A forward test did not establish a recall forecast gain from channel fixation and encoding power in 13 SEEG patients under source trial order

This forward test did not establish better recall forecasts from channel fixation and encoding power in 13 SEEG patients. It used OpenNeuro ds006136 version 1.0.1, with 14 recordings and 2,134 matched trials. Source trial IDs were assumed to give task order. Models learned on 1,383 early trials and were tested on 751 later trials. Penalty selection and risk thresholds used only early forward folds. Channel beta and gamma power ended before the memory delay. The comparison controlled task position, earlier errors, and source channel availability. Mean Brier gain was +0.000959; its descriptive 95% patient interval was [-0.001075, +0.003547], with sign-flip p = 0.559570. Seven of 13 patients improved. Later flag fractions ranged from zero to one despite an early-set 80th-percentile threshold. The package preserves matched features, all forecasts, models, figures, source hashes, code, and retrieval instructions. All ten stable numerical products repeated exactly from source bytes. This is an exploratory follow-up on a reused cohort. Source artifact and channel selection are offline. The test does not establish a live trigger or any benefit from brain stimulation.

stereoelectroencephalography · working memory · recall forecasting · Brier score · power features · forward validation