I've already answered your questions with my previous reply, but allow me to elaborate....
Let's look specifically at CPU resource allocation:
Pool 3 = 8000 shares
Pool 2 = 4000 shares
Pool 1 = 2000 shares
Pool 3 is twice as likely to be allocated CPU resource as Pool 2 WHEN there is CPU contention.
Pool 2 is twice as likely to be allocated CPU resource as Pool 1 WHEN there is CPU contention, but only half as likely to be allocated CPU resource as Pool 3 WHEN there is CPU contention
Pool 1 is half as likely as Pool 2 to be allocated CPU resource as Pool 2 WHEN there is CPU content, and only a quarter as likely to be allocated CPU resource as Pool 3 WHEN there is CPU contention
The same logic applies when you consider the VMs inside each of the resource pools, WHEN there is CPU contention - as an example let us assume each of the 12 VMs in Pool 3 has a single virtual CPU then each VM will have 1000 CPU shares:
VM3A in Pool 3 has 1/12 of the total shares of all the VMs in that pool, therefore it is likely to be allocated 1/12th of the available CPU resource of Pool 3 WHEN there is CPU contention.
VM3B in Pool 3 is the same as VM3A
And so on for the other VMs in Pool 3
If any of the VMs in Pool 3 has 2 vCPUs that VM will have 2000 CPU shares rather than 1000, but the same logic is true for each of the individual vCPUs for that VM.
Apply the same logic to the 300 VMs in Pool 2, and the 200 VMs in Pool 1.