Building an AI lineup optimizer for a Discord esports bot (the algorithm, not the hype)

Short summary

The author built a lineup optimizer for their Discord esports bot using classical constraint satisfaction, not an LLM. The system enumerates feasible player assignments, scores them by role coverage, availability confidence, and starter priority, then returns the best lineup in under a second. The key insight: LLMs can violate hard constraints, so they should handle unstructured input while a solver handles the actual optimization.

• •Lineup optimization is a constraint satisfaction problem, not an LLM problem
• •Solver enumerates candidates and scores by coverage, confidence, and priority
• •LLMs should parse input, not make assignment decisions

Generated with AI, which can make mistakes.