The writhe of a knot diagram is a simple geometric measure of the complexity of the knot diagram. It plays an important role not only in knot theory itself, but also in various applications of knot theory to fields such as molecular biology and polymer physics. The mean squared writhe of any sample of knot diagrams with n crossings is n when for each diagram at each crossing one of the two strands is chosen as the overpass at random with probability one-half. However, such a diagram is usually not minimal. If we restrict ourselves to a minimal knot diagram, then the choice of which strand is the over-or understrand at each crossing is no longer independent of the neighboring crossings and a larger mean squared writhe is expected for minimal diagrams. This paper explores the effect on the correlation between the mean squared writhe and the diagrams imposed by the condition that diagrams are minimal by studying the writhe of classes of reduced, alternating knot diagrams. We demonstrate that the behavior of the mean squared writhe heavily depends on the underlying space of diagram templates. In particular this is true when the sample space contains only diagrams of a special structure. When the sample space is large enough to contain not only diagrams of a special type, then the mean squared writhe for n crossing diagrams tends to grow linearly with n, but at a faster rate than n, indicating an intrinsic property of alternating knot diagrams. Studying the mean squared writhe of alternating random knot diagrams, also provides some insight into the properties of the diagram generating methods used, which is an important area of study in the applications of random knot theory.