This dataset accompanies the study “Kinetic Differences Between Orthodox and Southpaw Stances in Four Fundamental Boxing Punches.” It contains synchronized inertial measurement unit (IMU) and force plate recordings from 30 male boxers performing jab, cross, lead hook, and rear hook punches in both orthodox and southpaw stances. The dataset provides raw signals of fist acceleration and ground reaction forces during maximal punches, along with computational routines for filtering, peak detection, and event segmentation. Data were collected at Jan Długosz University, Poland, under controlled laboratory conditions. This resource enables researchers to reproduce the published analyses or perform alternative custom computations on stance-related performance differences. Data Files Raw Data (.xlsx) – Per-punch recordings with IMU and force plate values. Summary Tables (.xlsx) – Aggregated descriptive statistics across punch types and stances. Processing Code (.ipynb) – Python notebook implementing signal filtering, peak detection, and event extraction. Variables and Column Names Each raw file includes: Time – Sampling index (s). 1x, 1y, 1z – IMU acceleration along three axes (converted from milli-g to m/s²). fx, fy, fz – Ground reaction force components (N) from AMTI plate: fx – mediolateral force, fy – anteroposterior force, fz – vertical force. Derived variables created by the notebook: 1x_filtered, 1y_filtered, 1z_filtered – Bandpass filtered acceleration signals (25–250 Hz Butterworth, order 9). resultant_acceleration – Vector magnitude of filtered IMU acceleration. Total_GRF – Vector magnitude of ground reaction force. Data Processing Workflow The included Jupyter notebook (butterworth filtered-3.ipynb) automates computation in the following steps: Filtering: Applies a 9th-order Butterworth bandpass filter (25–250 Hz) to acceleration channels, removing noise and offset. Resultant Calculations: Computes resultant acceleration and total ground reaction force. Peak Detection: Identifies the five strongest strikes by ranking Total_GRF values above a 300 N threshold, ensuring a minimum 0.5 s separation. Window Extraction: For each peak, a 0.4 s window (±0.2 s around the peak) is exported into separate files for detailed event analysis. Summary Table: Generates a table with event time, peak force, and corresponding resultant acceleration values. Ethics All participants gave informed consent. The study was approved by the Human Subjects Research Committee of Jan Długosz University (KE-O/4/2022). Potential Applications Comparative stance analysis in boxing biomechanics Force–acceleration relationship modeling Signal processing benchmarking (filtering, event detection) Sports science teaching and reproducibility studies