We present the Log Analyzer for generating CPM-GOMS models from human performance data. Built on top of the SANLab tool for stochastic CPM-GOMS modeling (Patton & Gray, 2010), the Log Analyzer uses event-driven parsing to map experimental log files into SANLab interactive routines used to generate CPM-GOMS activity networks. Identical models within and across participants are averaged to obtain estimates of performance times and variability, which are then used to drive stochastic simulations. In this report, we apply our tool to human data collected during a simple eyetracking calibration task and compare the resulting models to existing models in the literature. The generated models show good predictive performance and raise questions about different strategies not captured in the literature. Keywords: stochastic, human performance modeling, CPM-GOMS, automated protocol analysis Copyright 2012 by Human Factors and Ergonomics Society, Inc. All rights reserved.