This site contains a meta-analysis of MST3K episode ratings. I've identified 10 different sources of rankings and combined them all to create a "Community Score" for each episode.

If you know of a ranking of episodes not listed here, let me know and I'll add to the data.

Sources

Some of these sites are no longer live. However, I've kept the URLs here for reference purposes.

Site Name	URL
MST3K Discussion Board	http://forrestcrow.proboards.com/thread/20122
Mr. Atari	http://forrestcrow.proboards.com/thread/13633/ataris-word-reviews
IMDB	http://www.imdb.com/title/tt0094517/epdate
Geos.tv	http://geos.tv/index.php/index?sid=216
Andrew's MST3K Site	http://www.magicmarkerweb.com/mst3kreview/
Kernunrex	http://kernunrex.blogspot.com/2006/01/mystery-science-theater-3000.html
Lefty's House O' MST	http://svamcentral.org/mst/
Mighty Jack	http://community-2.webtv.net/Mason85/MIGHTYJACKSMST3K/
MST3K Review	http://mst3k-review.home.mindspring.com/
Rhino Survey	http://www.weathergraphics.com/tim/mst3k.htm
Paste Magazine	https://www.pastemagazine.com/articles/2017/03/ranking-every-mst3k-episode-from-worst-to-best.html

Methodology

1. I dug around for any publically-accessible listings of MST3K episodes. Some of these lists are from individuals who reviewed each episode, and others are online polls with hundreds of respondents. A few of these lists are no longer available; I pulled them before they dissapeared from the Internet.

2. People used a variety of scales when ranking episodes. Some reviewers gave each episode a letter grade, others gave a rating out of 5 stars, and some used a numerical scale from 1-10. To make things consistent for the meta-analysis, I mathematically normalized each list to a 1-10 scale. A+ episodes became a perfect 10, 4-star (out of 5) episodes were now an 8, and for those lists where the lowest-ranked episode was a 2 -- those 2s were now 1s.

3. In pulling together this data, I discovered that reviewers naturally tended to skew their ratings high. After converting everything to a 1-10 scale, the average episode rating for most lists was around 7. While I agree that most episodes are pretty darn good, when comparing epiodes against each other, the ideal distribution should follow a bell curve. The average score across all episodes should fall right in the middle (at 5.5), with approximately the same number of 1's as 10's, 2's as 9's, and so on. This meant I needed to adjust each lists' rankings to correct for the skew.

For the mathematically-minded of you out there, I combined a Rank Transformation with an Empirical Cumulative Distribution Function to adjust the data to a specified mean. For those that aren't math majors, I just skewed each person's rankings so their average score was now 5.5.

4. Now I had a normalized table of ratings from 9 different sources. I then dropped the highest and lowest score that each episode received to remove any outliers, and averaged the remaining scores to get an overall "Community Score".

Data

You can download the Excel file I used to store the data here.