This paper develops an empirical behavioral model of price formation using high-frequency data from the London Stock Exchange. The authors analyze order placement and cancellation patterns, uncovering surprising regularities: the probability of placing a limit order at a given price relative to the best quote follows a symmetric Student distribution with less than two degrees of freedom, independent of the bid-ask spread and identical for buys and sells. This contradicts strategic reasoning that would predict asymmetry between orders inside the spread and inside the book. The fraction of orders that result in immediate transactions is accurately captured by the tail of this distribution. For cancellations, lifetimes follow a power-law distribution rather than an exponential, with rates depending on the order's distance to the opposite best price and the imbalance of buy/sell orders in the book. The authors combine these empirical findings with a long-memory model for order signs (based on strategic order splitting) into a stochastic representative agent simulation. Without fitting any parameters to price data, the model successfully reproduces the magnitude and functional form of return distributions and bid-ask spreads, including heavy tails. The work provides a quantitative behavioral alternative to rational equilibrium models, demonstrating that simple statistical regularities in order flow can explain key market properties.