Every One Piece Faction In One Network

Added: 2026-05-05

[00:00:00]Hello everyone. Recently I've been having a lot of fun working [music] on this thing, a network of 1,600 characters from One Piece connected by whether or not a pair of characters have ever belonged [music] to the same faction. It's a One Piece affiliation network. Characters in One Piece generally have pretty complex relationships and they're not always obvious. Take the Sun Pirates for example. All members of the Arlong Pirates have an [music] Arlong tattoo, but some of them also have a sun tattoo.

[00:00:27]We see this a dozen or so episodes into the show, but in the anime this isn't ever commented [music] on. It won't be for hundreds of episodes later until we learn what this tattoo means to the people that have it. To me, it just gives a sense that characters have a life [music] and a past that's independent of the viewer. It's one of the things that I love most about the show and what made me want to try to visualize those relationships. Because [music] even if you'd argue that maybe One Piece is too big for its own good and as someone getting through the Dressrosa arc, I might not disagree.

[00:00:56]That's also exactly why it's fun to look [music] at it just from a strictly data science point of view. Because as we'll talk about, this network is in some ways strangely realistic and really [music] interesting. So, that's what today's video is about. Honestly, it's mostly just me nerding [music] out about this thing, what it's trying to show, and how actually we can do much better than this. There's a link in [music] the description to an interactive version you can play with and a link to the data for the nerds on Patreon and Nebula, as well as an extended discussion at the end where I talk about how this network compares with my previous [music] attempt on the main channel that couldn't really find a place for here.

[00:01:30]But okay, what is this thing? How did I make it?

[00:01:32]Well, code-wise it's stupidly simple.

[00:01:35]For each character I have their affiliations, [music] and so if two characters have the same affiliation, they'll get a link. And that's it.

[00:01:42]That's all my code is doing. The vast majority of nodes are in this giant component here. That is the technical name for it, by the way. And we'll talk about those in a second, but some characters don't have affiliations and they form the bulk of these nodes flying around the periphery here. They don't connect to anything. There are a couple of groups that make these small affiliation sub-networks, but don't connect to the main body. Most of these will form small triangles and polygon structures, but a few of them, like the Drum Island arc characters, can be somewhat complex. And in some sense, this is actually exactly what we would expect. A giant component, fittingly, usually contains most of the nodes in the network. If you add in a new node, it's overwhelmingly likely to make a connection to a node that's already part of the giant component. So, this makes these small islands very unlikely and relatively rare. So, in that sense, this is exactly what we would expect. It's genuinely what we see in real life networks. But in another sense, there's really no reason for this to be the case. This is a fictional show about pirates hopping from island to island.

[00:02:44]If anything, it would have been very reasonable for the network to be very split into a bunch of small factions on distinct islands. And I'm sure for a large, long-running show, that would have been the much easier route to go down. like Pokémon, for example, with a character-of-the-week style approach would look very, very different to the One Piece faction network. So, the fact that it does have this structure tells us a little bit about how Oda chose to approach this world. To what extent was this always the plan? I'm not sure. I do think that the early parts of the series do sort of have this more isolated islands feel to it, which is honestly my favorite part of the show. But even then, we do have hints of a larger world, like the Sun Pirates tattoos, for example. But yeah, back to the network.

[00:03:28]One of the things we typically want to know when we look at a network is what are the most important nodes in the network? The fancy name for which is centrality. [music] Like if you've watched the show, then of course you know who they are, but we want to see how that's reflected in the structure itself. A node can be important in many different ways. A node can be important because it's connected to a lot of other nodes, that's called degree centrality, or maybe you only have one link, but that one link is really important, which in turn raises your importance. That's called the Katz centrality. The Marines, the Beast Pirates, and the Charlotte Family are the three largest organizations. So, they're naturally pretty easy to see in the network, but they're mostly filled with unimportant characters, despite all of them technically having a lot of connections. I mean, that's not entirely true. Garp and Coby are both here, but for the most part, it's clear that just because you have a lot of connections in this network doesn't mean you're actually super important to the plot.

[00:04:17]Really, the most interesting characters are the ones that bridge together multiple factions. That's characters like Ace, Shanks, and Trafalgar Law. In other words, they may or may not have a lot of connections, but their role is to bridge together different communities, which if you've watched my other videos, you might remember more formally as the betweenness centrality. So, here I've resized nodes according to betweenness, which shows that the most important character, at least according to this centrality, is Jinbe, which isn't a bad pick. He seems to know almost everyone and is really involved with a lot of the big players in the world. Maybe best evidenced by his sun tattoo. Not sure I would put Drake at second, though. I mean, he's not unimportant, but he's definitely not Jinbe. This is more a consequence of Drake being between both the Marines and the Beast Pirates, which just artificially inflates his score.

[00:05:03]That being said, if we look at a list of the characters with the most betweenness, it is really important characters. So, I think the trend is generally true. Plot-important characters tend to be the ones with high betweenness.

[00:05:17]Which, again, just from a story point of view, I guess makes sense. While it could be interesting to have a show where 1,600 characters get an equal amount of screen time, the more sensible approach is to instead have a few characters who really drive most of the plot. Oh hey, sorry to jump in. I was editing that and I kind of realized, I think that statement might explain why I personally didn't really like Dressrosa.

[00:05:40]I think the story was really good and Donquixote was a really, really good villain, but a lot of Dressrosa was dedicated to a lot of the side characters. Each episode kind of going from different side character side character, seeing kind of what they're doing in that particular situation. And I remember even as I was kind of watching it, I was like, this is a really cool idea, but I just don't like it.

[00:06:04]>> [laughter] >> I just kind of want to get back to the main thing. [music] And I really struggled with that because I think there is some sort of there is a novelty in terms of shows like this where I do want to see what other characters are doing and and what the rest of the world is doing. But I just felt like it was sort of poorly executed maybe is the right way to say it. Anyway, sorry. I don't know. Just have to get that out of my brain. I was just as I thought of it.

[00:06:30]Back to the video. Okay, never mind. I really had to try and visualize what I was just talking about there, and this is a day before I am planning to on posting the video, so I just really made this sort of last minute. The idea is relatively simple. The animation sweeps across episodes and I'll connect nodes together if one, they appear in that episode, and two, they belong to the same affiliation. Now, prior to the time skip, we see that most of the nodes are these connected clusters are relatively small as the episodes kind of go along.

[00:07:03]And that's what I was talking about, sort of more self-contained island-to-island adventure. It's really sort of around episode 500 where there's this sort of fundamental shift in how many characters are appearing in episodes. And that's kind of what I was talking about in terms of Dressrosa, but it seems to be less of a Dressrosa thing and definitely more of a post-time skip One Piece thing. Anyway, I mainly just wanted to show this cuz I thought it it looked cool.

[00:07:31]And I think it's sort of backs up one, what I was feeling, and two, what I was saying. So, with that, yeah, now we can go back to the video. Now, there is a bunch of other stuff I'm sure that we could do with this network, but doing so would ignore one important fact here, which is that this network kind of sucks.

[00:07:51]I mean, don't get me wrong. I think it looks cool, but it's just not really efficient. There are 130 named Marines, which requires 8,400 links for characters who mostly never even interact. And just in general, it's kind of hard to see which organization someone belongs to just by looking at this, especially if they belong to multiple organizations. Take Jinbe and Arlong, for example. They have a link between them, but it's not clear why. Is it because they were both part of the Sun Pirates? Were they both part of the Arlong Pirates? Were they part of both?

[00:08:23]It's really not clear why they have this link unless you personally know why.

[00:08:29]Luckily, we can do much better.

[00:08:32]We'll introduce a new type of node, an affiliation node, and instead of connecting characters according to affiliation, we connect characters to affiliations. This is called a bipartite network. Bi meaning two, in reference to the fact there are two types of things in the network, characters and affiliations. Instead of using 8,400 links to join 130 Marines together, just connect them to the Marine affiliation node, which only needs 130 links. For reference, this reduced my file size from 7.6 megabytes to 0.8 megabytes, a factor of 10 smaller. And overall, that just makes the network much easier to read. The Arlong Pirates and their relationship to the Sun Pirates is more clear and also detailed. We see some of the former Sun Pirates never joined the Arlong Pirates. We even see the Marco Pirates, a group I had completely forgotten about until I made this network. Previously, none of this information was obvious. There's actually another way to see this, though. Something you might notice from looking at this, or if you just think about it for a second, is that a link can only ever exist between characters and affiliations. That means that if we were to arrange the network so that affiliations were on one side and characters were on the other side, links will only ever go between the two sides, never within them. This is in fact sort of a sanity check. If I give you a random network and you could arrange it this way, you'd know that you were dealing with a bipartite network.

[00:09:54]Though, that being said, I do think that the original layout just kind of looks cooler for this. So, not only is this cleaner and much easier to read, it also genuinely has more information in it. So then, why did we even look at that original network at all? Well, mostly because in the real world, that's all you'll ever get. Think of it this way.

[00:10:14]People in the same organization interact with each other, not with the organization explicitly. [music] If you're some random in the One Piece universe trying to create a network of pirates, what you'd quote-unquote [music] measure or observe would be that original network that we had, not the bipartite network. In this sense, while it was frustrating that we originally couldn't tell why Arlong and Jinbe were connected, it was also realistic. One Piece is this weird edge case where someone has meticulously documented every affiliation for 1,600 characters, which is information you just can't expect in the real world. Having that kind of complete information [music] is like the One Piece of network science.

[00:10:54]Because if you have a bipartite network, you can do anything, including getting that original one.

[00:10:58]>> [music] >> It's actually pretty easy. First, we connect all the nodes that belong to the same affiliation, and second, just remove all the affiliation nodes. That's it. That's all you need to do to go from the bipartite network to the original one, which more formally is called the forward bipartite projection. That's just the fancy name for the thing that we started off with. The key point though is that if you start with the projection, you can't go back to the bipartite network. So, in reality, I've been playing a bit of a narrative trick.

[00:11:26]While we started [music] off talking about the projection, in the background, I actually started with the bipartite network and then made [music] the forward projection. Though, you might be wondering if this is called the forward projection, what is the backward projection? Well, it's not to go back to the bipartite, like I said, you can't do that. Instead, if we start with the bipartite, you can do the same thing, but this time connect affiliations and remove the characters. Now, instead [music] of 1,600 characters, we have a network of 200 organizations and how they relate to each other. This is called the backward projection. I also made it so that the wider the links are, the more members they share. Honestly, I think this is probably my favorite network of the three just because it's small enough to be comprehensible, but big enough to be interesting.

[00:12:09]>> [music] >> But hopefully, it's even more clear here why if you started with this thing and that was all the information you had, you couldn't make the bipartite network.

[00:12:17]Which always kind of sucks as a network nerd because bipartite networks, when you can build them, are [music] just really cool. You can look at relationships between characters and how they're impacted by the story and how they in turn impact the story back. We already talked about how in the original network, it's not really clear why Arlong and Jinbe are connected together.

[00:12:37]It's only when we look at the bipartite network that we see that while yes, they were both part of the Sun Pirates, Jinbe did not join Arlong. It's a divide that reflects what the Sun Pirates meant to both of them and why they continue to wear the tattoo, but for very different reasons. And it's not even the only time we see Oda pull something like this. In episode 91, we meet Luffy's brother Ace.

[00:12:58]On his arm is a tattoo that says Ace, but spelled with a crossed-out S.

[00:13:03]>> [music] >> He never brings this up, neither does Luffy, and there isn't anyone saying, "Hey, explain your tattoo." We won't know for a long time why that S is crossed-out. Is it loss? Is it revenge?

[00:13:12]It's not really clear, and honestly, it's entirely possible that even Oda didn't know at that point. But it is clear that it means something to Ace from a prior life that we, the audience, are just not privy to. Just like the sun tattoo means something to Arlong, Hatchan, and Jinbe. It's oddly real, and if nothing else, this network is just my little payment of respect to that. And [music] finally, a chance for me to nerd out about bipartite networks. So, yeah, if you like this video, then check out the others I have on this channel, or maybe check out the anime character network video I made on the main channel a while back. This video was essentially inspired by that one. Uh thanks for watching.