Wellbore Genius
DFIT benchmarks
Site DFIT scoring
Paste site samples, pick a benchmark entry, and score the live-kernel prediction. Scoring uses the canonical fixture inputs (v1 scaffold — site-tuned inputs land in a follow-up).
3 samples
Accuracy metrics
FAIL
N compared [—]
3
rms log10 [—]
0.8790
mean |log10| [—]
0.8771
max |log10| [—]
0.9320
bias log10 [—]
0.8771
Pass when rms_log10 ≤ 0.13. Bias > 0 ⇒ kernel over-predicts the site; < 0 ⇒ under-predicts.
Per-sample residuals
| t_sec | observed | predicted (kernel) | analytical | delta | log10(pred/obs) |
|---|---|---|---|---|---|
| 1.000 | 4.1000e-4 | 0.002564 | 0.002973 | 0.002154 | 0.7961 |
| 1.500 | 4.8000e-4 | 0.003841 | 0.003404 | 0.003361 | 0.9032 |
| 2.000 | 5.4000e-4 | 0.004617 | 0.003746 | 0.004077 | 0.9320 |
Validation history [—]
Saved runs appear here. Click "Save run" above to capture the current accuracy metrics, parsed samples, parse issues, and scoring grid for later re-download as CSV/PDF. Use Import JSON to restore a history file exported from another browser or machine.