Three pulses per group reduced 20–500 ms SEEG score variance by 41.7 percent across recorded choices in eight sessions

We tested six retained pulses in eight public SEEG sessions. The source was OpenNeuro ds004696 version 1.0.1. The test used 886 readout edges at 86 stimulation sites. The score compares early and later mean responses. At 20–500 ms, three early and three later pulses gave 41.7 percent lower variance than four early and two later pulses. The descriptive patient interval was 38.4 to 44.6 percent. All eight patients and 74 of 86 sites had a gain. These variances use all choices from two recorded pools of five pulses. They do not measure new-experiment error. The gain before stimulation was 38.8 percent. Under independent errors, equal covariance, stable means, and a fixed scale, the proof gives different gains. New pulses give a 33.3 percent reduction. Expected fixed-pool variances give a 42.5 percent reduction. Unequal noise can reverse the gain. The package preserves exact inputs, scores, code, figures, provenance, and replay instructions. Both fresh source rebuilds and all numerical checks passed. The result does not establish a clinical schedule or live BCI benefit.

stereoelectroencephalography · pulse allocation · variance · brain-computer interface · finite-pool sampling