Fair means declared

Two students compare worksheets after class. One got 3-digit multiplication, the other got times tables. Both sheets said "Multiplication practice" at the top, both were graded out of 10, and one of those students has a legitimate complaint.

The fix isn't to make every sheet identical. It's to declare what you varied.

"Hard" is not a property of a sheet

Difficulty feels like a quality — this sheet is hard, that one is easy — but it's a range you drew from. Say the range out loud and it stops being an opinion:

Now "level 2" means something a colleague can reproduce and a parent can question. It also means you can be honest with yourself when a class average drops: did they get worse, or did you quietly move to level 3?

That's the first half of fairness — the sample space is stated. The second half is that two students who were told they got equivalent work actually got equivalent work.

The seed is a contract

Every sheet the generator makes is built by a seeded random number generator. Same seed, same topic, same level, same count → the same sheet, forever. The seed is printed on the page and carried in the URL.

Three consequences that matter more than they sound:

• Versions without inequity. Change only the seed and you get a different sheet drawn from the same declared range. Row A and row B get different problems of provably equal difficulty — which is what you actually wanted when you said "make a version B", and not what you get when you hand-edit a few numbers and hope.
• The answer key cannot drift. The key isn't stored alongside the sheet; it's regenerated from the same seed. There is no version of this system where the key belongs to last week's worksheet.
• A link is a document. Send the URL to a substitute, a parent, or your own future self in March, and the exact page comes back. Nothing to attach, nothing to lose.

There's a rule holding all of that up, and it's a promise the tool makes to you: published sheets are frozen. New topics get added as new topics; the existing generators are never "improved", because improving one would silently rewrite every sheet anyone had already shared.

What you still owe the student

The declaration only helps if it reaches them. Two habits cost nothing:

1. Put the level on the page, not just in your gradebook. "Level 2 — unlike denominators" tells a student what they're being asked to do, which is different from telling them how they did.
2. Say when the version changed. If row B got a different seed, that is not a secret — it's the reason their neighbor's answers won't match, and knowing it prevents the exact argument you were trying to avoid.

Run it

Open the worksheet generator with a fixed seed loaded:

1. Note the seed on the sheet and in the address bar. Reload the page — identical problems, identical key.
2. Change only the seed. New problems, same declared difficulty: that's version B, and you can defend it.
3. Move the difficulty to level 3 and watch the operations change kind, not just size — the sample space really did move, and now you can say so.