Contact order changed half-strength ordinal prediction gain in paired 1–25 Hz SEEG from 36 epilepsy patients

We tested whether bipolar contact order changes prediction gain for a three-point ordinal SEEG feature. We used two 60 s records for each of 36 epilepsy patients in OpenNeuro ds004100 version 1.1.3. Each selected shaft kept the same contact pair and sign in both records. The predictor used the source mean plus 0.5 times the source departure from that mean. We enumerated all 683,268 relative sign choices across patients. The equal-patient mean gain ranged from -13.26% to 19.01%. The original order gave 9.62%. Gain changed sign in 24 of 36 patients. These finite bounds use both records. They are not prediction rules or confidence limits. The uniform sign mean remained positive at 8.65%. A proof shows why the source mean can change while raw direction agreement remains unchanged. A scalar feature that is unchanged by contact order gave 1.75% mean gain. Its descriptive 95% patient interval included zero. All source hashes, direct loss checks, and a complete analysis rerun passed. Fresh extraction matched stored signals for eight files at four native sample rates. The package preserves code, signals, control counts, complete sign results, source pins, and reproduction commands. The study tests a feature benchmark. It does not establish a neural mechanism or clinical use.

SEEG · stereoelectroencephalography · ordinal patterns · bipolar polarity · contact order · signal prediction · time reversibility · OpenNeuro ds004100 · epilepsy