We revisit the optimal weights for the weighted log-rank test for nonproportional hazards\u00a0data. It is noted that the optimal weight function can be derived by assuming a stable\u00a0distribution for an exponentiated omitting covariate from the proportional hazards model,\u00a0which induces the nonproportionality. A special case is the weight function for the popular\u00a0Harrington-Fleming\u2019s G test statistic. However, in practice it is not straightforward for investigators\u00a0to determine the optimal value of the tuning parameter for the weight function\u00a0in the G test statistic. We propose a maximum likelihood method to estimate the parameter\u00a0from the observed data, noticing that the parameter is inversely related to the index\u00a0parameter from the gamma distribution commonly assumed for the frailty model. The simulation\u00a0results indicate that the test statistic with the estimated weight function from the\u00a0data are more powerful than the commonly used Harrington-Fleming test with = 1. We\u00a0also propose a different weight function that possibly gives more power than existing ones\u00a0to detect middle difference. Three datasets from phase III clinical trials on breast cancer\u00a0are illustrated as real examples.<\/p>\n