However, there are some problems with the simplex method, such as that if there are a number of local minima, the simplex could end up in any one of them, and due to its methods of finding the minimum point, once it is in a dip, it would be impossible to get back out because any direction it goes in would result in a worse fit of the model. Also, unless the starting point is specified to the simplex, it will start at a random point, and each time it runs, it may result in different parameter estimates each time if it has found different minima, leading to problems fitting the curve. This problem can also be the solution however, as the simplex can be started with different starting paremeters a number of times, and the ones that appear most often can be taken as the best set of parameters. However, the results will not ever show for certain whether the global minimum has been found, but the more sets of starting parameters run will just make it more likely.