Mechanical in-row weed control of crops like sugarbeet require precise knowledge of where individual crop plants are located. If crop plants are placed in known pattern, information about plant locations can be used to discriminate between crop and weed plants. The success rate of such a classifier depends on the weed pressure, the position uncertainty of the crop plants and the crop upgrowth percentage. The first two measures can be combined to a normalized weed pressure, \lambda. Given the normalized weed pressure an upper bound on the positive predictive value is shown to be 1/(1+\lambda). If the weed pressure is \rho = 400/m^2 and the crop position uncertainty is \sigma_x = 0.0148m along the row and \sigma_y = 0.0108m perpendicular to the row, the normalized weed pressure is \lambda ~ 0.40$; the upper bound on the positive predictive value is then 0.71. This means that when a position based classifier predicts that a certain plant is a crop plant 71% of the times it will be correct.