Text this: Greedy Action Selection and Pessimistic Q-Value Updating in Multi-Agent Reinforcement Learning with Sparse Interaction