Let’s suppose the player picks a door, then Monty has to open another door and offer a “stay or switch?” choice. But let’s hold off for now on pinning down exactly what the process is by which Monty chooses which door to open.
On any given occasion, one of three things happens: [ul]
[li]The player initially picks the prize door, then Monty opens one of the goat doors. Let’s call the proportion of times when this happens A.[/li][li]The player initially picks a goat door, then Monty opens the other goat door. Let’s call the proportion of times when this happens B.[/li][li]The player initially picks a goat door, then Monty opens the prize door. Let’s call the proportion of times when this happens C.[/li][/ul]
The proportion of time when Monty opens a goat door is A + B. Out of those times, the relative frequencies of the stay door having the prize vs. the switch door having the prize are A to B. [If the only information you have is this probabilistic model and that Monty opened a goat door, and your goal is to make the choice with the highest probability of getting the prize, then you should stay or switch according as to whether A or B is bigger].
But we haven’t actually answered the question: Out of A and B, which is bigger?
In fact, we haven’t said much at all yet about what A, B, and C are.
Of course, one constraint is A + B + C = 1. Typically, we assume the player’s initial choice has no particular correlation to the location of the prize and goats, so that A = 1/3 while B + C = 2/3.
But we haven’t pinned down exactly what B and C are beyond that. If B accounts for more than half of B + C, then B will be larger than A. If B accounts for less than half of B + C, then B will be smaller than A. If B accounts for precisely half of B + C, then B will equal A.
So there you go.
If the player initially picks a door at uniform random and Monty then opens a non-player goat door whenever possible, then A = 1/3, B = 2/3, C = 0, and, over all the cases where Monty opens a goat door, the stay door has the prize half as often as the switch door.
If the player picks a door at uniform random and Monty then opens a non-player door at uniform random, then A = 1/3, B = 1/3, C = 1/3, and, over all the cases where Monty opens a goat door, the stay door has the prize equally as often as the switch door.