Introduction

Game of Thrones is a critically acclaimed television show that has captivated audiences around the world. Its complex and intricate story, richly developed characters, and stunning visuals have earned it numerous awards and accolades, making it one of the most successful and influential TV shows of all time. For this project, we used a data set which contains Game of Thrones episode information from Wikipedia, IMDb, Rotten Tomatoes, and Metacritic. The data set consists of 73 rows and 19 columns. Each row corresponds to a particular episode, and each column represents a different attribute describing these episodes. We briefly describe the variables here:

Research Questions

One of the key indicators of any TV show’s success is its ratings, which reflect the popularity and engagement of its audience. Naturally, this was a important metric of interest in our project. We formed three research questions surrounding this topic, listed as follows:

  1. How has the popularity of Game of Thrones changed over time?

  2. Is there an apparent relationship between the rating of an episode and other factors?

  3. Does there tend to be a consensus of episode quality between one rating platform and another? (IMDb, Rotten Tomatoes, Metacritic)

Research Question 1

We determine two primary ways to determine the popularity of a show, viewership and rating. We will first examine how the rate of viewership for a Game of Thrones episode has changed over time.

Game of Thrones is a pop culture sensation and has experienced monumental growth season after season. Viewership is a metric that indicates the popularity of the show, and we see that the shown has grown more and more popular over time.

A second metric of popularity is episode rating. The rating of a show speaks towards the quality and appeal to the viewers, as even shows with high viewership can still be seen in a negative light. Here, we plot the score from an assortment of media rating platforms against the episode number.

All three rating platforms follow a similar trend over time, with ratings dropping substantially near the end of the show. When measuring overall popularity, it is interesting to note the contrasting results in our two plots. Viewership was an at an all time high in the later episodes of the show, while rating was at an all time low. It may suggest that the show’s creators did not adequately fulfill the growing audiences’ expectations and deliver a satisfying conclusion to the story, resulting in a widely viewed, yet unpopular final portion of the show.

We can see as the show progressed, the popularity and ratings of each episode hovers around 90. But as the show reaches the point where there is no more source material starting at season 5 or episode number 41 we can see a trend as the shows popularity and rating decrease over time.

Research Question 2

The steep drop in ratings begs the question,

First, we evaluate the extent in which rating deviates over episodes. Note that we use an aggregated rating, which is an average of the rating platform scores.

We can see that the majority of episodes have an above average rating, which we have compiled from each rating website, with a relatively high average rating of 86. There are a few outliers such as the episodes “The Iron Throne”, “The Last of the Starks,” and “The Bells” which are all very below average and all from the last season. We can deduce that there is something going on with certain episodes that deviate from the mean either negative with those three episode or positive with “The Rains of Castamere” with an average rating of 99.

Next, we expand upon our previous graph by incorporating the season of the show into our analysis.

Over the first four seasons of Game of Thrones, rating is impressively consistent. A slow decline is observed in seasons 5-7, followed by a steep decline in season 8. It is clear that, at least for later seasons, there is a distinct difference in quality of the show between seasons. To verify this conclusion, included is R output (and an interpretation) of a linear regression model, where we fit average rating to the seasons of Game of Thrones:

## 
## Call:
## lm(formula = avgRating ~ factor(Season), data = data)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -19.667  -2.286   0.400   3.400  17.333 
## 
## Coefficients:
##                 Estimate Std. Error t value Pr(>|t|)    
## (Intercept)       92.600      2.024  45.743  < 2e-16 ***
## factor(Season)2   -2.000      2.863  -0.699 0.487291    
## factor(Season)3   -2.300      2.863  -0.803 0.424674    
## factor(Season)4    0.200      2.863   0.070 0.944519    
## factor(Season)5   -6.000      2.863  -2.096 0.039997 *  
## factor(Season)6   -9.600      2.863  -3.353 0.001335 ** 
## factor(Season)7  -12.314      3.155  -3.903 0.000228 ***
## factor(Season)8  -34.933      3.306 -10.568 9.47e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 6.402 on 65 degrees of freedom
## Multiple R-squared:  0.7071, Adjusted R-squared:  0.6756 
## F-statistic: 22.42 on 7 and 65 DF,  p-value: 4.155e-15

We notice significant negative coefficients for seasons 5, 6, 7, and 8, meaning that we have evidence that the decline in rating associated with the later seasons of the show compared to that of season 1 is significant.

Next, we explore the impact of the writers on the ratings of the show.

There are noticeable differences in distribution of show ratings between each writer. as an example, we see that for the episodes that George R. R. Martin (the author of the book series) is credited as writing, his ratings are the highest across all quartiles compared to other writers. Using this conclusion, it appears that the writer of the episode is a possible contribution to the rating of the episode.

From previous analysis, we observed that ratings were consistently high between seasons 1-4, and experienced a slow decline across seasons 5-8. The plot may be a contributor to this decline, so below we examine the most frequent words from the synopses of the first half of the show, and then from the second half of the show.

Most frequent words in synopses from seasons 1-4 (Left) and seasons 5-8 (Right)

This plot highlights the differences in frequency of characters, actions, locations, and events throughout the synopses of early Game of Thrones episodes and then of later Game of Thrones episodes. This is relevant to ratings of the show because there was a notable difference between the ratings in the first half and the ratings in the second half of the show. In the second half, we see Jon and Daeneris rise in prevelence over the other characters, and decline in Tyrion and the Lannister crest. Common actions, such as attack and watch, are are more heavily used in the early synopses compared to the later synopses. This is an interesting indicator of changes in characters and plotlines, which may influence the ratings of the show.

Diving further into analysis of the episode synopses, we are also able to pull positive and negative sentiments from the text. Below, we examine positively-connotated words from the early seasons vs positively-connotated words from later seasons, as well as negatively-connotated words from the early seasons vs negatively-connotated words from later seasons.

It is apparent that there are different themes being played out across the earlier seasons and the later seasons, both in terms of the positive aspects and negative aspects. There is little overlap in terms of the most frequent positive words in episode synopses in seasons 1-4 compared to the most frequent positive words in seasons 5-8, nor in the two sets of words on the negative side. When we consider that seasons 1-4 are rated higher, it is possible that the show is of better quality when dealing with themes of magic and freedom on the good side, rather than faith and support. The same claim could be made for words with negative connotations, where the show may be of better quality (or at least rated higher) when incorporating plots with fear, attacks, and prison, compared to topics of overwhelming actions, the dead, and ambushes.

Research Question 3

In parts of our analysis, we were relying on an average score between the media rating platforms. The reality is these scores may need to be considered with more nuance, as they are collected, formulated, and interpreted in slightly different ways.

To answer this question, we will first step away from individual episodes and examine how the scores from each platform are distributed across episodes of Game of Thrones.

We see that between the rating platforms, there are differences in distribution of rating. Ratings from all platforms are unimodal and have a long tail to the left, as it is uncommon for an episode of Game of Thrones to receive a low score. IMDb is the most bell shaped, and is centered around 85 with decreasing occurrences of ratings to either side. Rotten Tomatoes gives out high scores with the most frequency, with a mode at above 95 and a steadily dropping tail to the left. Lastly, we see that metacritic has lower densities of the higher values, while the low values to the left are at a higher density, indicating that Metacritic may be a harsher critic than the other two platforms.

We hone in on this comparison by revisiting our graph from question 1, which splits rating by individual episode and platform over time. This time, we overlay standard error ranges on the fit curves.

For the first 40 episodes, the standard error ranges overlap, meaning that we cannot say with confidence that the platforms are rating differently. However, as the episodes advance, Metacritic assigns far lower scores for each episode than IMDb and Rotten Tomatoes. This implies that Metacritic is overly critical compared to the other two platforms, which also see a decrease in ratings over time, but continue to overlap in error ranges throughout.

We can see that the difference between the three rating websites are relatively minimal with most dots clustering near each other in the top right. The dots are labelled using the average rating. There are a few outliers such as the three dots near the bottom left which are the last three episodes and we can see that they are very low on x axis and y axis and as such we can see that Metacritic rated the last three episodes the lowest out of the three rating websites which might be because Metacritic scores are user submitted and not professional critics. There are also some dots in between the large cluster and the last three episodes which does indicate that there is some disparity among the rating websites although we can see that most episodes are rated similarly with a slightly more critical evaluation from Metacritic.

Conclusion

Game of Thrones has consistently received high ratings, with many of its episodes ranking among the most-watched in television history. However, the ratings for the show’s final season saw a significant drop compared to previous seasons, sparking debate and speculation among fans and industry insiders. The possible implications of this drop in ratings are significant and wide-ranging. For one, it may suggest that the show’s creators did not adequately fulfill audience expectations and deliver a satisfying conclusion to the story. This could damage the show’s legacy and reputation, and potentially harm the future success of its creators and cast. We find that viewership steadily increased over time, while ratings began to drop near the end of the show’s lifetime, clouding the popularity of the legendary show. The most evident predictor of rating was the season of the show, but other factors, such as the writer of the episode and the language used in the plot synopses, changed alongside rating and were likely related in some extent to the decline of the show’s critical response. Lastly, we looked at the differences between the how IMDb, Rotten Tomatoes, and Metacritic rate Game of Thrones, and concluded that Metacritic was a more critical platform than its peers when it came to evaluating the show.

The importance of Game of Thrones and its episode ratings cannot be overstated. The show’s lasting impact on the television industry and its fans is undeniable, and the potential implications of a drop in ratings should not be ignored. It is therefore our mission to seek out the factors that made the show a smashing success and the factors that disappointed fans and the industry as a whole, so that future shows may learn from both the successes and failures of Game of Thrones. This topic could be further explored with a more expansive data set, containing more detailed records such as cast, plot, script, and budget of each episode to better theorize went went wrong in the later seasons.