For this reason, a historian friend of mine really only likes history you can do with statistical regressions, because the regression includes a calculation of the counterfactual, i.e., of the likelihood that your variables do not matter. But we need not go so far as that to acknowledge that any argument would be strengthened by making the counterfactual explicit, and thereby assessing the degree to which a particular event or personality mattered. For example, if you argue that Truman's racism influenced the decision to drop the atomic bomb on Japan, you might consider making an explicit and plausible argument that had FDR survived a few months longer, things would have turned out differently.
My personal favorite category of counterfactuals have to do with Reconstruction's failure, which has been supposed to have so many causes. Lots of us think it had something to do with Andrew Johnson's terrible policies on assuming office. (See e.g. Michael Perman, Reunion without Compromise.) What if the Lincoln assassination attempt had turned out the other way around, with Lincoln surviving and Seward dying? Then you would have the war-hero president with Seward's bloody shirt to wave. Worth a few minutes' thought, I should think.
In the London Review of Books you can find philosopher Slavoj Zizek on counterfactuals, history, and the tedium of overdetermined arguments.