Binghamton Team's Shannon Entropy Wordle Tactic Hits 99% in Simulations but Requires Software Assistance

The approach treats each guess as an entropy query that partitions the remaining word list by color-pattern outcomes. In simulations across the full 12,972-word solution space, the method reduced average guesses needed compared with frequency-based starters like SLATE. Unlike static lists of optimal openers, it recomputes the highest-entropy candidate after every feedback vector, producing recommendations that often look statistically unlikely to human players.

Existing public solvers already use similar decision trees or reinforcement learning, yet the Binghamton work explicitly frames the problem in Shannon entropy terms and quantifies the gain from non-probable guesses. This connects to earlier Mastermind analyses from the 1970s and recent arXiv preprints on Wordle that reached 99%+ success with different heuristics. The paper understates that real-time deployment still demands an external program because humans cannot compute conditional entropy on the fly.

A key limitation is the simulation-only design; no human-subject trials measured whether players can translate the suggestions into faster solves under time pressure. Larger validation would require testing the algorithm against daily Wordle archives with actual participants and logging both solve rate and cognitive load.

Future work could embed the entropy calculator in a browser extension that reads the game state automatically, lowering the barrier for non-programmers and allowing field experiments on whether the strategy improves completion rates outside controlled simulations.