Fun post, but wrong measure. You correctly note that eigenvector centrality is a measure of "network importance, which considers both direct connections and the influence of second and third-order connections. A higher score means an actor is more central to our casting network."
In other words, it captures not only how well-connected someone is, but how well-connected those connections are. Samuel L Jackson is not *only* well-connected, he's well-connected to other well-connected people (like Bruce Willis, with whom he appears in multiple movies). That's why he ranks so highly.
That's not what the Bacon Number is about, the *quality* of connections isn't where the emphasis is placed, the "reach" is what we're after. Bacon didn't say he "worked with only the important people in Hollywood, and the important people those people worked with". He said he worked with everyone: and "everyone" includes people with both high and low eigenvector centrality. Working with people with low eigenvector centrality (e.g., the cast of Tremors) gives him further network reach than someone like Meryl Streep, who does not appear in these kinds of pictures.
The Bacon Number is a fun application of the Erdos Number that was applied to scientific collaboration networks. Erdos did not emphasize second-order influence, just overall connectivity, so I'd recommend using a measure that emphasizes network paths rather than overall influence. Closeness coefficient might be a good one, or even basic betweenness centrality. For a true test of the "small worldiness" of Hollywood you could calculate the shortest paths for every node to every other node, then see how many of those paths include Kevin Bacon.
(PS, the sum of *my* Erdos and Bacon numbers is less than 10, that's pretty low!)
Whether Kevin Bacon is the best actor for this game today, is kind of moot, since it was conceived in the mid-90s, well before the era of enormous ensemble-cast action flicks and movie universes, both of which significantly increase the connectedness, especially of A-listers. It would be a challenge, I think, to find someone as good as Kev for this game in, say, 1999.
Also, you have to remember that in the early days we were playing this game without access to IMDB, or even Google. The point is not just to find the connecting films, they actually have to be ones that are memorable. And the actor has to be well known enough to casual movie fans as well. "Six Degrees of Richard Jenkins, no, you know him, the dead guy from Six Feet Under" is not a great party game, haha. Kevin Bacon is also well-known enough and recognisable enough that you can easily spot him in a bit part early in his career, or a later cameo.
Another factor that is important in the game is interconnectedness between different universes of actors. There are certain actors that are important nodes in connecting between different movie eras, or studios, or genres, or geographical markets. A memorable challenge was someone claiming it would be impossible to connect Reg Varney, British star of comedies such as "Mutiny on the Buses" to Bacon, correspondence on the subject which has fortuitously been preserved by New Scientist. https://www.newscientist.com/letter/mg15921439-500-ubiquitous-varney/
George Cole, probably relatively unknown outside of the UK, proves to be a key connector from 90s era Hollywood A-Listers to the 60s UK comedy scene via a co-credit with Julia Roberts. Then you have people like Roddy McDowall, who as a child actor in the 30s and 40s was working with actors from the silent era and was still appearing in support and cameo roles through the 1990s.
I think a more complete analysis would need to weight these kind of 'long distance' links higher than ones which link actors and films that are more similar in terms of where they start and end.
That's another point about Bacon actually, he was clearly not snobbish at all as he would appear in films in a such a wide range of genres compared to most actors, so he'd end up working with horror, comedy, romcom and 'serious drama' actors who were relatively unlikely to cross paths with each other.
Christopher Lee is an excellent vector across geography and decades via a lot of fun and memorable movies. My father was no good at Six Degrees of Kevin Bacon but he was a mean player of The Allseeing Eye of Dracuman.
It would be really cool to see an interactive visualizer for this data (see: “An example network of actors connected by co-starring film credits.” above). Maybe something like Visual Thesaurus. Thanks for this excellent post. Your work is always super interesting and top notch. Much appreciated.
The final thoughts of this article speak volumes to me. As movie person who has random actor knowledge and strong "that guy" game i often wonder why I need this info. I also love trying to write about a passion just to get information out there for no reason other than I have a nerdy fandom of something and I often struggle to find motivation (also data related). I get it haha.
This immediately made me yearn for an ensemble movie with just the supporting actor list. Walken, Jenkins, Tucci, Cox, Turtutto and Cheadle all together.
The last time I played, we had a "no using any MCU movies" rule to connect the two actors. Had to make it harder since my cousin was playing and she used to work at Blockbuster and AMC 😃
It's become cool to hate on Pulp Fiction and Tarantino but I've long-argued it's one of the most important movies ever made, not just for the movie itself, but the way it resurrected the careers of so many stars (e.g., Willis, Travolta) and made others household names (e.g., Jackson, Thurman). Very cool to see some data supporting that maybe it's the connectedness of the cast that made it the cult and then pop phenomenon that it became.
Fun fact: The game/concept of 6 degrees from Kevin Bacon was created by students at Albright College. Depressingly... It remains one of Albright's top claims to fame.
I first heard of this game when I discovered #movies on IRC in the '90s, though mostly we'd play "Links" instead, which is like the Kevin Bacon game, but without Kevin Bacon. Someone would suggest an actor, someone else another, and then we'd compete to see how few connections it took to link them.
It was helpful to have some Old Hollywood and New Hollywood connections on hand, because you'd often find yourself trying to link, say, Joel McCrea to Morgan Fairchild. Movies like Xanadu or On Golden Pond served to link big names of the past to modern actors, so it was important to have some of those at your disposal.
I'd just like to mention another venue that plays the Kevin Bacon game. Genealogy. There's a site called WikiTree that encourages everyone to link their family tree to one big tree that includes everyone in the world. One of the ways they encourage the linking is by playing a game they call CC7, which stands for connection count within seven degrees. Besides telling you all the ancestors and others you're linked to, the CC7 table they created chooses two celebrities to be permanent fixtures in the CC7 game. Those celebrities are Queen Elizabeth II and Kevin Bacon.
Fun post, but wrong measure. You correctly note that eigenvector centrality is a measure of "network importance, which considers both direct connections and the influence of second and third-order connections. A higher score means an actor is more central to our casting network."
In other words, it captures not only how well-connected someone is, but how well-connected those connections are. Samuel L Jackson is not *only* well-connected, he's well-connected to other well-connected people (like Bruce Willis, with whom he appears in multiple movies). That's why he ranks so highly.
That's not what the Bacon Number is about, the *quality* of connections isn't where the emphasis is placed, the "reach" is what we're after. Bacon didn't say he "worked with only the important people in Hollywood, and the important people those people worked with". He said he worked with everyone: and "everyone" includes people with both high and low eigenvector centrality. Working with people with low eigenvector centrality (e.g., the cast of Tremors) gives him further network reach than someone like Meryl Streep, who does not appear in these kinds of pictures.
The Bacon Number is a fun application of the Erdos Number that was applied to scientific collaboration networks. Erdos did not emphasize second-order influence, just overall connectivity, so I'd recommend using a measure that emphasizes network paths rather than overall influence. Closeness coefficient might be a good one, or even basic betweenness centrality. For a true test of the "small worldiness" of Hollywood you could calculate the shortest paths for every node to every other node, then see how many of those paths include Kevin Bacon.
(PS, the sum of *my* Erdos and Bacon numbers is less than 10, that's pretty low!)
Couple of thoughts on this analysis.
Whether Kevin Bacon is the best actor for this game today, is kind of moot, since it was conceived in the mid-90s, well before the era of enormous ensemble-cast action flicks and movie universes, both of which significantly increase the connectedness, especially of A-listers. It would be a challenge, I think, to find someone as good as Kev for this game in, say, 1999.
Also, you have to remember that in the early days we were playing this game without access to IMDB, or even Google. The point is not just to find the connecting films, they actually have to be ones that are memorable. And the actor has to be well known enough to casual movie fans as well. "Six Degrees of Richard Jenkins, no, you know him, the dead guy from Six Feet Under" is not a great party game, haha. Kevin Bacon is also well-known enough and recognisable enough that you can easily spot him in a bit part early in his career, or a later cameo.
Another factor that is important in the game is interconnectedness between different universes of actors. There are certain actors that are important nodes in connecting between different movie eras, or studios, or genres, or geographical markets. A memorable challenge was someone claiming it would be impossible to connect Reg Varney, British star of comedies such as "Mutiny on the Buses" to Bacon, correspondence on the subject which has fortuitously been preserved by New Scientist. https://www.newscientist.com/letter/mg15921439-500-ubiquitous-varney/
George Cole, probably relatively unknown outside of the UK, proves to be a key connector from 90s era Hollywood A-Listers to the 60s UK comedy scene via a co-credit with Julia Roberts. Then you have people like Roddy McDowall, who as a child actor in the 30s and 40s was working with actors from the silent era and was still appearing in support and cameo roles through the 1990s.
I think a more complete analysis would need to weight these kind of 'long distance' links higher than ones which link actors and films that are more similar in terms of where they start and end.
That's another point about Bacon actually, he was clearly not snobbish at all as he would appear in films in a such a wide range of genres compared to most actors, so he'd end up working with horror, comedy, romcom and 'serious drama' actors who were relatively unlikely to cross paths with each other.
Christopher Lee is an excellent vector across geography and decades via a lot of fun and memorable movies. My father was no good at Six Degrees of Kevin Bacon but he was a mean player of The Allseeing Eye of Dracuman.
It would be really cool to see an interactive visualizer for this data (see: “An example network of actors connected by co-starring film credits.” above). Maybe something like Visual Thesaurus. Thanks for this excellent post. Your work is always super interesting and top notch. Much appreciated.
Have you considered looking into the segmented populations, such as the top ten women or non white people, in your analyses?
It would be cool to see this applied to the crew of movies. Who is the most interconnected Director or writer or cinematographer?
The final thoughts of this article speak volumes to me. As movie person who has random actor knowledge and strong "that guy" game i often wonder why I need this info. I also love trying to write about a passion just to get information out there for no reason other than I have a nerdy fandom of something and I often struggle to find motivation (also data related). I get it haha.
Wow, graph theory. Great to see that.
Debra Winger: There are three roles for actresses: babe, and Driving Miss Daisy.
And I should have mentioned the Erdös Number
This immediately made me yearn for an ensemble movie with just the supporting actor list. Walken, Jenkins, Tucci, Cox, Turtutto and Cheadle all together.
The Kevin Bacon game is old by now. Perhaps at the time where it was introduced, Mr Bacon was higher in the rankings.
The last time I played, we had a "no using any MCU movies" rule to connect the two actors. Had to make it harder since my cousin was playing and she used to work at Blockbuster and AMC 😃
I am surprised you didn't conduct the same analysis within the time frame of Kevin Bacon's career (movies up to 2016).
Kevin Bacon is still in tons of movies! He’s far from done. 🥓
Same thought popped into my head - although I would have put the cut-off at around 2005 or so.
It's become cool to hate on Pulp Fiction and Tarantino but I've long-argued it's one of the most important movies ever made, not just for the movie itself, but the way it resurrected the careers of so many stars (e.g., Willis, Travolta) and made others household names (e.g., Jackson, Thurman). Very cool to see some data supporting that maybe it's the connectedness of the cast that made it the cult and then pop phenomenon that it became.
Fun fact: The game/concept of 6 degrees from Kevin Bacon was created by students at Albright College. Depressingly... It remains one of Albright's top claims to fame.
I first heard of this game when I discovered #movies on IRC in the '90s, though mostly we'd play "Links" instead, which is like the Kevin Bacon game, but without Kevin Bacon. Someone would suggest an actor, someone else another, and then we'd compete to see how few connections it took to link them.
It was helpful to have some Old Hollywood and New Hollywood connections on hand, because you'd often find yourself trying to link, say, Joel McCrea to Morgan Fairchild. Movies like Xanadu or On Golden Pond served to link big names of the past to modern actors, so it was important to have some of those at your disposal.
A nice update on the chapter I just read in Tipping Point (Gladwell) — nice to see stats with some more modern actors than he noted in his analysis.
I'd just like to mention another venue that plays the Kevin Bacon game. Genealogy. There's a site called WikiTree that encourages everyone to link their family tree to one big tree that includes everyone in the world. One of the ways they encourage the linking is by playing a game they call CC7, which stands for connection count within seven degrees. Besides telling you all the ancestors and others you're linked to, the CC7 table they created chooses two celebrities to be permanent fixtures in the CC7 game. Those celebrities are Queen Elizabeth II and Kevin Bacon.