The wheel is a map of what gets mixed with what

Okay, so this is a chord diagram. Every ingredient in the collection gets a small colored arc on the inner ring, and the outer ring bundles those arcs into ten families: dark liquor, light liquor, liqueurs, wine and beer, juices and sodas, mixers and sweeteners, fruit and citrus, herbs and spices, bitters and aromatics, and a catch all for everything else.

The ribbons flying across the middle are the actual point. A ribbon between two arcs means those two ingredients showed up in the same recipe at least once, and the fatter the ribbon, the more recipes they share.

Hover a ribbon and it names the pair and counts the drinks behind it. Click an arc and it holds the highlight, fading every ribbon that does not touch it, and the family bars underneath swap over to a ranked list of that ingredient's partners. Honestly, that ranked list is the fastest way to an answer. Click gin, read down the list, done. No squinting at ribbon widths required.

The Connections dropdown decides how many of the strongest pairings get drawn at once. The full wheel is 1,902 connections between 300 ingredients, which looks incredible and tells you almost nothing. The 100 and 250 settings strip it back to the skeleton. Ingredients that lose all their connections drop off the ring completely, so whatever you are looking at is always what is still genuinely connected.

Everybody is basically drinking citrus

I went in expecting spirits to dominate. They do not.

Lemon juice shows up in 379 recipes and lime juice in 312, which means roughly a third of every cocktail in here is built on citrus acid. Gin leads the spirits at 353 and beats vodka at 184, which is the opposite of what modern liquor sales would predict, and tells you a lot about when these recipe books were written. Simple syrup turns up 235 times. Sweet and dry vermouth appear almost exactly as often as each other, 151 and 150, because the classics will not stop pairing them.

Recipes are also shorter than bar menus lead you to believe. The median drink has four ingredients, and 89% of the collection sits at five or fewer. Drinks with eight or more, which is mostly tiki and punch, add up to 30 recipes total. Of the 654 recipes that bother recording a technique, shaking beats stirring about two to one, and building in the glass is a distant third.

Where alcohol by volume is recorded, the median finished drink lands around 18%. A little stronger than wine, way below the spirit it started as. That gap is the whole craft, really. Dilution, citrus and sugar dragging a 40% spirit down into something you can have two of.

A very short history of mixology

The word cocktail turns up in print in 1806, in a New York paper called The Balance and Columbian Repository, and the definition is almost aggressively simple: spirits of any kind, sugar, water, and bitters. That is the entire category. If you read that list and thought "that is an Old Fashioned", you are right, and so is the name. The Old Fashioned is called that because by the time it needed a name, people were asking for the old fashioned kind of cocktail.

Jerry Thomas published the first real bartender's guide in 1862 and the job turned into a craft with literature attached. Then American Prohibition ran from 1920 to 1933 and cut the thread. Bartenders left for Havana, London and Paris, and the drinking that stayed home got worse, because when the spirit is bad you bury it in something sweet.

Mr. Boston's guide arrives in 1935, two years after repeal, which is exactly why it reads the way it does. It is a book about rebuilding. That guide is one of the four sources here, and you can see it sitting in the data: powdered sugar, blended whiskey and grenadine all show up far more often than any bar would reach for now.

Then the long slump, roughly the 1970s through the 1990s, when sour mix came out of a gun and vodka took the whole shelf. The International Bartenders Association had already formed in 1951 and kept publishing specs through all of it, which is part of why the 90 drinks on its official list read as a spine running through the century. The revival started in the late 1990s and is the reason you can order a properly made Daiquiri today.

Here is the part I keep thinking about. That 1806 definition put bitters into the base case of the entire category, and two hundred years later they still appear in 357 recipes in this collection, in doses measured in dashes. Something that small staying that necessary for that long is rare.

I sorted everything into ten families

The families exist so the wheel has some structure at a glance. Worth saying upfront that the boundaries are editorial, not chemical. Here is how I drew them.

Dark liquor
Aged and barreled spirits. Bourbon, rye, Scotch, Irish whiskey, cognac, the other brandies, and the aged rums that behave like them in a recipe.
Light liquor
Clear and unaged. Gin, vodka, blanco tequila, mezcal, cachaça, pisco, white rum, aquavit, grappa.
Liqueurs
Anything sweetened and flavored. Triple sec and curaçao, maraschino, the crèmes, Chartreuse, amari, and bittersweet aperitifs like Campari and Aperol.
Wine & beer
Fermented instead of distilled. Still and sparkling wine, both vermouths, sherry, port, Madeira, sake, cider, beer.
Juices & sodas
Nonalcoholic liquid volume. Citrus and other fruit juices, tonic, soda water, cola, ginger beer, coffee, tea.
Mixers & sweeteners
Syrups, sugar in all of its forms, honey, orgeat, falernum, grenadine, cream, milk, egg.
Fruit & citrus
Solid fruit, plus the parts of it that go in as garnish or muddle. Wedges, peels, twists, zest, berries, olives, cucumber.
Herbs & spices
Mint, basil, thyme, lavender, cinnamon, nutmeg, cardamom, ginger, chili, vanilla.
Bitters & aromatics
Angostura, Peychaud's, orange, and the newer flavored bitters like mole, grapefruit, cardamom and rhubarb. Tiny volumes, 357 appearances, and the difference between a drink and a mixture.
Other ingredients
Salt and saline, vinegars and shrubs, brines, sauces, eggs, ice, and the long tail of single use oddities that every crowdsourced recipe set collects.

808 distinct ingredients get sorted this way, and about 99% of them land in a family either by name or from the source's own ingredient tree. The leftovers are brand names that appear in exactly one recipe, and they sit in "other ingredients" until somebody files them properly.

Bitters got their own family late. They were sitting in the junk drawer next to soy sauce and crushed ice until I actually counted them. 357 appearances, third behind spirits and citrus. Oops.

This is not a flavor pairing engine

A thick ribbon means two ingredients kept showing up in recipes that people wrote down, and that is a record of habit and commerce as much as taste. Gin and lemon juice are joined by a fat ribbon because the sour template is three hundred years old, not because an algorithm decided they go together.

It is also not a complete census of drinking. Recipe collections carry their compilers' eras and appetites. The Mr. Boston guide is heavy on American drinks from the middle of the last century. Crowdsourced entries wildly overrepresent shots and novelty drinks. The curated sets lean modern and craft. So reading this as "what the world drinks" would be wrong. Reading it as "what these 1,857 written recipes contain" is exactly right.

I think that is still the more interesting question anyway. What people bothered to write down is its own kind of data.

Where I got all of this

Four openly licensed sources: Bar Assistant (657 recipes), IBA Official Cocktails (2 recipes), TheCocktailDB (356 recipes) and Mr. Boston Official Bartender's Guide (842 recipes). Every recipe keeps a link back to wherever it came from, and a drink that appears in more than one source gets recorded once, under whichever version is most detailed. The 90 recipes on the International Bartenders Association official list are flagged as official.

The wheel itself runs on a separate ingredient association matrix that I built earlier and kept as its own thing, so the picture stays stable while the recipe collection keeps growing underneath it. The sources page has the full breakdown of what each dataset brings and what license it carries. The about page covers the method, and where it falls apart.