Combination of expert opinion is frequently used to produce estimates in software projects. However, if, when and how to combine expert estimates, is poorly understood. In order to study the effects of a combination technique called planning poker, the technique was introduced in a software project for half of the tasks. The tasks estimated with planning poker provided: 1) group consensus estimates that were less optimistic than the mechanical combination of individual estimates for the same tasks, and 2) group consensus estimates that were more accurate than the mechanical combination of individual estimates for the same tasks. The set of control tasks in the same project, estimated by individual experts, achieved similar estimation accuracy as the planning poker tasks. However, for both planning poker and the control group, measures of the median estimation bias indicated that both groups had unbiased estimates, as the typical estimated task was perfectly on target.