Earthquake early warning (EEW) systems can quickly identify the beginning of a significant earthquake rupture, but the first seconds of seismic data have not been found to predict the final rupture length. We present two approaches for estimating probabilities of rupture length given the rupture initiation from an EEW system. In the first approach, bends and steps on the fault are interpreted as physical mechanisms for rupture arrest. Arrest probability relations are developed from empirical observations and depend on bend angle and step size. Probability of arrest compounds serially with increasing rupture length as bends or steps are encountered. In the second approach, time‐independent rates among ruptures from the Uniform California Earthquake Rupture Forecast, Version 3 (UCERF3), are interpreted to apply to the time‐dependent condition in which rupture grows from a known starting point. Length probabilities from a Gutenberg–Richter magnitude–frequency relation provide a reference of comparison. We illustrate the new approach using the discretized fault model for California developed for UCERF3. For the case of rupture initiating on the southeast end of the San Andreas fault we find the geometric complexity of the Mill Creek section impedes most ruptures, and only are predicted to reach to San Bernardino on the eastern edge of the greater Los Angeles region. Conditional probabilities of length can be precompiled in this manner for any initiation point on the fault system and thus are of potential value in seismic hazard and EEW applications.