这是最优停止问题（Optimal Stopping）的经典策略：在不可回头、按顺序出现的选择中，先观察前约 37% 的样本，只记录当前最优；之后遇到第一个超过历史最优的选项，就停止等待并行动。
This comes from the classic optimal stopping problem: when options arrive sequentially and you cannot go back, observe roughly the first 37% without committing, remember the best-so-far, then take the first later option that beats it.
观察前 -- 个选项，只记录，不承诺。
Observe the first -- options without committing.
过了第 -- 个后，遇到第一个超过历史最优的选项，就停止等待并行动。
After option --, take the first option that beats your best-so-far.
记录你已经评估到第几个选项。如果已过阈值，开始执行 Leap Rule。
Record which option number you are currently on. If you've passed the threshold, begin the Leap Rule.