Say it in plain English.

“No team plays back-to-back road games.” “Field 3 is dark after 6pm in October.” “Coach Oyelaran has two teams and can't be in two places at once.” Say it the way you'd say it to another human and Smart rules works out what you meant.

Most of the time what you described is a rule Big Sports already has, and it tells you so plainly — here's the rule for that, here's how it's set, here's what it does. When it isn't, it builds you a new one that behaves like a first-class rule: it can be hard or soft, weighted, switched off, edited, and deleted like anything else in your list.

You are never trapped in the text box. The rule catalogue is one click away, and it stays the working path for anyone who — reasonably — would rather pick from a list.

This is the part that matters, and it's the part nobody else does. A rule that has been generated for you is worth nothing until you know what it will do. So before you accept it, Big Sports runs it against your real season and shows you the result: how many of your games it flags, which teams and dates those are, and a couple of specific fixtures with a verdict on each — this one breaks the rule, this one doesn't.

Those are your actual games, not examples the model wrote to sound convincing. If the rule you described would blow up half your October, you find that out while it costs you nothing.

There's also a line you can open to read the exact logic it built, in full, if you're the kind of person who wants to see it. Most people never open it once. It matters that they could.

One honest caveat: the dry run needs a schedule to run against. If you're writing rules during setup, before you've generated anything, there are no games to check yet and the rule is simply accepted as written. Generate the season and the evidence is there from then on, including for every rule you add afterwards.

Language models are good at reading a sentence and bad at being accountable. So Big Sports uses one for exactly the first job and none of the second. The model reads what you typed and produces a rule. The engine — deterministic, reproducible, no model anywhere in it — takes your rules and produces the schedule.

That line is why “same inputs, same schedule” survives having an AI in the product at all, and it's why the evidence step above can exist: there's something real to check the rule against. A tool that lets a model write fixtures directly can show you a confident summary and nothing more.

And every rule it builds is a change like any other — it lands in your history with your name and the time on it, marked as one Smart rules wrote, and you can take it back out.

Smart rules writes rules. It does not set up your league for you, import your teams, or move games around on your behalf. There is no chat window you can ask to reshuffle October.

Those are real things we want to build, and describing them as though they already exist is how scheduling software earns the reputation it has. When they ship, they'll ship on this page.