What are some good ways to estimate the capacity of a strategy from historical data (including full market depth)?
Obviously, a naive approach is that you want the strategy's returns to exceed its market impact. PBs mostly provide their institutional clients with software packages that estimate market impact for large orders. However, many strategies have unique and more intricate entries/exit than large-order-buy-and-hold. Moreover, there are strategies that are low-capacity and contingent on market depth events, both of which I'm guessing require very different underlying assumptions than the market impact models used for large orders.