AI boxing predictions with statistical edge

CompuBox tracks total punches thrown and landed per round, split into jabs and power shots. We compute landing percentage, punch output rate and the differential between a fighter's output and their opponent's for recent bouts. High landing percentage at high output correlates strongly with future performance, particularly in longer fights where stamina and accuracy compound.

Boxer performance typically peaks between 26 and 32 depending on style — punch resistance, hand speed and footwork all decline at different ages. A 38-year-old former world champion entering a fight against a 29-year-old contender carries a different risk profile than their record suggests. We model career-stage adjustment as a factor alongside recent performance.

Yes. Fighters returning after 12 months or more of inactivity often show ring rust — slower reflexes and reduced conditioning — particularly in their first few rounds. We track the gap since last fight and apply a cautionary signal for fighters returning from extended layoffs, especially into high-level competition.

Mismatched fights often produce very short odds on the favourite, leaving little value over fair probability once the bookmaker margin is stripped. Our model does not publish picks where the edge after margin-stripping does not clear the 2% post-tax threshold — many mismatches fall into this category regardless of which fighter is expected to win.

Yes. Every settled boxing pick is logged in the public track record on /track-record. The full history is available publicly — win rate, ROI and the edge that was published before each bout. Boxing has fewer picks per month than daily sports, so the sample-size chip will appear frequently to reflect the early-stage data status.