You have an optimization problem, you have no constraints, you cannot compute derivatives, evaluating the function costs a bundle, and you want to optimize with the least amount of effort. What solver should you use? This is a reasonable question, but the answers that you get from the professional optimizers vary all over the place.

In this presentation we attempt to settle this question. The presentation is based on joint work with Stefan Wild on developing tools for benchmarking optimization solvers.

We provide the background for this research, and an overview of current work. Our conclusions are controversial, so (constructive) comments from the audience will be appreciated.