Phys. Rev. E 98, 020301(R) (2018)

When making decisions in a competitive environment, access to comprehensive prior information should intuitively confer an advantage. This is the rationale behind the accumulation of big data: we expect that the more information we store, the better our predictions.

As Vijayakumar Sasidevan and colleagues have now shown, the relationship between information availability and prediction accuracy is in truth much subtler. They considered an adaptive system composed of agents that compete for a scarce resource, making decisions on the basis of previous outcomes. By varying the ratio between agents that have access to good or poor data — in terms of the length and resolution of their record — and the details of their strategy, they concluded that better information availability doesn’t always lead to a larger payoff.

The origin of this phenomenon, the authors argue, resides in the collective information that is available only at certain level of coarse-graining of the data, which might vanish at higher resolution.