After prodigious work over several decades, binary black hole mergers can now be simulated in fully nonlinear numerical relativity. However, these simulations are still restricted to mass ratios q = m2/m1 > 1/10, initial spins a/M < 0.9, and initial separations r/M < 10. Fortunately, analytical techniques like black-hole perturbation theory and the post-Newtonian approximation allow us to study much of this region in parameter space that remains inaccessible to numerical relativity. I will use black-hole perturbation theory to establish a fundamental upper limit to the final spin that can be attained through binary mergers, and show how this limit can be used to improve predictions of final spins for finite mass ratios as well. I will also show that post-Newtonian inspirals between 1000 M < r < 10 M can align or anti-align black hole spins with each other, dramatically changing the distributions of final spins and recoil velocities that would be expected in astrophysical black hole mergers.