Knock-limited engine operation is one of the most important constraints on fuel efficiency and performance that must be considered during the design, control algorithm development and calibration of spark-ignition engines. Since a complete fundamental explanation of knock phenomenon over the full range of engine conditions is not available, estimating knock onset at different in-cylinder thermodynamic conditions is very challenging. Various approaches have been proposed, either stochastic or deterministic, in an effort to capture the complex phenomena related to engine knock for the purposes of engine control. This research evaluates the accuracy of control-oriented physics-based knock prediction routines under various engine operating conditions. Two common methods of knock prediction, a generalized chemical kinetics model and an induction-time integral, are evaluated and compared against experimental data. The experimental investigation is conducted using a naturally aspirated 3.6L V6 engine, retrofitted with cooled EGR. Data are acquired from spark timing sweeps under knocking conditions at various engine speeds and loads in an engine dynamometer cell. The knock prediction models utilize inputs derived from experimental in-cylinder pressure data. Chemical kinetics model parameters are calibrated based on engine data and the resulting model prediction of knock onset location is compared to experimental observations. Multiple combustion phasing parameters are examined in an effort to provide a threshold of comparison for knock onset timing, and to distinguish between light and heavy knock events. Finally, the effect of cooled EGR on knock onset prediction is assessed for different operating conditions. Evaluating the accuracy and feasibility of these methods for real-time modeling and control of engine knock, neither of the combustion phasing parameters provides a single universal threshold for knock determination. Depending on the threshold, the chemical kinetics model shows better prediction than the induction-time integral method, with a maximum knock borderline prediction error of 3 CAD. However, with increasing EGR levels, this error is augmented.