Solver makes it possible to try many different search strategies fairly easily and without having to change the representation of domain constraints. We urge you to take advantage of this facility to try out different ways of solving a given problem to find the one most advantageous to your situation.

There is no single rule of thumb for discovering what is a good strategy. You should try strategies that:

However, there are two main kinds of strategies: the one used for constraint satisfaction problems, and the one used for optimization problems.

It is our experience that when we consider satisfaction problems, a strategy that is often successful for selecting variables is to choose the most constrained one. In effect, those variables are where propagation is the most significant, and as a consequence, where entire portions of the search tree can be pruned as quickly as possible. A predefined Solver strategy--IlcChooseMinSizeInt--can be most profitably applied in this respect.

When we consider optimization problems, strategies based on generation, such as IloGenerate, often first generate those variables and their values that are statistically the most likely to lead to a low-cost solution.