The forced regional pattern of sea level is a key consideration in adapting to sea level rise as it doubles the global mean rate in some regions while offsetting it entirely in others. However the forced pattern in the altimeter record is accompanied by strong internal variability, leaving models as the main tool used in its estimation. Here we show that the current approach for estimating forced patterns of rise from averages across standard-resolution coupled models is likely an underestimate due to two main factors. First, we show that standard-resolution models systematically produce weaker patterns than high-resolution models, which better resolve ocean fronts and eddies. Secondly, using large ensembles we show that model structural uncertainty in forced patterns is substantial, rendering multi-model average patterns susceptible to destructive interference. Together, we estimate using both the altimeter record and large ensembles that these factors lead to an underestimation in the regional pattern of rise by roughly a factor of two, suggesting paths forward for producing improved estimates.