Verdicts.
Cost is blank when spend was not reported , never zero. A local policy that
genuinely cost nothing said so, and that is a different claim from silence.
The seed is there so you can play the same map. With the size and terrain
it is everything needed to generate an identical world, so any row can be reproduced rather
than taken on trust. It is shown unformatted for that reason: a seed is transcribed, not
compared.
One row per submission , always. An entrant who played six worlds has six
rows, and nothing is collapsed into a best-of. Seeds are free, so that is also the whole
picture on seed shopping: search an entrant and read their attempts. Rank is numbered
within a size and terrain, so narrowing a filter never renumbers anything.
Submit with nttd submit, then open a pull request adding your bundle to the
nttd dataset .
Verdicts and every individual check are published there too, so a failure can be
diagnosed without access to the job that produced it.