when scaling user-generated sorting mechanisms, a basic arithmetic mean (sum of stars / total votes) critically fails. for instance, a newly listed bot or channel on telegramic with a single 5-star review would mathematically outrank a heavily used asset with one hundred 5-star reviews and two 1-star reviews.

to eliminate this sorting bias, the rating engine for telegramic was migrated to a bayesian average algorithm.

the bayesian approach introduces a "prior" (a baseline belief) that pulls items with very few votes toward the global average. as an item accumulates more empirical data (votes), the algorithm trusts the actual user ratings over the prior.

this ensures that a highly rated item with significant statistical volume will always correctly outrank an item with a perfect score but zero statistical weight. the legacy arithmetic mean remains visible on the frontend to prevent user confusion, but the backend sorting pipeline is strictly driven by the bayesian weight.