As van der Pauw noted in his original paper, the incisions in the cloverleaf shape "substantially reduce the error due to the finite dimensions of the contacts" compared to simple circle. A cloverleaf is effectively a small central circular sample with four large contact pads, and you can be a little less precise in where connect to those pads.
According to Koon, to achieve resistivity measurements with 1% errors, a typical cloverleaf is about 4 times more forgiving than a simple circle for small displacements of the contact points on the edge of the sample. Even a simple square sample with contacts at the four corners is about 3 times more forgiving than a simple circle.
For 1% Hall coefficient measurements, the square is actually the best, being about 12 times more forgiving than the circle, while the cloverleaf is only about 7 times more forgiving.