Only your logged matches are used. Draws don’t count towards win probability. We start with your role record blended towards a neutral 50% baseline, blend in each hero’s record, then blend in that hero’s results on the selected map. Small samples are pulled towards the broader record so one lucky win doesn’t dominate. Recommendations require at least three wins or losses on a hero. A map without hero-specific games uses the hero’s broader history and is labelled accordingly.

This estimates past performance, not a guaranteed outcome. Team composition, opponents, ranks, balance patches and hero swaps aren’t captured. Log one primary hero per match. Different queues can be analysed separately.