| Strategy | Expected amount | Mean amount |
|---|---|---|
| Always stay | - | - |
| Always switch | - | - |
| Switch when it helps | - | - |
Two envelopes paradox
You are given two envelopes. One holds twice as much as the other. You pick one. Should you switch? The tempting argument says yes, because the other envelope seems to average 1.25 times yours. That argument treats half and double as equally likely after the amount is fixed, and they are not. Pick a rule for how the amounts were chosen, open an envelope, and see when switching helps, when it hurts, and when it changes nothing. One envelope holds twice the other. Choose how the amounts were picked, then pick an envelope. Leave it closed, or open it and see whether staying or switching has the higher expected amount. Run 10,000 pairs to compare always staying, always switching, and switching only when the opened amount says so.
How it works
- Choose how the amounts were picked: one fixed pair, any smaller amount in a range, or powers of two.
- Pick an envelope. Leave it closed, or open it and read whether to stay or switch.
- Run 10,000 pairs and compare always staying, always switching, and switching only when the opened amount says so.
The tempting argument
Suppose your envelope holds X. The other then holds either X/2 or 2X. If those were equally likely, the other envelope would average (X/2 + 2X) / 2 = 1.25X, so switching would always look better. The mistake is the phrase "equally likely". Once X is a particular amount, half and double are two different ways the pair could have been chosen, and those ways need not have the same probability.
What a definite rule shows
If the pair is always the same two amounts, opening the envelope tells you which one you hold: switch only if it is the smaller. If you do not open it, staying and switching have the same expected amount. If the smaller amount is drawn from a range, switching helps for small amounts and hurts for large ones, and the gains and losses cancel when you switch blindly. With the powers-of-two rule, every amount you can open makes the other envelope look better, but the average amount is infinite, so there is no finite gain to collect.
Frequently asked questions
Should you switch envelopes?
Not as a rule. With the envelopes closed, switching does not change the expected amount. After you open yours, it depends on how the amounts were chosen and on the amount you see. Sometimes you should switch, sometimes you should stay.
What is wrong with the 1.25 times argument?
It treats "the other envelope has half" and "the other envelope has double" as equally likely after you already know your amount. Those are not equally likely in general. The probability of each one depends on a rule for how the smaller amount was chosen.
Can switching look better for every amount and still not help?
Yes, when the average amount is infinite. For the powers of two used here, the other envelope's expected value is 1.1 times yours for every amount above the smallest, and double for the smallest. Both envelopes still have the same infinite expectation, so 1.1 times is not a gain you can count on. A sample average jumps whenever a rare large pair appears.
Are my numbers sent anywhere?
No. The envelopes are dealt in your browser, and your last settings stay in this browser's local storage.
Ready to try it?
Open PictorX in your browser — no install, no credit card required to start.
Go to Editor