Mouthguard sensors recorded 3,851 sports impacts and found peak linear acceleration predicted concussion better than rotational acceleration. The model placed 50% risk near 100 g, but only 47 impacts caused concussion, so independent datasets must test the proposed threshold.