Unveiling Trends in Movie Remakes

Introduction

You are a film director working for a well-known company - “Lausanne pictures”. It’s been a while since you shot your last movie that took people’s breaths away and you wish to get back into the spotlight without “dropping the ball”. But …. you are out of original ideas. “What if I take an already existing storyline but make it bigger, better?”, you think to yourself. “But then, what movie should I choose to reenact so that success is inevitable?”. Tough question. And here is where our team comes to assist you. MADAYAM is about to unravel the MAD secrets of the remake industry "which movie should you remake? Will your remake match or even surpass the success of the original? What characteristics of a movie increase its likelihood of being remade?" - these are the questions we will answer for you based on the data across years of evolution and development of film industry. Are you ready for the most comprehensive analysis involving the most popular movie datasets?

Meet Our Team

(randomized order)

Aryan Ahadinia

Master in Computer Science

Amirmahdi Hosseinabadi

Master in Financial Engineering

Matin Ansaripour

Master in Computer Science

Sabri Yigit Arslan

Master in Computer Science

Maria Yuffa

Master in Neuro-X

Remakes and Movies Through Chronus

Let’s dive deeper into the world of films and remakes. First, you might as a question “How many remakes have been made throughout the filmographic history? Is there a difference in the number of films made across the decades? How about remakes?” All of these are valid questions. And let me tell you, there is a difference. While there are more remakes and more films made each year, the number of original movies are worth re-capturing seems to decline. Well, at least you are not tricking yourself by choosing to re-shoot the film instead of trying to aimlessly search for worthy new ideas. This is different with remakes and the whole dataset: more movies are made - more remakes are produced. Cooking some statistics to see how close our power law fits the log-log distributions, we observe that R2 metric for the whole dataset is 0.86 suggesting that the fitted power law distribution explains the data well. This is not the case for the remakes and originals with only 0.58 and 0.18 respectively.

It is important to note that a movie being shot before 1980 had a higher probability to be remade. This is a good point! We don’t want the recreated movies to be based on the films released recently. We want to have the nostalgic factor to appeal to our audience and make the movie a success!

Finally, we can notice that the mean number of remakes across the cinematic history is much more recent than for the whole dataset. Understandably, before you shoot a remake you have to shoot an original.

How Long Should Our Creative Struggle Take? Or Runtimes

Well, ok, we have more remakes. But it is not about quantity but quality of those, mes amis, that we are interested in. To start us off, we should constrain our movie to a reasonable time. Compared to the whole dataset, there are not so many outliers and there is not a single remake exceeding 200 minutes.

People and Their Opinions

We want your movie to be great. But to know what “great” actually means, we first need to take a look at what people’s opinions are on different movies, be it an original movie, a remake or any other. Let’s now take a look at the trends in the vote scores.

As we would have predicted anyways, vote scores for originals are shifted to the right meaning that movies that really resonated with the audience were chosen to be remade. This perfectly aligns with the regression analysis since the vote average significantly affected the movie to be remade.

Rather surprisingly, the remakes didn’t seem to keep that sparkle and performed similarly to the rest of the movies. Well, something to think about.

The remakes appeared to generate significant buzz, receiving a higher average number of votes compared to the rest of the dataset. Interestingly, the originals maintained a slight edge in average votes over both remakes and other films. This trend could suggest that fans of the original movies felt compelled to share their opinions on the recreated versions of the beloved films.

Bombastic Genre Combo

Now, we’ve thought a lot about data generally and have made up our mind about some technicalities, but what about the movie itself? Let’s think holistically for now and consider a possible genre combination between the original movie and the remake.

Uff, some tenners do seem appealing. But despite the success of the combination, we should approach this with some skepticism. Specifically, there might be only few votes for those compared to the ones with a lower rating. Overall, it is not just about the score, it is also about the wide public reach and the urgent need of the viewers to share their opinion. Plus, it is simply fairer when the votes are given by a variety of people. This also applies to the case where there are more votes just because there are more movies. To account for all that, we average the vote counts across movies and apply logarithm scaling to make the plot descriptive. After checking the vote counts we see that for some genre combinations the value is indeed low. This coincides with the tenners that we observe, tough luck. Hence, we multiply vote averages by the mean logarithm of vote counts to get a better look.

Seems like you can’t really go wrong with Comedy, Mystery and World cinema.

You might ask, what genres were typically remade? The movies were likely to be recreated if their main genre was Thriller, Romance or Comedy which constitutes one of the most prevalent genres. Hence, for an original being either of those three was significant to being remade. That was also true for Black-and-White movies, which is easy to explain, since Black-And-White movies are typically old movies (of course some masterpieces) that were waiting for technological development to be shot in full glory. Therefore, going for one of these options seems like a safe choice. Well, for now that seems fair.

Laughing or Crying? Hating or Loving? Sinking into Despair or Rising in Joy?

Genres are sorted, but what about the emotions we want to evoke in our viewers? This is where Large Language Models (LLMs) step in to help. To analyze the sentiment of each movie, we utilize a BERT model alongside plot summaries from our merged dataset. For shorter summaries within the model's input limit, we directly perform sentiment analysis. For longer summaries, we first condense them using a Seq2Seq BERT model, ensuring all plots are effectively analyzed.

Now, let’s go back to the analysis. Even though the sentiment labels for whole movie dataset are more or less evenly split, when it comes to remakes and originals there seems to be a prevalence of movies with sad or negative plots. Sad stories often have a strong emotional core that can resonate deeply with audiences, providing a built-in emotional investment and a higher potential for audience engagement. Due to nostalgic factor and emotional depth, these movies might have higher commercial appeal to the audience. Nonetheless, looking at the data, specifically vote averages, it seems that on average there was only 0.23% higher score for films with negative sentiment. There were no remakes that had a drastically different plotline from the originals: all of the sentiment labels for originals matched those of remakes. Overall, it is up to us to choose the sentiment of the story.

All of these stats is interesting, but is sentiment label significant when it comes to predicting whether the movie is a remake, or in other words, does it play a role in the remake? To answer this question we perform regression analysis and indeed, sentiment of the movie affects the remakes. Specifically, the movies with negative sentiment are likely to be a remake, hence the negative coefficient we observe in the plots later. Sentiment, however, seems to have no effect on the movie being remade, meaning that whether we have a happy or sad ending of the original story, whether it will be remade remains a mystery. To conclude our brief analysis, it is up to us to choose an original, making the audience cry tears of sadness or joy.

Who Should We Invite to Star?

Now, you might start thinking, who should we invite to join our movie to make it truly splendid - an impressive cast! For one, we should make sure our cast is diverse to maximize our monetary gains.

What about the actual ethnicities? How represented is the movie industry?

Despite higher revenues gained by movies with diverse nationalities and a lot of them represented in the filmmaking industry likely appealing to wider public, when it comes to race the movie industry should definitely do better..

Previously, remakes seem to have involved crews with huge fanbases, with the total popularity score across movies higher than those of originals and other movies. However, despite doing so, as we’ve seen before the vote average scores for remakes were similar to the average across all movies. So, should we invite a star cast? This is something we will need to think about further when doing the regression analysis.

Movies with diverse nationalities achieve higher revenues, reflecting their appeal to a broad audience. However, the stark racial imbalance in representation—dominated by "White" individuals—highlights the industry's shortcomings in inclusively. To truly embrace diversity, the film industry and we as a part of it, must prioritize equitable racial representation, not just nationality diversity, unlocking both ethical and market potential.

Remakes tend to feature more prominent crews compared to originals and other movies, often involving individuals with large fanbases. This is reflected in their higher total popularity scores of the crew. Furthermore, higher popularity scores for the cast meant that the movie was more likely to be a remake. However, despite this, the vote average for remakes remains similar to the overall movie average. Should we rely on a star-studded cast? This is a question worth exploring further in our regression analysis.

Future Directions

Following our exploratory data analysis (EDA) and initial investigation into patterns and differences between movie remakes, originals, and other movies, we delve deeper into two key areas of analysis:

Comparison of Movie Originals vs. Their Remakes: In this section, we take a closer look at how specific features of original movies differ from their remakes. By comparing attributes such as runtime, budget, revenue, and popularity, we aim to uncover the defining characteristics of remakes and understand the changes made to adapt these movies for modern audiences.

What Movie Should I Build From, and How Should I Make It? Here, we shift focus to explore what makes a movie more likely to be remade and how certain features contribute to its potential success.

What Movie Should I Build From, and How Should I Make It?

When planning a remake, it's crucial to evaluate which past movies are worth remaking and to identify the key attributes and logistics—such as cast, budget, genre, and more—that should be taken into account. These factors play a significant role in the success of your remake.

In this analysis, we aim to determine the most important attributes that make a movie a strong candidate for a remake or that contribute to the success of a remake. To achieve this, we employ regression and causal analysis to identify the specific attributes associated with remakes.

First, we conduct regression analysis with the outcome variable indicating whether a movie is a remake. From this, we extract impactful features by examining their p-values under the null hypothesis that the corresponding regression coefficients are zero. In the second step, we validate the attributes identified through regression using causal analysis and treatment effect estimation to confirm their significance.

Feature Engineering

Before conducting regression analysis, we performed feature engineering on the dataset. First, we calculated the maximum popularity of the crew and the average popularity of the top three cast members, as they are the most influential individuals in the movie. We then identified the most common genres and created dummy variables for them. Additionally, we incorporated the sentiment score of the movie, its language, and the country of origin, specifically noting whether it was from the USA. Then, missing values were imputed using the Datawig library, and each column was filled out based on the others as much as possible.

We also considered the movie's release year and grouped it into intervals. Based on prior regression analysis, we observed that remakes or movies with remakes were more common before the 1980s, a period associated with the advent of color television. Budget and revenue were also included, with a new feature created as the logarithm of the revenue-to-budget ratio. Revenue itself was excluded since we were primarily interested in its proportional relationship to the budget.

Regression Analysis with All Features

We perform regression analysis on the dataset. We consider two types of regression: considering the features (columns) to regress one-hot encoding of 1) being a remake or 2) having a remake for each movie. To make the regressions more interpretable, for the first regression we remove those movies that have remakes and for the second, we exclude those that are remakes themselves. We then plot the significance of the coefficients and their confidence intervals. We then try to verify the causal effect of having a remake or being a remake on the feature that we see their significance. We consider those features as significant if their p-values are less than 0.05 in the regression. The null hypothesis is that the coefficient is zero.

As we can see, factors such as sentiments, adjusted budget, cast average popularity, genres, and the country of production play critical roles in determining whether a movie is a remake. On the other hand, runtime, vote average, sentiments, adjusted budget, genres, black-and-white status, year and country of production, movie language, and the revenue-to-budget ratio also appear to have significant impacts. Next, we should examine whether these factors have valid relationships with a movie being or having a remake.

Causal Analysis

To assess the impact of being a remake or having a remake on the significance of specific features, we proceed as follows: for each column, we regress the feature of being a remake or having a remake. Next, we compute propensity scores and match samples accordingly. We then plot the average treatment effect (ATE) for the treated and control groups, including 95% confidence intervals. A feature is deemed to have a causal relationship with being a remake or having a remake if the ATE confidence interval excludes zero and the corresponding regression coefficient is statistically significant.

Here, treated samples are remakes or movies with remakes, and control samples are the rest. The outcome for the categorical variable is their value, and for the continuous variable, it is if the value is above the mean plus two times the standard deviation. This value is chosen to ensure that the feature has a significant effect on the outcome. We assess the validity of the causal relationship using the bounded odds ratio (OR). Specifically, we constrain the matched pairs to have an upper bound of 5 for the odds ratio. To evaluate sensitivity, we plot a histogram of the odds ratios for the matched samples. The causal relationship is considered valid if the odds ratio is below 4, with the majority of matched samples exhibiting an odds ratio less than 2.

Let’s Cook!

To conclude, when considering movies that have at least one subsequent remake, the following features showed a significant causal relationship:

• High Vote Average: Movies that receive higher viewer ratings are more likely to inspire a remake. A possible interpretation is that higher-quality or more critically acclaimed films are deemed valuable intellectual property worth revisiting or updating for newer audiences.
• Being in Specific Genres (Comedy, Thriller, Romance): Certain genres appear more likely to inspire remakes.
  • Comedy: Comedies that resonate well with audiences may be remade to align with contemporary humor, and cultural shifts, or to reach new demographic segments.
  • Thriller: Thrillers can be easily adapted to modern contexts and technologies, keeping their suspense elements fresh.
  • Romance: Romantic films that achieve a timeless emotional resonance may be reinterpreted for new eras or cultural backdrops.
• Black and White Movies: Historically, many black-and-white classics have been remade in color or updated to reflect modern production values. This suggests that older, iconic films (often black and white) serve as prime candidates for remaking, potentially due to nostalgia, cinematic legacy, or the desire to revitalize stories for contemporary audiences.
• Made in the USA: U.S.-produced films—given the global prominence of Hollywood—are more likely to be remade. This may stem from the wide international reach, market potential, and cultural influence of American cinema.
• Made Before 1980: Older films, particularly from before 1980, are more frequently remade. As time passes, production values, storytelling techniques, and audience expectations shift, making these older stories appealing targets for modern reinterpretation.
• High Log of the Revenue-to-Budget Ratio: Films that proved financially successful relative to their production costs are likely to catch the attention of filmmakers and studios looking for profitable properties to revisit. This financial indicator suggests that the perceived commercial viability drives the decision to invest in a remake.

For movies that are themselves remakes of earlier works, the following features have a significant causal relationship:

• Negative Sentiment: Remakes tend to arise even when the original film has a negative sentiment. Filmmakers might see an opportunity to reinterpret or enhance the emotional resonance of a sad or somber story to better connect with modern audiences or have deeper feelings in the movie.
• High Budget: Modern remakes are often given higher budgets, possibly to leverage updated technology, special effects, or star power. Studios may feel more confident investing heavily in known intellectual property, anticipating stronger box-office returns due to brand recognition.
• High Average Popularity of the First Three Cast Members: Remakes often get well-known or rising star actors to generate publicity and attract audiences. Casting high-profile talent can rekindle interest in a familiar story and differentiate the remake from its predecessor.
• Being in Specific Genres (Thriller, Romance): Similar to the pattern seen for having a remake, certain genres—thrillers and romances—are common subjects of remakes. These genres rely on timeless emotional or suspenseful narratives, making them ripe for re-interpretation and modernization.
• Made in the USA: The U.S. film industry’s influence is again evident. Remakes are frequently developed within Hollywood, likely due to the robust infrastructure, global distribution networks, and a history of successful adaptations.

The analysis suggests that certain film characteristics are causally linked to the likelihood of a movie having a remake or being a remake. Films that are older, successful, and produced in the U.S., and that belong to certain genres, appear more likely to inspire future remakes. For movies that are themselves remakes, factors like higher budgets, star-studded casts, and fitting into “remake-friendly” genres are associated with their production. They are mostly made by the US movie industry and they mostly have negative sentiments, probably to have deeper feelings about the movie. Additionally, the sensitivity checks reinforce the credibility of these findings, indicating that these relationships are robust to concerns about potential unobserved confounders.