Explore or exploit? What the maths of slot machines says about choosing books

Every time you choose a book you face a small dilemma. You can pick something very like the books you already love, which is a near-certain good time. Or you can try something different, which might become a new favourite and might be a slog. Statisticians have studied this exact trade-off for decades. It's called the explore–exploit dilemma, and the maths behind it gives some surprisingly practical advice for readers.

The casino problem

The classic version is a gambler facing a row of slot machines, the “one-armed bandits” that give the multi-armed bandit problem its name. Each machine pays out at a different, unknown rate. Every pull, the gambler chooses between two things:

• Exploit: play the machine that has paid best so far.
• Explore:try a different machine to learn whether it's better.

Only exploit and you may spend your whole life on a decent machine while a great one sits untried next to it. Only explore and you never cash in on what you've learned. The best strategies mix the two, and how they mix is the interesting part.

Swap “machine” for “kind of book” and it's your reading life. Your safe corner, perhaps tense, fast, standalone thrillers, is the machine you know pays out. Every other combination of mood, pace and genre is a machine you've pulled a few times or never.

Three lessons from the maths

1. Explore more when you have more time ahead

The value of discovering something new depends on how long you'll get to enjoy it. A new favourite genre found at twenty pays off for decades. That's why optimal strategies explore heavily early on and exploit more as time runs out. For readers the time scale is long, and there's always another book, so most of us should explore more than we do.

The same logic applies at smaller scales. On a two-week holiday with a stack of books, one gamble costs little. When you get one hour a week, a safe bet may be the better call.

2. Be optimistic about what you haven't tried

One family of good bandit strategies follows a rule called optimism in the face of uncertainty: when you know little about an option, assume it might be good, and try it until you know better. The logic is that you can only learn something is bad by trying it, but you can never learn something is good by avoiding it.

In reading terms: if you dismissed a whole mood or genre on the strength of one or two books, you haven't learned much. Two data points aren't a verdict. Give it a fair sample.

3. Explore close to home first

Exploration doesn't have to mean a random leap. The cheapest experiments change one thing at a time. If you love dark, slow literary fiction, try a dark, slow fantasy. Keep the mood and pace you trust, and change the genre. If it works, you've found a whole new shelf. If it doesn't, you've learned something specific.

This is the reading version of a controlled experiment: vary one factor, hold the rest steady, and the result tells you something clean.

Your data can only see what you've tried

There's a deeper reason to explore. Your ratings only describe the books you chose to read. If you've never read a funny novel, your data can't say whether you'd love one. It can only say you don't pick them. Any recommender trained purely on your history, including your own instincts, inherits that blind spot. Exploration is the only way to collect evidence about the possibilities you've been ruling out. The taste guide calls this selection bias.

How Match Book builds this in

The Match Book app has three ways to choose, running from pure exploit to deliberate exploration:

• Exact match is pure exploitation: books that fit your taste on every dimension you care about.
• Dials is the controlled experiment. Move a single dimension (darker, shorter, set somewhere new) and keep everything else yours.
• Spin is structured exploration. It flips some of your strongest preferences to their opposites on purpose, with an intensity slider that runs from a gentle nudge to full chaos.

You can try the idea by hand on this site. Open a book you loved, look at its dimensions, then go to a collection that keeps most of them and changes one. For example, go from fast-paced thrillers to fast-paced books in other genres, or from cosy books to cosy fantasy.