Best Fit vs First Fit

Why might first-fit allocation outperform best-fit in practice, despite best-fit minimizing wasted space per allocation?