We present an algorithm to compute the (pre)-kernel of a TU-game N, v with a system of n 2 1 linear programming problems. In contrast to the algorithms using convergence methods to compute a point of the (pre)-kernel the emphasis of the chosen method lies not on efficiency and guessing good starting points but on computing large parts or in good cases the whole (pre)-kernel of a game. The chosen algorithm computes on a first step by relying on linear programming the n 2 largest amounts