50 Years of Oscars: Acting Success and Collaboration

I am delving into IMDb data and employing network science to unravel the hidden dynamics behind the most celebrated actors and actresses throughout history.

The Academy Awards, commonly referred to as the Oscars, have stood as a hallmark of excellence in the film industry for nearly a hundred years, recognizing both individual and cinematic achievements across various categories, including Best Picture, Best Director, Best Documentary Feature, and Best Sound. While there are currently over twenty award categories, the most recognized figures remain the actors and actresses, who are intricately linked within a network of stars. To understand how these connections might influence their Oscar success, I examined historical Oscar data spanning the last fifty years, focusing on the major milestones of award winners and their collaboration patterns.

Data Collection

To begin, I compiled a list of nominees and winners from 1973 onward sourced from the Internet Movie Database (IMDb), concentrating on categories such as Best Performance by an Actor or Actress in a Leading Role, and in Supporting Roles. This effort yielded a dataset comprising 1,200 nominations, 200 Oscar wins, and 548 artists.

Subsequently, I gathered the individual filmographies of these artists using the IMDbPY package in Python, resulting in nearly 40,000 titles. After removing generic titles and collections like The Oscars or The EE British Academy Film Awards, I narrowed down the filmography to approximately 35,000 unique titles.

Oscar Distribution

With my dataset of over 500 artists and 35,000 titles compiled, I began to navigate the Oscar landscape to pinpoint key figures and uncover intriguing facts. For example, a ranking based on nominations and wins reveals that Meryl Streep leads with 21 nominations and three wins. Jack Nicholson also boasts three awards but from only ten nominations since 1973. Daniel Day-Lewis and Frances McDormand have similarly secured three Oscars each, with six nominations each. Notably, McDormand, recognized for her role in Nomadland (2020), also contributed to a Best Motion Picture win, which is not included in this analysis as it is not an individual award.

The dataset includes a select few double winners, like Christoph Waltz, Hilary Swank, Kevin Spacey, and Mahershala Ali, who each received Oscars on both of their nominations. However, winning multiple awards is rare; 200 Oscars have been awarded to 174 artists, with only 22 receiving multiple statues. This indicates that 68% of nominees have never won. Notably, Glenn Close, nominated eight times, and Amy Adams, nominated six times, both left empty-handed. Conversely, Al Pacino received nine nominations, including for The Godfather and The Irishman, before winning once. Leonardo DiCaprio also garnered six nominations, ultimately winning an Oscar in 2015. A comprehensive chart detailing the top actors and actresses based on nominations and wins is shown in Figure 1.

Network Approach

While the individual award landscape offers valuable insights, it is reasonable to hypothesize that collaborations and connections may influence award outcomes, especially considering that the roughly 7,000 members of the Academy of Motion Picture Arts and Sciences determine the awards. Previous research has highlighted various network effects on success in film, prompting me to construct and analyze the social networks of winners and nominees.

In this network, each node represents an Oscar nominee or winner (the 548 artists), with connections established between individuals who appeared in the same films. The frequency of these shared appearances strengthens their link. After processing the filmographies, I developed a densely interconnected network of 546 nodes and 17,140 links, indicating the need for further data refinement. Despite eliminating several titles like The 76th Annual Academy Awards, some generic collections remained in the dataset, which did not represent significant collaborations and needed exclusion.

Instead of further manipulating the original dataset, I opted to clean the network by filtering out statistically insignificant edges while retaining as many nodes as possible to minimize information loss. Utilizing a previously established method for cleaning noisy networks, I refined my analysis to yield a graph consisting of 526 connected nodes and 1,299 links, illustrated in Figure 2.

Network Description

One notable aspect of this network is the absence of distinct communities, contrasting with networks that feature well-defined groupings. Instead, the network displays two major denser clusters—one side featuring actors like Christoph Waltz and the other including figures like Jodie Foster, with a sparse region in between.

Moreover, the most prominent nodes representing successful figures are distributed relatively evenly throughout the graph. This observation suggests that awards are fairly distributed within certain cliques. This is further corroborated by low clustering coefficients and a minimal rich-club coefficient, indicating a lack of a central core where top nodes predominantly connect.

An intriguing trend emerged when I colored the nodes by the debut years of the artists in relation to their Oscar nominations, revealing a gradient from bright (earlier) to dark (more recent) from left to right, as illustrated in Figure 3. This visualization indicates a few all-time Oscar stars, while the majority of new names and trends continue to emerge, reflecting the evolving landscape of the film industry.

Stars in the Network

To further understand the network dynamics, I examined individual nodes to identify the top networkers and whether their networking intensity correlates with award success. I calculated both the Degree (number of connections) and Weighted Degree (sum of edge weights) for each artist. The results are summarized in Table 1, highlighting Robert De Niro, Diane Keaton, and Burgess Meredith as leading figures. Their network metrics illustrate varied patterns, with their wins and nominations ranging from zero to two wins and one to eight nominations.

Do these network metrics indicate a relationship with Oscar performance? A correlation analysis suggests not, as the values remain low (Table 2). Furthermore, the average Weighted Degree shows minimal variance between those with zero, one, or two Oscars (ranging from 8.76 to 9.44), with only a slight drop for those with three wins (5.75), likely due to the small sample size. While this analysis focused on basic metrics, additional computations using more sophisticated centrality measures yielded similar insights.

Conclusion and Limitations

In this brief exploration, I examined the leading Oscar winners and investigated whether their successes stemmed from robust networking within Oscar-centric communities. My findings revealed minimal biases, suggesting that networking patterns exhibit surprisingly low correlation with Oscar achievements.

However, this analysis is not without limitations. A clearer picture could emerge by comparing Oscar winners and nominees alongside actors who were never nominated, providing a baseline for analysis. Additionally, defining collaboration networks solely based on shared films may not capture the full extent of relationships, as it does not account for alternative forms of networking. Future studies could enhance this understanding by incorporating social media interactions or news articles.