New grading system based on player data

[Poobslag]

• [x] New level format includes goal for 'm', 'sss', 's-', etc...
• [x] New level format no longer supports 'weird rank fields'
• [x] Add demo which can calculate, inject goals for existing levels
  • [x] Demo extracts 'weird rank fields' to big json repository
  • [x] Demo reports oddities in big json repository
  • [x] Demo injects goals
  • [x] Demo rounds goals to nice numbers
  • [x] Demo injects gentler 's-' cutoffs for levels in first 2 areas (50%, 80%)
  • [x] Demo reports outliers based on player data

[Poobslag]

I've implemented the new rank algorithm which defines one or more thresholds in each level file.

I'd like to have a mechanism to tweak these manually based on playtesting data, I'll work on that over the next few days. My current idea is to have something like a "poobslag.json" file with a bunch of text like this:

practice/marathon_normal seconds=223 score=2609
career/asparagus_vault seconds=120 score=1742

For the player's best performance (highest score or lowest time), it'll store the seconds and score for that run. We can then compare their performance to the master rank for that run and get a list of percentages, and determine outliers.

If we get more playtesting data and add ways to import it, we might be able to do more complex statistical generation of these master times. It's possible I'm just really bad or really good at certain levels. With additional data we can hack together an algorithm which twists knobs and figures out things like "on asparagus vault, 78% of player's scores fall outside the standard deviation for where we'd expect given their other scores. by increasing asparagus vault's master score from 1700 to 2500, their scores fall in line with where we'd expect them"