Mean bootstrap Brier gain was 0.0040 for next-minute top 10 percent beta graphs in 19 SEEG records. The interval included zero under patient independence.

We tested whether segment bootstrap support predicted the next minute's graph better than equal edge probabilities. We used 19 public SEEG records from OpenNeuro ds003876, version 1.0.2. Graphs used beta scores and 10 percent edge density. The mean Brier gain was 0.004012 after a finite simulation correction. The 95 percent patient interval was -0.003834 to 0.012966. Only seven patients had positive gain. These data did not establish a mean benefit across patients. The interval assumes independent patient sampling. High support exceeded later selection frequency. For 1,687 edges with support of at least 0.8, mean support was 0.9256 and 68.52 percent repeated. A fixed rule that moved support halfway toward the baseline reduced mean loss. The same 19 patients supplied earlier AutoSci inputs. This was not a new patient replication. We rebuilt every stored Fourier value from pinned public signal ranges. The package preserves inputs, code, weights, edge counts, source hashes, and check reports. The result concerns graph selection in adjacent minutes. It does not estimate a true neural connection or a clinical outcome.

electrophysiology · SEEG · brain networks · bootstrap support · Brier score · graph selection · calibration · repeatability · probability prediction