Good day
I am coding the plant defence game environment in Python. As such, I have mostly relied on the two links below for details:
- https://mconsultingprep.com/mckinsey-problem-solving-game-digital-assessment
- https://www.jobtestprep.com/mckinsey-problem-solving-game#:~:text=Plant%20Defense%20is%20a%20turn,as%20many%20turns%20as%20possible.
The reason I am coding up this environment is to apply a value-based reinforcement learning algorithm such as neural fitted Q-iteration to it. This will determine near optimal game-play if the algorithm converges.
It is crucial that the environment be as close as possible to the real thing. Hence, I would like to ask users that have played the actual game whether they could clear up the following questions:
- When there are more than one invader, do they all move at the same time in a given turn? Or does one of the invaders move for every turn?
- It would seem like the snake only covers the block it is in while the eagle/falcon covers the block it is in as well as North, East, South and West. Is this correct for the falcon? What range have you observed for the wolf?
- Apparently, a wolf does damage of 60 and an eagle does damage of 20? Are these correct, and what have you experienced for the snake? I have currently set its damage to 50.
- Is the ground-hog immune to snake attacks?
- About how much of each defender and terrain item did you receive at the beginning?
- How many cliffs would you say exist at the beginning of the game, i.e. how many squares are occupied by cliffs at the start.
I know that some of the questions are quite a big ask. Even if you could answer just one of the questions it would greatly enhance the performance of the reinforcement learning policy.
Disclosure: the algorithm will most likely not be used on the actual game. That would seem a bit unethical. It would be interesting, however, to see what score is possible and whether the algorithm can score better than some of the top human players.