That's it. Win and losses in prior games don't necessarily matter to their algorithm as much as net offensive and defensive efficiency weighted towards strength of schedule - with a healthy amount of prior seasons' results baked into the algorithm as well.
It all makes for a tool that is fairly good at predicting results int he aggregate, but it generally discounts good performances by teams that were bad the previous season(s) and props up teams that have struggled a bit but historically (over the past season or two) have been very good.