No, that's not how it works. There is a chance that one of the first 3 candidates is the best one, but as the total number of candidates increases the probabili

No, that's not how it works. There is a chance that one of the first 3 candidates is the best one, but as the total number of candidates increases the probability of that being the case decreases. Different failure mode for this algorithm is when you get your candidates presented in the "worst to best" order - it makes you set your threshold low, and then settle for an average candidate.