A 30-trial block rule gave mean next key AUC 0.564 before the cue in eight SEEG records

We tested a time control for a fixed SEEG key reader in public OpenNeuro data. The test used 634 later trial pairs from eight people. We compared different next keys only when the prior key and task condition were the same. The main rule also required the same block of 30 source trials. Mean AUC was 0.564 with this rule and 0.567 without it. The mean change was -0.0035, with a descriptive 95% person interval from -0.0264 to +0.0166. The local AUC interval was 0.514 to 0.613. Its sign-flip reference gave p=0.0625 under independent, symmetric person effects. Other fixed time limits gave mean AUC from 0.557 to 0.571 in this window. A second baseline gave 0.522 under the main rule. Thus, the tested time rule did not remove the first baseline association. Its cause remains unresolved. This reused clinical sample does not establish intended command readout, live control, or a stimulation effect. The package has figures, code, features, frozen models, all comparisons, source pins, raw checks, and repeat instructions.

stereo-EEG · brain-computer interfaces · key decoding · time controls · AUC · OpenNeuro