I show how to combine large numbers of forecasts using several approaches within the framework of a Bayesian predictive synthesis. I find techniques that choose and combine a handful of forecasts, known as global-local shrinkage priors, perform best.