The Shapley value concept from cooperative game theory has become a popular\ntechnique for interpreting ML models, but efficiently estimating these values\nremains challenging, particularly in the model-agnostic setting. Here, we\nrevisit the idea of estimating Shapley values via linear regression to\nunderstand and improve upon this approach. By analyzing the original KernelSHAP\nalongside a newly proposed unbiased version, we develop techniques to detect\nits convergence and calculate uncertainty estimates. We also find that the\noriginal version incurs a negligible increase in bias in exchange for\nsignificantly lower variance, and we propose a variance reduction technique\nthat further accelerates the convergence of both estimators. Finally, we\ndevelop a version of KernelSHAP for stochastic cooperative games that yields\nfast new estimators for two global explanation methods.\n