------------------------------------------------
  Actual / Predicted            Normal >=5      Failure <5
  ------------------------------------------------------------------
  Normal >=5                         4,625             198
  Failure <5                            92              75
  ------------------------------------------------------------------
  Normal correctly rejected: 95.89%
  Failures successfully detected: 44.91%
     7    6657   3.26% |  0.1597  0.8197 |  0.3636  0.0553  0.0960  0.0666    0.50% |  0.1603  0.4055  0.2298  0.3105    8.25%

  Confusion matrix — fold 7 — fixed development threshold 0.341621
  ------------------------------------------------------------------
  Actual / Predicted            Normal >=5      Failure <5
  ------------------------------------------------------------------
  Normal >=5                         6,419              21
  Failure <5                           205              12
  ------------------------------------------------------------------
  Normal correctly rejected: 99.67%
  Failures successfully detected: 5.53%

  Confusion matrix — fold 7 — drift-aware rate 8.250%
  ------------------------------------------------------------------
  Actual / Predicted            Normal >=5      Failure <5
  ------------------------------------------------------------------
  Normal >=5                         5,979             461
  Failure <5                           129              88
  ------------------------------------------------------------------
  Normal correctly rejected: 92.84%
  Failures successfully detected: 40.55%
------------------------------------------------------------------------------------------------------------------------------------
POOLED   11647   3.30% |  0.2361  0.8451 |  0.3210  0.2031  0.2488  0.2192    2.09% |  0.1983  0.4245  0.2703  0.3456    7.06%

  Confusion matrix — POOLED — fixed development threshold
  ------------------------------------------------------------------
  Actual / Predicted            Normal >=5      Failure <5
  ------------------------------------------------------------------
  Normal >=5                        11,098             165
  Failure <5                           306              78
  ------------------------------------------------------------------
  Normal correctly rejected: 98.54%
  Failures successfully detected: 20.31%

  Confusion matrix — POOLED — drift-aware warning policy
  ------------------------------------------------------------------
  Actual / Predicted            Normal >=5      Failure <5
  ------------------------------------------------------------------
  Normal >=5                        10,604             659
  Failure <5                           221             163
  ------------------------------------------------------------------
  Normal correctly rejected: 94.15%
  Failures successfully detected: 42.45%

### DEVELOPMENT [model=committee, beta=2]
  fold    test   base% |      AP     AUC |     P@t     R@t    F1@t    Fβ@t   rate@t |     P@r     R@r    F1@r    Fβ@r   rate@r
     1    7393   0.47% |  0.1112  0.7903 |  1.0000  0.0286  0.0556  0.0355    0.01% |  0.0667  0.4571  0.1164  0.2105    3.25%
     2    5584   1.36% |  0.0909  0.8927 |  0.1019  0.2105  0.1373  0.1735    2.81% |  0.1207  0.0921  0.1045  0.0967    1.04%
     3    6920   1.49% |  0.1204  0.8968 |  0.1771  0.1650  0.1709  0.1673    1.39% |  0.1353  0.2718  0.1806  0.2262    2.99%
     4    5107   2.04% |  0.4713  0.9579 |  0.2456  0.8077  0.3767  0.5541    6.70% |  0.3832  0.6154  0.4723  0.5489    3.27%
     5    6995   2.22% |  0.3319  0.9355 |  0.2285  0.6516  0.3384  0.4755    6.32% |  0.2428  0.4903  0.3248  0.4073    4.47%
------------------------------------------------------------------------------------------------------------------------------------
POOLED   31999   1.48% |  0.2616  0.9063 |  0.2110  0.4630  0.2899  0.3737    3.24% |  0.1939  0.4038  0.2620  0.3319    3.08%

!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
HOLDOUT: pooled development threshold and warning ratio are frozen. No holdout label is used for threshold tuning.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

### HOLDOUT (test) [model=committee, beta=2]
  fold    test   base% |      AP     AUC |     P@t     R@t    F1@t    Fβ@t   rate@t |     P@r     R@r    F1@r    Fβ@r   rate@r
     6    4990   3.35% |  0.3394  0.8835 |  0.2459  0.5329  0.3365  0.4320    7.25% |  0.2922  0.4251  0.3463  0.3897    4.87%

  Confusion matrix — fold 6 — fixed development threshold 0.155193
  ------------------------------------------------------------------
  Actual / Predicted            Normal >=5      Failure <5
  ------------------------------------------------------------------
  Normal >=5                         4,550             273
  Failure <5                            78              89
  ------------------------------------------------------------------
  Normal correctly rejected: 94.34%
  Failures successfully detected: 53.29%

  Confusion matrix — fold 6 — drift-aware rate 4.863%
  ------------------------------------------------------------------
  Actual / Predicted            Normal >=5      Failure <5
  ------------------------------------------------------------------
  Normal >=5                         4,651             172
  Failure <5                            96              71
  ------------------------------------------------------------------
  Normal correctly rejected: 96.43%
  Failures successfully detected: 42.51%
     7    6657   3.26% |  0.1597  0.8197 |  0.2169  0.2949  0.2500  0.2752    4.43% |  0.1575  0.3548  0.2181  0.2837    7.35%

  Confusion matrix — fold 7 — fixed development threshold 0.155193
  ------------------------------------------------------------------
  Actual / Predicted            Normal >=5      Failure <5
  ------------------------------------------------------------------
  Normal >=5                         6,209             231
  Failure <5                           153              64
  ------------------------------------------------------------------
  Normal correctly rejected: 96.41%
  Failures successfully detected: 29.49%

  Confusion matrix — fold 7 — drift-aware rate 7.344%
  ------------------------------------------------------------------
  Actual / Predicted            Normal >=5      Failure <5
  ------------------------------------------------------------------
  Normal >=5                         6,028             412
  Failure <5                           140              77
  ------------------------------------------------------------------
  Normal correctly rejected: 93.60%
  Failures successfully detected: 35.48%
------------------------------------------------------------------------------------------------------------------------------------
POOLED   11647   3.30% |  0.2395  0.8439 |  0.2329  0.3984  0.2939  0.3488    5.64% |  0.2022  0.3854  0.2652  0.3263    6.28%

  Confusion matrix — POOLED — fixed development threshold
  ------------------------------------------------------------------
  Actual / Predicted            Normal >=5      Failure <5
  ------------------------------------------------------------------
  Normal >=5                        10,759             504
  Failure <5                           231             153
  ------------------------------------------------------------------
  Normal correctly rejected: 95.53%
  Failures successfully detected: 39.84%

  Confusion matrix — POOLED — drift-aware warning policy
  ------------------------------------------------------------------
  Actual / Predicted            Normal >=5      Failure <5
  ------------------------------------------------------------------
  Normal >=5                        10,679             584
  Failure <5                           236             148
  ------------------------------------------------------------------
  Normal correctly rejected: 94.81%
  Failures successfully detected: 38.54%
