Continuous ECG monitoring on wearables is constrained by the high energy and bandwidth costs of native-rate sampling (250–360 Hz). We introduce Sparse Min-Max Sampling, a pre-processing pipeline that retains only local amplitude extrema and their precise temporal offsets within signal blocks. This approach preserves native-rate timing fidelity while enabling the standard Pan-Tompkins algorithm to operate at low effective rates. Across three benchmarks (MIT-BIH, QT, and Noise Stress Test databases), our method matches or exceeds native-rate accuracy down to 10 Hz. On MIT-BIH, the 10 Hz configuration achieves an F1 score of 99.40%, surpassing the 99.19% baseline of the full 360 Hz signal, while reducing computational load by 10.5×. When combined with a specialized 9-bit amplitude + 7-bit position codec and LZ4 compression, the pipeline achieves a 38:1 data reduction (2.6% of original size) while exceeding the baseline F1 by +0.12% points. In noisy conditions (NSTDB), the 10 Hz pipeline improves F1 from 87.38% to 88.50% by leveraging sparse extrema selection to suppress sub-extremal noise artifacts. Sub-20 Hz sparse sampling with exact offset preservation is not only feasible but superior to traditional native-rate methods, simultaneously delivering higher detection accuracy, 10× faster processing, and 38× data compression for resource-constrained cardiac monitoring.