I like Ryan’s paper and (obviously) I think complexity matters for behavior. But I think Florian’s updating too much from one experiment. I want to see replications and results from ‘nearby’ designs before saying the lottery anomalies have nothing to do with risk.