Erdos source is dead
The below reference:
Vazsonyi, Andrew (December 1998 – January 1999). "Which Door Has the Cadillac?" (PDF). Decision Line: 17–19. Archived from the original (PDF) on 13 April 2014. Retrieved 16 October 2012.
... which seems to be the reference for Erdos, seems to have been scrubbed.~2025-35048-75 (talk) 21:18, 20 November 2025 (UTC)
Actual computer simulation
I had trouble myself grasping this problem intuitively so I created a computer simulation, that, I think, is quite easy to understand—not just the result but the actual code—it's written in Scala 3 (using a clear functional programming style that should be understandable even by non programmers) and is available here: https://github.com/2072/MontyHallProblemSimulation under an MIT license.
In the main article "solutions by simulation" are often mentioned but I did not find any referenced (I'm not sure if this is the right place to put this but here it is) but since it helped me understand the problem once and for all I guess it might help others too. Inogan (talk) 00:41, 24 November 2025 (UTC)
Fallacy
Today I find this in the article of which this is the Talk page:
QUOTE: What is the probability of winning the car by always switching? What is the probability of winning the car by switching given the player has picked door 1 and the host has opened door 3? UNQUOTE
But those are not remotely equivalent questions. The Monty Hall Problem is the computation of your TOTAL odds of winning by an "always switch" strategy over EVERY possible permutation of the game (each permutation being weighted in the probabilities by the likelihood that it will come up). That doesn't allow for stipulating things like "it's a given that the player first picked Door 1" and "it's a given that the host then opened Door 3". Where do those "givens" come in. For instance, it's never a given that the Contestant will pick Door 1. The odds of any one selected permutation being a win or a loss have nothing to do with the odds pertaining to the sum of all possible permutations.
The 2nd question above is about as relevant as asking "What is the probability of winning by always switching in the cases where switching causes the Contestant to lose?" Well, yes, they're 0%, but you've limited the test to cases where the Contestant loses by switching, so, why would it NOT be 0%?~2026-20619-07 (talk) 20:48, 3 April 2026 (UTC)Christopher Lawrence Simpson
- 1. "But those are not remotely equivalent questions" → Indeed; and it fact this is exactly the point of that section: "the distinction between [these questions] seems to confound many" and "the following two questions have different answers".
- 2. " Where do those "givens" come in." → See e.g. Law of total probability.
- Malparti (talk) 01:08, 4 April 2026 (UTC)
Problem statement is not complete
The italic text giving the problem statement is not complete. In order for the MH problem to have the solution "switching wins 2/3 of the time", the rules of the game must be made clear to the contestant prior to the game starts. That is, the contestant must know that the host will always open a door, no matter the outcome of the contestant's initial guess. If the host only reveals a door when the contestant was initially correct, the a switching strategy looses with 100% probability.
This is a frequent mistake of phrasing the MH problem. The "standard assumptions sorts this out, but I think it could be given already in the formulation. — Preceding unsigned comment added by 2A00:1310:202:3013:0:DDDD:1:5 (talk) 07:42, 24 March 2025 (UTC)
- Yes, this is a valid point. I have made some edits to address this problem. Specifically, there is a sharp gap between the probability problem (i.e., here is a probability puzzle) and the decision-theory problem (i.e., what should a rational contestant do), because the contestant's state of belief is not captured in the likelihood model of the problem. If the contestant doesn't know the rules, then a whole bunch of possibilities are "rational" depending on the contestant's assumptions/beliefs. Gill's paper actually sets this up in a fascinating way as a Nash equilibrium problem for an adversarial game between the contestant and host. He does ultimately arrive at the 2/3 result, through a rather different set of assumptions. (And I think just personally, his model is unconvincing from a purely decision theory perspective, but still very interesting. E.g., I don't think Gill's assumptions are any more realistic than those of the "standard" problem.) To be clear, those assumptions are different from the standard setup in important ways, and may also be different from those that the contestant has reason to believe. Sławomir Biały (talk) 09:18, 8 April 2026 (UTC)