Abstract
I show that frequent batch auctions for stocks have the potential to reduce the severity of stock price crashes when they occur. For a given sequence of orders from a continuous electronic limit order book market, matching orders using one-second apart batch auctions results in nearly the same trades and prices. Increasing the time interval between auctions to one minute significantly reduces the severity stock price crashes. In spite of this and other advantages pointed out in the literature, frequent batch auctions have not caught on. There is a need for carefully designed market experiments to understand why and what aspect of reality academic research may be missing.
Using Deep Learning for price prediction by exploiting stationary limit order book features
