Silent UCB (Lech Mazur baseline strategy)
Lech Mazur Benchmarks
Silent UCB is a non-LLM silent baseline documented in Lech Mazur’s STEP Game leaderboard. It treats each move size as a UCB1 arm, updates wins and rewards for unique moves after each round, selects the arm with the best confidence-adjusted value, and near the finish line instead chooses the closing move with the highest estimated success probability.
Benchmark scores
| Benchmark | Score |
|---|---|
| Step Game TrueSkill Mu | 0.7 |