In this work, we explore the application of a modified Recursive Least Squares (RLS) algorithm—referred to as degenerate RLS—for the detection of epileptic seizures in multichannel EEG signals from pediatric patients. The algorithm was tested using real-world data from the publicly available CHB-MIT Scalp EEG Database, recorded at Boston Children's Hospital. This dataset comprises long-term EEG recordings from individuals with drug-resistant epilepsy, offering a robust foundation for clinical and computational analysis. Our approach focuses on left fronto-parietal electrodes (FP1-F7, F7-T7, and T7-P7), applying unconventional filtering settings, particularly with a filter order of zero and a forgetting factor λ = 2. Despite the apparent simplicity of this configuration, the method achieved surprisingly high detection performance. We evaluated its effectiveness using standard classification metrics—precision, recall, F1-score, accuracy, and confusion matrix—with results surpassing 90% for several patients. These findings indicate that the degenerate RLS algorithm is a promising tool for low-complexity, real-time seizure detection systems, especially in resource-constrained or embedded environments where computational efficiency and responsiveness are critical.