Two new perturbation simplex variants for solving linear programs without introducing artificial variables are proposed. Each uses the dual pivot rule to achieve primal feasibility and then the primal pivot rule to achieve optimality. The second variant is designed to handle degenerate problems more efficiently. The author claims that the new algorithm outperforms the conventional simplex algorithms with the tested problems significantly.