Yeah... I question the efficacy of the methodology. A show that had nothing but perfect 10 episodes during its entire run wouldn't make this list. But a real data-driven way to assess this escapes me... It's the kind of question that would do better by survey, because elements a single number doesn't capture, like how memorable the episode was and how it impacted the series as a whole which the episode ranking does not capture are absolutely relevant to what makes a good finale. I guess the fairest data-driven way to measure a good finale would be to capture the general ranking of the series as a whole before the finale, and then after it, and the biggest shift would indicate the finale with the strongest impact, but this is still not a perfect estimate for "best".