10 follows out of 1,000 is noise. 10 follows out of 12 is a strong signal. Without normalizing by your network size, you can't tell the difference.
And follow counts ignore mutes entirely — which means high-volume posters and well-connected bad actors score well simply by being active.
Bayesian scoring asks the right question: given everyone in my network who has an opinion on this account, what fraction think positively? Follows are evidence for. Mutes are evidence against. The ratio is the score.
One catch: ratios break on thin data. One mute, one follow looks identical to 500 mutes and 500 follows. The prior fixes this — it anchors weak evidence toward neutral rather than letting small samples produce extreme scores.
Without Bayesian logic, WoT is either a follower count or a coin flip. Neither is trust.