Home page All episodes sorted by rank All episodes sorted alphabetically All episodes sorted by writer All episodes sorted by director All episodes sorted by season Episodes with the most polarized opinions Episodes largely forgotten by fans Best and worst streaks of episodes An analysis of what makes episodes great Angel writers sorted by the quality of their episodes Angel directors sorted by the quality of their episodes Seasons of Angel sorted by the quality of their episodes Angel characters sorted by their possible contribution to episode quality Predicted episode quality based on writer, director, characters, and when aired Episodes that are much better or much worse than predicted How the results on this site were obtained Frequently Asked Questions The links page Sitemap

Frequently Asked Questions

(or, Questions Anticipated to be Frequently Asked)

Although this page is called a "F.A.Q.," it is not really a F.A.Q. because it was written before the many of these questions were asked. Instead, a better name would be a "Q.A.F.A." (Questions Anticipated to be Frequently Asked).

How did you get the ratings?

Simply put, the webmaster searched the internet for lists of the best and worst episodes of Angel. He used these lists to create average ratings for each episode of the series. More detail is on the Method Page.

What do you mean by points for episodes?

The points for each episode represents its average ranking on best and/or worst episode lists, with 110 points meaning that the episode was rated the best episode on all lists and 1 point meaning that the episode was rated the worst episode on every list. Fans were not unanimous on any episode, so point totals for all episodes fall between these two extremes. Again, more detail is on the Method Page.

What do you mean when you say that something explains a certain percentage of episode quality? For example, what does it mean when you say that "only 1.9% of the quality of an episode can be attributed to the season when it aired"?

Suppose that someone wrote the titles to all 110 episodes on separate cards, drew one an random, and asked me to guess how many points that episode has. My best guess would be 55.5, the average score for all episodes. However, no episode has exactly 55.5 points. Some have more, and some have fewer. In this case, suppose that the person picked "Smile Time" with 76.16 points. I would be off by 20.66 points. These 20.66 points would be error due to the fact that "Smile Time" is a much better than average episode.

Suppose that the person drawing "Smile Time" gave me one hint, that it was a fifth season episode. Using this information, I would guess 58.10, the average score for fifth season episodes. Now I am off by only 18.06 points, a 12.6% improvement.

Sometimes, this information could be misleading. If the person drew "The Girl in Question" (with 49.66 points) instead of "Smile Time," knowing that the episode is a fifth season episode would make my guess worse, not better. That is, I would be off by 8.44 points instead of being off by 5.84 points, making my new guess 44.5% worse. Across all 110 episodes, however, my guesses would improve a little (1.9%) if I knew the seasons of each episode.

[Note to statisticians: I know that this concept is more complicated than I described. However, I would like to see you explain this concept to lay people in only three paragraphs without boring them.]

What does it mean when you say that you analyzed the effect of a director, actor, or character "after taking into account when the episode aired."

All this means is that the page looks at how much better we can predict the quality of an episode once we already know what season the episode aired in and whether the episode was the first, second-to-last, or the last episode of a season. This way, characters, writers, and directors are neither punished for not participating in good episodes nor rewarded for not participating in bad episodes if they were not part of the show for that season. This also takes into account the fact that season finales tend to be better than other episodes.

How is what you did for the Angel Phenomenon (and The Buffy Phenomenon) different from what you did for the Phi-Phenomenon?

The Phi-Phenomenon is a study of films and film tastes. There have been at least a few hundred thousand feature films made throughout the world. Many films have been lost forever and new ones are being made every day. Nobody has come close to seeing every film. It is not reasonable to include all feature films ever made in any analysis. Instead, only a sample of the best films can be studied. There have been only 110 episodes of Angel ever made, and it is very unlikely that there will be any more (possible movies do not count). This means that the entire population of Angel episodes can be analyzed, and lists of worst episodes are as useful as lists of best episodes. In fact, they are more useful because they are rarer.

Also, data analysis is much simpler with a population of a manageable size. Rather than the complicated points formula that the Phi-Phenomenon uses, the Angel Phenomenon can simply calculate the average ranking of an episode. Even if a list includes only the top-ten episodes, one can simply infer that the remaining 100 episodes are tied for eleventh.

Your statistical procedures are completely unwarranted. For example, you violate the assumptions of the [insert statistical jargon that would confuse and bore the ordinary reader]. How can you justify your procedure?

First, look at the answer above and note that I am dealing with the entire population of episodes, not a sample. I can make more definite conclusions when I am studying the entire population. Second, I am simply presenting rankings of episodes of a television show; I am not trying to land a rover on Mars. Precision is not crucial.

How can you possibly say that X is a great episode? It was awful.
How can you possibly say that Y is a bad episode? It was great.

The webmaster is not saying that X is a great episode or that Y is a bad episode, other fans are. Some episodes seem to divide fans. These are listed on the Polarized Episodes Page. Even well-liked or hated episodes have their critics or fans. Some people hate "I Will Remember You" and a few love "She." It is also possible that you missed something in episode X that others see or that you see something in episode Y that others do not.

If it is any consolation, the webmaster's #3 episode is barely in the top 50, and he is not so fond of some highly ranked episodes.

How confident are you in your rankings? Can you really be sure that episode #54 is better than episode #55?

Not very confident. A given rank means that we are at least 50% confident that the episode is better than any given lower ranked episode and weaker than any given higher ranked episode. On the Episode Ranking page, there is a column (labeled "+/-") indicating the margin of error for each episode. If the point difference between two episodes is less than the sum of the margins of error for the two episodes, then we are less than 95% confident that fans prefer the higher ranked episode.

Furthermore, this only addresses the margin of error that can be attributed to the fact that not every list on the internet is included in the analyses. This does not include error caused by other factors such as the fact that most lists do not rank all episodes or the possibility that fans who post lists on the internet have different opinions that fans who do not post lists.

I have noticed numerous errors in your listings of writers and directors. Don't you know that Joss Whedon ghostwrote parts of several episodes?

Only credited directors and writers are listed on this site. The Internet Movie DataBase is the final arbiter as to who is the credited writer and director for a given episode. This site ignores uncredited writers and directors.

The Phi-Phenomenon has found that there are three different tastes in film. Are there different tastes in Angel episodes such as preference for funny episodes versus dramatic, heartwrenching episodes or for early episodes versus later episodes?

Most of lists that this site analyzes skip episodes or mention only a handful of episodes. These lists do not lend themselves to an analysis of tastes in Angel episodes. An analysis was performed on a subsample of 78 lists from people who have seen all episodes and mention at least 10 episodes. The results suggest that there might be two tastes in Angel episodes, but those tastes are still quite ill-defined.

Will you do these analyses for other televison shows like Star Trek, The Simpsons, or Babylon 5?

There may be analyses of more shows if there is enough interest. Any analyses that take place will be only on long-running shows, so there will be none on Firefly unless a network picks it up for a few more seasons. Ideally, no analyses will be done on a series that is still airing new episodes, like The Simpsons, but they may be done over a summer hiatus if there is sufficient interest.

This page was last modified on August 23, 2004