This modified chord chart depicts all ingredient interactions found in the 77 cocktails officially recognized by IBA. A chord between two ingredients represents cocktails that contain both.
Ingredient arcs and chord widths are proportional to the number of cocktails represented, with chords additionally color-coded. The relative size of a chord compared to one of its ingredient arcs therefore indicates the conditional frequency of the pair given the ingredient.
Since cocktails often contain more than two ingredients, an ingredient can have more pairings than total cocktails associated with it and chords have to overlap. The areas of overlap have no significance; they're simply a consequence of the algorithm used to distribute the chords.
For example, more than a third of cocktails which contain orange liqueur (the liqueur used in the most cocktails) also contain lemon. This pairing is fairly common among cocktails, but roughly half as frequent as the pairing of rum and lime.
This visualization focuses on flavor combinations, so similar-tasting ingredients are grouped together. There are some judgement calls involved and a shared grouping is not meant to imply that ingredients are fully interchangeable.
Today's plot is inspired by Surviving The World's Lesson #1225. Sorry, Dante; it turns out you can't mix just anything.
The choice of visualization for this was tricky. The chord chart isn't good for spotting clusters (For which I probably would have used a force-directed graph because adjacency matrices aren't pretty...) but it's great for judging relative frequencies of interactions. I really wanted to preserve conditional frequencies as well, which meant letting chords overlap (and spending hours coming up with styling that makes them readable while overlapped). The result isn't perfect, but it allows for a lot of different comparisons.