My Python-based Fantasy Premier League optimiser already generates legal transfer plans and scores them with a well-tested evaluation function. Where it falls short is in how close those scores get to the true mathematical optimum over a multi-week horizon. I’m convinced the core logic can do better if its search strategy is opened up: less aggressive pruning, smarter node ordering, and broader exploration without exploding run time. Right now the solver takes around 20 seconds on my cloud setup; I’m happy for this to increase up to roughly 90 seconds if it buys significantly better solutions. I’d like you to dig into the existing code, keep the current evaluation function intact, and enhance the algorithm that drives the search. I’m not looking to swap in a totally new meta-heuristic, but rather to refine what is there—revisit pruning and termination criteria, improve how states are ranked and expanded, and expand the way the tree/graph is walked so higher-scoring plans aren’t left unexplored (e.g. better beam search, memoisation, caching of states, etc.). You’ll deliver: An updated solver module that consistently returns solutions measurably closer to the true optimum on my test scenarios. Inline comments plus a short technical note explaining what you changed and why, including any new parameters I may want to adjust later. A lightweight benchmark script so I can reproduce your before-and-after results (runtime and evaluation score). If you enjoy algorithmic optimisation and can show past wins squeezing extra accuracy out of an existing search routine, let’s talk.