A numerically stable dual algorithm for strictly convex quadratic programming is presented. The algorithm uses the unconstrained minimum of the objective function as a starting point and employs Cholesky and QR factorizations for implementation. Computational results show that the dual algorithm outperforms primal algorithms when a primal feasible point is not readily available. The algorithm is compared to modified-simplex type dual methods and is illustrated with a numerical example. The paper discusses the need for efficient and robust algorithms for solving strictly convex quadratic programming problems, which are important in nonlinear programming. The algorithm is based on projection methods, which are generally more efficient and require less storage than modified simplex type methods. The algorithm is implemented using a phase 1 approach that starts from the unconstrained minimum of the objective function, which can lead to a near-optimal feasible point. Computational testing indicates that this approach usually finds a feasible point that is also optimal, requiring only a few additional iterations in phase 2 to achieve optimality. The algorithm is presented as a projection type dual algorithm for solving the quadratic programming problem, which was first introduced by Idnani. The paper highlights the importance of efficient and robust algorithms for solving quadratic programming problems, particularly in the context of nonlinear programming.A numerically stable dual algorithm for strictly convex quadratic programming is presented. The algorithm uses the unconstrained minimum of the objective function as a starting point and employs Cholesky and QR factorizations for implementation. Computational results show that the dual algorithm outperforms primal algorithms when a primal feasible point is not readily available. The algorithm is compared to modified-simplex type dual methods and is illustrated with a numerical example. The paper discusses the need for efficient and robust algorithms for solving strictly convex quadratic programming problems, which are important in nonlinear programming. The algorithm is based on projection methods, which are generally more efficient and require less storage than modified simplex type methods. The algorithm is implemented using a phase 1 approach that starts from the unconstrained minimum of the objective function, which can lead to a near-optimal feasible point. Computational testing indicates that this approach usually finds a feasible point that is also optimal, requiring only a few additional iterations in phase 2 to achieve optimality. The algorithm is presented as a projection type dual algorithm for solving the quadratic programming problem, which was first introduced by Idnani. The paper highlights the importance of efficient and robust algorithms for solving quadratic programming problems, particularly in the context of nonlinear programming.