Why can an anti-markovnikov radical addition of haloalkane only happen in presence of hydrogen peroxide?

In a "normal" Markovnikov attrition of $\text{HBr}$ to an alkene, the $\text{H}$ adds to the carbon with more hydrogen atoms in order to form the more stable carbocation.
That means that the $\text{H}$ must go on the carbon with fewer $\text{H}$ atoms.