The course's own rules — late work, collaboration, AI use, extensions, and what "in numbers" means when you submit something.
Policies and support#
One page, so the four briefs don’t each carry their own version of the rules and slowly disagree with one another. These are the course’s rules; anything university-wide is on the university’s own pages, not restated here.
Late work#
Five percent of the item off per working day, up to five days, after which it isn’t marked.
The fight logs are the exception, because they’re a running artefact: a missing week costs that week’s sheet and nothing else. Twelve sheets handed in on time with one gap beats eleven handed in late.
Collaboration#
Your log is yours. The twelve tables are your reading of twelve fights, and two students who logged the same fight should be able to disagree about a row — that disagreement is most of what week 1 is for.
Your models can be compared. Take somebody’s spreadsheet apart, argue about whether their Haste halves the reset or the rate, rebuild yours after losing the argument. A model is a claim about a system, and claims get better in public. Say whose model you looked at; that costs you nothing.
The design document is yours. Week 12 has you implement a classmate’s and hand back a question log — their document stays theirs, and what you submit is your implementation and your questions.
Using an AI assistant#
Allowed for anything you can verify against a recorded fight. Not allowed as a substitute for the log.
That’s the whole rule, and it follows from the course’s position rather than from anywhere else. If an assistant writes a spreadsheet formula, you can step the model forward and check its turn order against the recording — the fight is the ground truth, and a wrong formula fails visibly. If an assistant writes your ten logged hits, there’s nothing to check them against, and the item you’re submitting no longer exists.
So: fitting a curve with an assistant, fine. Generating the numbers you fitted it to, not fine. Drafting the prose of a design document with one, fine. Inventing the damage formula it specifies, pointless — that formula is the assignment.
Every submission carries a short note naming what you used and for what. “Wrote the sheet formulas with an assistant, checked the turn order against the week 6 recording myself” is complete. A submission with no note claims you used nothing.
Extensions#
Ask before the deadline if you can, after it if you couldn’t. A sentence saying what happened and what date you want is enough; no documentation is required for anything under a week.
The design document is the one piece with a hard edge, because marking has to finish — an extension past the last teaching day needs a reason the university recognises, and that process is theirs, not this page’s.
What “in numbers” means#
It appears in every brief, so: a submission is in numbers when somebody else can recompute your result without asking you anything.
A formula with every term defined is in numbers. “Damage scales with Strength” is not. A crit chance of 1/24 at 1.5× is in numbers. “Crits are rare but satisfying” is not. A threshold of 300% with a decay rate is in numbers. “The gauge drains fairly quickly” is not.
Prose is welcome everywhere and does a job no table does. It just can’t stand in for the arithmetic, because the arithmetic is the thing being marked.
Getting help#
Idris first for anything practical — a spreadsheet misbehaving, a fit that won’t converge, a recording you can’t find. Marisol for the design document, marking and dates.
If a question has gone unanswered for two working days, send it again. That isn’t rudeness, it’s the agreed escalation.