aditijc/snooker-testbed-phase5-main-v1

这是一个斯诺克台球测试床的数据集，记录了第五阶段主要运行的实验结果。数据集包含20行数据，每行有8个特征列，包括全局PPO时间步、课程阶段、平均得分、最高得分、平均击球数、平均效率、平均犯规率和回合数。实验使用了PPO算法（多离散动作空间）和特定的超参数设置，如总时间步数为500万，学习率为0.0003等。实验结果显示，平均得分为5.39，最高得分为28，犯规率最低为89.3%。数据集还记录了犯规乘数从1.0到1.5的渐变过程及其对犯规率的影响。

This is a dataset for a snooker testbed, documenting the results of the Phase 5 main run. The dataset contains 20 rows with 8 feature columns, including global PPO timestep, curriculum stage, mean score, max score, mean shots, mean efficiency, mean foul rate, and episodes. The experiment used the PPO algorithm (MultiDiscrete action space) with specific hyperparameters, such as total timesteps of 5M, learning rate of 0.0003, etc. Results show a mean score of 5.39, a peak score of 28, and a foul rate floor of 89.3%. The dataset also records the impact of a foul-multiplier ramp from 1.0 to 1.5 on the foul rate.