Probability Puzzles and Instrument Tricks
Late-night study sessions have a way of drifting from one rabbit hole to another. For several engineering students, that meant revisiting the classic Monty Hall problem—and then discovering that similar counterintuitive tricks run straight through the physics lab.
The Monty Hall Problem, Revisited
The setup is familiar: three doors, one prize. You pick a door. Before the reveal, the host opens a different door that he knows is empty, then offers you the chance to switch your pick. Common intuition says the two remaining doors are equally likely, so switching buys nothing. That intuition is wrong—switching is always the better move.
Initially your selection has a 1-in-3 chance of being right. After the empty door is removed, the choice does become a 50/50 proposition if you randomize again. But a randomized choice isn't optimal. The reason is subtle. Since the host always opens an empty door and never opens the one you picked, the remaining unopened door inherits the combined probability mass of the two doors you didn't choose. Switching therefore carries a 2-in-3 chance of winning, while sticking with your original pick holds only that initial 1-in-3 probability.
There is no penalty for switching since you cannot distinguish between the unopened doors anyway. Picking a specific door each time doesn't change the randomness of the event—it just aligns your selection with the higher-probability outcome.
A Better Spectrum From Nothing But Zeros
The same kind of counterintuitive logic appears in nuclear magnetic resonance (NMR) spectroscopy. A longer sampling time ordinarily improves effective resolution, so a 30-second scan should produce sharper signals than a 5-second acquisition. The complication: the data get noisier as time accumulates. The trick that works in practice is to record only 5 seconds of genuine signal, pad the rest with zeros, and treat the data as if the signal simply decayed away. After the Fourier transform, the resolution is actually better than either the plain 5-second run or the full 30-second dataset—you can distinguish finer peaks by deliberately injecting null data.
Uncertainty as a Feature
Pulse sequences rely on another exploit of the same physics. An NMR instrument can only emit one frequency at a time, but a full spectral range is needed. The solution is to tune the pulse to the center of the desired window and make the pulse so short that, by the Heisenberg uncertainty principle, its frequency becomes unresolvable. In effect, a single-frequency pulse blurs across the entire range, exciting all frequencies at once. The result is a complete data set across the whole band from one shot—an impossible reading if the pulse were long and frequency-precise.
The universe keeps insisting that the rules at the small scale don't map neatly onto everyday intuition. Between game-show doors and spectrometer pulses, the same theme emerges: what seems like a disadvantage can be engineered into a benefit once the underlying probability and quantum behavior is accounted for. Quantum mechanics does indeed beat any tendency toward realism out of you.
