1: \begin{abstract}

2: Best and Chebyshev approximations play an important role in

3: approximation theory. From the viewpoint of measuring approximation

4: error in the maximum norm, it is evident that best approximations

5: are better than their Chebyshev counterparts. However, the situation

6: may be reversed if we compare the approximation quality from the

7: viewpoint of either the rate of pointwise convergence or the

8: accuracy of spectral differentiation. We show that when the

9: underlying function has an algebraic singularity, the Chebyshev

10: projection of degree $n$ converges one power of $n$ faster than its

11: best counterpart at each point away from the singularity and both

12: converge at the same rate at the singularity. This gives a complete

13: explanation for the phenomenon that the accuracy of Chebyshev

14: projections is much better than that of best approximations except

15: in a small neighborhood of the singularity. Extensions to

16: superconvergence points, spectral differentiation, Chebyshev

17: interpolants and other orthogonal projections are also discussed.

18: \end{abstract}