Jewels Issues Tracker

[lukeoftheshire]

Currently the code for Jewels has a few problems. It is essentially undocumented, which makes it difficult to figure out what each function does. In addition to this:

• How the number of jewels increase from level to level is extremely esoteric - as near as I can tell:

  1. at the end of each successfully completed level a certain number of bonus points is assigned, equal to the remaining time minus 2 times the number of errors made. (This number can be negative, which means total bonus points can be negative)
  2. This number is passed to a different function, which adds it to the running total of bonus points. The running total is then divided it by the bonus_point_count setting. By default bonus_point_count is 50. The running bonus total divided by bonus_point_count is floored and stored as level (a confusing name as it is NOT the same as the current game level).
  3. If level is exactly equal to the current game level, the game level is not changed. Otherwise, the game level increases by one (meaning a negative score still increases the game level)
  4. If level is greater than 1, level-1, divided by the x_changes_in_level_count then multiplied by x_diamond_count (from settings) is used to determine how many extra jewels should be added. Because of this, scoring a lot of bonus points or having a low bonus_point_count setting can lead to a big spike in Jewels from one level to another. It is basically impossible to predict how many jewels will be added from trial to trial. If level is less than or equal to 1, nothing happens.

  This seems like a very complicated way to handle this - it feels like an attempt to make a leveled test that adjusts for user performance but the fact that diamond count increases unpredictably to a max of 25, cannot decrease, and the level value does not necessarily increase work against this. Should we consider: a steady increase of one level each time and increasing diamond count by the same amount each trial?

  • The shape_count setting overrides the variant setting when determining how many jewel shapes are shown in the first trial - so the first trial for Jewels B is always actually a "Jewels A" trial with one shape.
  • There are only 8 shapes but there is no check to ensure shape_count falls within 0-8 so setting it to 50, for example, leads to issues.
  • There is no check to ensure that diamond_count is 25 or less for the first trial, so if it is set to 50, for example, the first trial has 50 jewels, and then the second has 25
  • The way that the number of shapes shown for Jewels B is similar in complexity to the way the number of jewels changes - but the limit here is 4 so this is less critical. Still, it would be nice to be able to change the ceiling to ensure, for example, the number of shapes never increases beyond 2 - currently there doesn't seem to be a way to do this.
  • There are no instructions which makes the task significantly more confusing.

[lukeoftheshire]

There are 4 settings that control how much time participants have for each test:

'beginner_seconds': 95,
    'intermediate_seconds': 30,
    'advanced_seconds': 25,
    'expert_seconds': 15,

However, there is no code that can change the difficulty during the game itself that I can see and mode itself is a seperate setting that appears to only control which of these seconds keys is used. Additionally, neither the mode variable nor the actual time allotted is stored in the ActivityEvent data.

I think we should consider removing these 5 variables and replacing them with a single time_allotted or similar variable. Additionally, I think it's important to store the time in static data in case either the time or the difficulty setting is later changed.

[lukeoftheshire]

Once this is resolved, the page at https://docs.lamp.digital/data_science/data_types/activity_types should be updated to reflect any changes.