跪拜 Guibai
← All articles
Algorithm

The Secretary Problem Says Men Should Lock In a Partner by 24.3

By AsherJ ·
Read original on juejin.cn ↗ Google Translate ↗ Alt translation

The 37% rule is a well-known algorithm for hire/no-hire decisions, but applying it to a fixed personal timeline makes the cost of delay concrete. Every year past 24.3 that someone keeps searching raises the odds they settle for whoever is left when the window closes.

Summary

The optimal stopping algorithm known as the secretary problem sets the observation period at 37% of the total window. For a man seriously dating from 18 to 35, that means spending the first 6.3 years building a baseline without committing, then immediately choosing the first person who beats everyone seen before. The math lands the inflection point at 24.3 years old, which matches the folk wisdom that men should find a partner before 25.

Simulations across 20,000 trials show the 37% rule succeeds 38.5% of the time, compared to roughly 5% for picking randomly, picking the first candidate, or waiting until the last. Parameter sweeps confirm the success-rate peak sits squarely in the 36–42% skip range, corresponding to ages 24.1–24.8. The curve collapses quickly on both sides: too little observation means no reliable baseline, and too much means the best candidate has already passed.

Several real-world factors can shift the exact age. Non-uniform candidate pools, the possibility of going back to an ex, and lowering standards from "best" to "good enough" all pull the decision point earlier. But for any reasonable estimate of total serious partners (N between 5 and 30), the optimal skip ratio stays locked between 33% and 40%, keeping the inflection point in the 23–25 range regardless.

Takeaways
With a dating window of 18–35 (17 years), the 37% rule dictates a 6.3-year observation-only phase, ending at 24.3 years old.
After 24.3, the optimal strategy is to commit to the first person who exceeds everyone met during the observation phase.
Monte Carlo simulations (N=20, 20,000 trials) give the 37% rule a 38.52% success rate versus ~5% for random, first-pick, or last-pick strategies.
Parameter sweeps show the success-rate peak spans a skip ratio of 36–42%, corresponding to ages 24.1–24.8.
Using a 50% observation period instead of 37% drops the success rate to 35.4% because the baseline is either too weak or the best candidate is already gone.
Lowering the target from "best" to "good enough" (satisficing) shortens the observation period and pulls the decision point closer to age 22.
As long as the total number of serious candidates N falls between 5 and 30, the optimal skip ratio stays in the 33–40% range, pinning the inflection point at 23–25.
Conclusions

The folk advice to "find a partner before 25" has a precise mathematical basis that most people giving it cannot articulate.

The 37% rule's success rate of ~38% sounds modest, but it is 7.5 times better than any naive strategy, which is a massive edge in a one-shot life decision.

Waiting longer does not increase the chance of finding someone better; it increases the probability that the best option has already appeared and been passed over.

The model exposes a harsh asymmetry: the cost of too little observation is a bad baseline, but the cost of too much observation is permanent loss of the best candidate.

Satisficing versus optimizing is not just a preference; it shifts the mathematically optimal decision point by 2–3 years.

Concepts & terms
Secretary problem (37% rule)
An optimal stopping problem where an interviewer must decide immediately after each candidate whether to hire, with no recalls. The mathematically optimal strategy is to reject the first 37% (1/e) of candidates to establish a baseline, then hire the next candidate who is better than all seen so far.
Satisficing
A decision-making strategy coined by Herbert Simon that aims for a satisfactory or "good enough" result rather than the optimal one. In mate selection, satisficing shortens the observation period because the threshold for acceptance is lower.
Monte Carlo simulation
A computational technique that uses repeated random sampling to estimate the probability of different outcomes. Here, it is used to empirically verify the 37% rule's success rate across thousands of simulated dating scenarios.
Source: juejin.cn ↗ Google Translate ↗ Backup ↗