Diagnostic transfert
Validation → Test
Objectif prioritaire: réduire les gros écarts négatifs, pas maximiser une Validation spectaculaire.
Runs économiques23avec signaux
Corr Val/Test-0.049plus haut = meilleur transfert
Risques de sur-sélection Validation
| Run | Arch | Val score | Test score | Gap | Worst Val | Worst Test | Seuil p | Diagnostic |
|---|---|---|---|---|---|---|---|---|
| run_0034_tcn_directional_adamw | tcn | 5.5619 | -70.6180 | -76.1798 | -3.38634 | -11.74080 | 70 | Val-overfit suspect |
| run_0049_tcn_directional_adamw | tcn | 6.0233 | -67.9209 | -73.9442 | -3.24185 | -11.60276 | 70 | Val-overfit suspect |
| run_0052_tcn_directional_adamw | tcn | 25.3432 | -48.3337 | -73.6769 | -0.92446 | -8.04724 | 70 | Val-overfit suspect |
| run_0055_tcn_directional_adamw | tcn | 26.9831 | -21.5358 | -48.5189 | -0.30748 | -8.13527 | 70 | Val-overfit suspect |
| run_0035_tcn_huber_adamw | tcn | 24.8437 | -22.7374 | -47.5811 | -0.06647 | -8.07493 | 80 | Val-overfit suspect |
| run_0042_cnn1d_pnl_adamw | cnn1d | 18.1429 | -23.6851 | -41.8280 | -2.90367 | -6.00086 | 70 | Val-overfit suspect |
| run_0038_tcn_directional_adamw | tcn | -0.0625 | -38.9064 | -38.8439 | -0.79731 | -8.27858 | 80 | Val-overfit suspect |
| run_0040_tcn_huber_adamw | tcn | 24.2932 | -10.6092 | -34.9023 | -0.77433 | -2.53041 | 70 | Val-overfit suspect |
| run_0046_tcn_huber_adamw | tcn | 20.3119 | -10.8614 | -31.1732 | -1.29727 | -4.63178 | 70 | Val-overfit suspect |
| run_0045_tcn_directional_adamw | tcn | 27.9405 | 0.1795 | -27.7611 | -0.00076 | -2.21381 | 80 | Val-overfit suspect |
| run_0039_tcn_huber_adamw | tcn | 26.5645 | -0.9783 | -27.5428 | -0.24354 | -3.49712 | 80 | Val-overfit suspect |
| run_0047_mlp_directional_adamw | mlp | 6.4690 | -6.3272 | -12.7962 | -0.61490 | -2.74823 | 80 | Val-overfit suspect |
| run_0048_mlp_directional_adamw | mlp | 6.4690 | -6.3272 | -12.7962 | -0.61490 | -2.74823 | 80 | Val-overfit suspect |
| run_0036_mlp_directional_adamw | mlp | 6.9279 | -1.9350 | -8.8628 | -0.37946 | -1.60743 | 70 | ok/à suivre |
| run_0037_mlp_directional_adamw | mlp | 6.9279 | -1.9350 | -8.8628 | -0.37946 | -1.60743 | 70 | ok/à suivre |
| run_0032_mlp_directional_adamw | mlp | 5.9080 | -2.0102 | -7.9182 | -0.00973 | -0.58466 | 70 | ok/à suivre |
| run_0033_mlp_directional_adamw | mlp | 5.9080 | -2.0102 | -7.9182 | -0.00973 | -0.58466 | 70 | ok/à suivre |
| run_0053_mlp_directional_adamw | mlp | 4.6296 | -1.5145 | -6.1441 | -0.23097 | -0.76794 | 70 | ok/à suivre |
| run_0054_mlp_directional_adamw | mlp | 4.6296 | -1.5145 | -6.1441 | -0.23097 | -0.76794 | 70 | ok/à suivre |
| run_0043_mlp_directional_rmsprop | mlp | 3.9022 | -2.1699 | -6.0721 | -0.36013 | -1.12967 | 80 | ok/à suivre |
| run_0044_mlp_directional_rmsprop | mlp | 3.9022 | -2.1699 | -6.0721 | -0.36013 | -1.12967 | 80 | ok/à suivre |
| run_0050_mlp_directional_adamw | mlp | 5.4737 | 2.7964 | -2.6773 | -0.29923 | -0.53784 | 70 | ok/à suivre |
| run_0051_mlp_directional_adamw | mlp | 5.4737 | 2.7964 | -2.6773 | -0.29923 | -0.53784 | 70 | ok/à suivre |
Règle de recherche
Les prochains runs doivent chercher à améliorer le Test économique médian et à réduire les gros Test négatifs via score pessimiste, no-trade Validation-only et modèles plus stables. Le Test reste interdit pour sélectionner un seuil ou un best.