Skip to content
Omnikalc

Secretary problem calculator: how many candidates should you reject first?

Enter the number of candidates and the number you reject to get your chance of picking the best and the optimal number to reject.

Updated on
Your inputs

Value out of range

Value out of range

Results

Chance of picking the best with your rule

37.1 %

Close to the best possible rule
Best number to reject (about n ÷ e)
37
Chance of picking the best with that rule
37.1 %
Chance by choosing at random
1.00 %
How many times better the best rule is than a random choice
37.1

Hiring, house hunting and dating all share a trait: you meet options one by one and must decide before you see the next. The secretary problem calculator takes the number of candidates and the number you reject first, and gives the chance of picking the best with that rule, the best number to reject, the chance with that best rule, the chance of a random pick and the advantage of the rule.

How the calculation works

  • Enter the number of candidates who will come one at a time in a random order, and who you must accept or reject on the spot, with no way back.
  • Enter the number of candidates you reject without choosing, only to learn what the field looks like.
  • After that, you pick the first candidate better than all those seen so far. The chance of landing the very best is (r ÷ n) × (H(n−1) − H(r−1)), where H is the harmonic number, computed here by a close approximation.
  • The calculator also gives the best number to reject, about n ÷ e, and the matching chance, which tends to 1 ÷ e = 36.8 % for large n.

Example

You will interview 100 candidates, one after the other. You decide to reject the first 37 and then take the first one who beats all of them. The chance that this picks the single best candidate is 37.1 %. The best number to reject is round(100 ÷ e) = 37, which gives the same 37.1 %, 37 times the 1 % chance of choosing at random. Rejecting only 10 candidates would drop the chance to about 23 %, and rejecting 70 to about 25 %.

Why it matters

The result shows that a little discipline pays: learning from a first batch and then committing beats choosing at random by a wide margin. It also shows the cost of stopping too early or too late, which is steeper than most people expect.

Good to know

The model has strict rules: random order, no recall, and a payoff only for the best. If you accept a good candidate, or can go back, the best strategy changes. Treat the 37 percent rule as a lesson about exploring and exploiting, not as an instruction for choosing a partner or a hire.

The Monty Hall calculator is another classic puzzle on decisions, the expected value calculator lets you compare payoffs, the birthday paradox calculator shows another surprising probability, and the probability calculator handles basic chances.

Frequently asked questions

What is the secretary problem?

It is a puzzle of optimal stopping: candidates arrive in a random order, you see each one once, you can rank it only against those already seen, and you must accept or reject on the spot, with no recall. The goal is to maximise the chance of choosing the single best candidate.

What is the 37 percent rule?

The best strategy is to reject roughly the first n divided by e candidates, about 37 percent of them, and then take the first one who is better than all those seen. For large n, this strategy picks the best candidate with a probability close to 1 divided by e, about 36.8 percent.

Does it apply to real choices?

It is a model, and real life seldom fits it. It assumes you only care about the very best, that you can rank candidates but not score them, that the order is random and that rejected options are gone for good. Looser versions of the problem give different rules, such as choosing a good option with a higher chance.

What about small n?

For small groups, the best number to reject is a bit different and the success chance is higher, since there is less to compare. With 10 candidates, rejecting 3 gives about 40 percent. The calculator uses a close approximation of the sum, so small values can differ from the exact one by a fraction of a percent.

Was this calculator helpful?

Thanks for your feedback!

Results are estimates provided for information purposes. They do not replace professional advice.