To simulate chance occurrences, a computer can’t literally toss a coin or roll a die. Instead, it relies on special numerical recipes for generating strings of shuffled digits that pass for random numbers. Such sequences of pseudorandom numbers play crucial roles not only in computer games but also in simulations of physical processes.
Researchers have long known that the use of particular methods for generating random numbers can produce misleading results in simulations. In one famous case in 1992, physicists discovered that even “high-quality” random-number generators, which pass a battery of randomness tests, can yield incorrect results under certain circumstances.
At that time, Alan M. Ferrenberg, a computational physicist at the University of Georgia, and his coworkers were interested in simulating the so-called Ising model, which features an abrupt, temperature-dependent transition from an ordered to a disordered state in a system in which neighboring particles have either the same or opposite spins.