Today the findRelevantStates method has been fixed to suit the variable sized states. Due to these differently sized states, some positions of Tetromino will not be encompassed by a SubState. Perhaps this could result in a policy return of -Something to dissuade the agent from going to that location.
Subsequently, many tests were easily resolved due to the method being fixed. Some had stale assertions that no longer applied. One method that struggled to cope was updatePolicy, which required some changing to accommodate the changes. The test for this method is still failing however, as a new assertion has been added which doesn’t have code to pass it. The assertion for updating superstates of smaller substates when updating. Such as updating {-1,1,0} and {1,0,1} when updating a smaller substate of {0, null, null}.
The same will need to go for finding rewards also, which hasn’t been tested properly yet.
Tests passing: 18/19 (17/19 if you count the findPieceRewards).
EDIT: Later on in the day, I got around to the superstate method. Well, writing some basic wall state code and a complete test for the method. Need to fill in the code fully later. Note that states of size 2 can have superstates of size 3 and 4. Perhaps recursion could be used. Haven’t used that in a while!