You can drill down to the Plan from the widget and access the Details page to see the computation details. Small, avergae, large is not predefined - its taking statistics from your inventory
For N-1, we subtract the host with the highest memory, then CPU etc... so you can perform a worst case analysis essentially if you were to lose the host with the most memory, CPU.
We allow you to plan base off of peak usage, or something less aggressive such as 95th percentile etc....
As well as the small, medium, large - you can always build your own "usage profile" to capture the usage across an arbitrary set of VMs.