# Minimum Questions Puzzle: Beat Binary Search in 7

*Science & Technology · Priya Nair · 2026-08-31*

## TL;DR

You can guarantee guessing any number from 1 to 1000 in seven questions if you allow three possible answers: yes, no, or "I don't know." Each question divides the remaining possibilities into three roughly equal groups, so seven questions cover up to 2187 numbers, well beyond the 1000 you need.

## Key facts

- The puzzle was popularized by Alex Bellos in his mathematics column, where he challenged readers to beat the standard binary search.
- A standard binary search with only yes or no answers needs ten questions to cover 1 to 1000, since 2^9 is 512 and 2^10 is 1024.
- With three possible answers, each question can split the number line into three parts, so seven questions cover up to 3^7, which is 2187 possibilities.
- The key is crafting questions where you honestly answer "I don't know" for about one third of the remaining numbers, which still gives the asker useful information.
- The optimal strategy is a ternary search, which guarantees a correct guess in seven questions for any number between 1 and 1000.

## The Classic Puzzle Gets a Twist: Meet Osman

Picture this: a friend named Osman picks a whole number between 1 and 1,000, and your job is to pin it down by asking questions. The catch? He can only answer "Yes," "No," or "I don't know." The puzzle here is deceptively simple to state and surprisingly tricky to solve, because that third option changes everything about how you play the game.

Most of us grew up with some version of "think of a number." You ask, "Is it higher than 500?" and narrow things down step by step. With only yes-or-no answers, the math is clean: each question halves the possibilities, and 10 questions guarantee you land on the right integer between 1 and 1,000. That works because 2 to the 9th power is 512, which isn't quite enough, while 2 to the 10th is 1,024, which covers the whole range.

But here's where Osman's third answer, "I don't know," sneaks in and scrambles the familiar approach. In the classic version of this puzzle, which reportedly gained wider attention through a column by Alex Bellos in The Guardian on February 8, 2021, the trick is to craft questions that genuinely allow all three responses. When you do that, each question can split the remaining possibilities into roughly three equal chunks instead of just two.

The shift feels small, but the payoff is huge. Three outcomes per question means 3 to the 7th power equals 2,187, comfortably more than 1,000. Six questions only give you 729 combinations, which falls short. So the answer to this puzzle drops from 10 to 7, and that's the moment Osman stops being a simple guessing game and becomes a genuine head-scratcher.

## Binary Search: The 10-Question Safety Net

Let's bring Osman back and watch him pick a number. Say he settles on 358, somewhere in that wide field between 1 and 1,000. Your first move is simple: ask if his number falls between 1 and 500. He says yes. Just like that, you've cut the possibilities in half.

The trick is to keep doing exactly that. Each question splits whatever range remains into two equal halves, and Osman's yes or no tells you which half to keep. Ask if it's between 1 and 250, and he says no, so now you're looking at 251 through 500. Narrow it again, and again, and again. The rhythm of the questions becomes almost musical: 251 to 375, yes; 251 to 313, no; 314 to 345, no; 346 to 361, yes. Each answer is a small door closing behind you.

Work through the sequence with Osman and you'll land on 358 in exactly ten questions. The final exchange goes like this: is it 355 to 358? Yes. Is it 355 to 356? No. Then the number must be 358. Ten questions, guaranteed, every single time.

Why ten? The math is clean and satisfying. With nine questions, the best you can do is distinguish 2 to the 9th power, which is 512 possibilities. That's not enough for 1,000 numbers. But 2 to the 10th power gives you 1,024 distinct paths, just barely more than you need. So ten yes-or-no questions form a perfect safety net, catching any number Osman might hide.

Notice something about those questions though. They forced Osman into a corner where only "yes" or "no" made sense. He technically had a third option available, the "I don't know" escape hatch, but the way the questions were phrased, he never got to use it. The classic puzzle plays out entirely in that binary world.

That's the elegant baseline. But here's the thing that makes this puzzle genuinely interesting: what if we actually let Osman use that third answer? What if we designed our questions so that "I don't know" carries real information? That single shift changes everything, and it's where the puzzle gets truly clever.

## The "I Don't Know" Secret: How Three Answers Shrink the Range

Here's where the puzzle gets genuinely clever. Instead of forcing Osman into a binary corner, you craft a question where his honest "I don't know" becomes a third, equally useful signal. The trick question goes like this: "I'm thinking of a number between 333 and 666. Is your number smaller than mine?"

Think about what each answer tells you. If Osman says "Yes," his number must be smaller than 333, so it sits somewhere in the 1 to 332 range. If he says "No," his number is bigger than 666, landing it between 667 and 1,000. And if he says "I don't know"? That's the beautiful part. He can't know whether his number is smaller than yours unless his number is actually below 333. If his number were in that lower range, he'd instantly answer yes. If it were above 666, he'd know it's bigger and answer no. The only way he's genuinely uncertain is if his number falls in that middle band between 333 and 666, the exact range you were thinking of.

With that single question, you've sliced the entire 1 to 1,000 range into three roughly equal chunks. Each answer narrows the possibilities to about one-third of what they were. Repeat the same logic with the new, smaller range, always anchoring your question to the middle third, and you'll reach Osman's number in just 7 questions. That's three fewer than the classic yes/no approach requires, all because you gave uncertainty a seat at the table. When you let "I don't know" do real work, the puzzle transforms from a binary hunt into something far more efficient.

## Seven Questions to Victory: A Step-by-Step Example

Let's watch the trick work in real time. Osman has picked the number 680, and we're going to corner it in exactly seven questions using the ternary method. Each question targets the middle third of whatever range remains, and Osman's honest answers do the heavy lifting for us.

The first question splits the whole 1 to 1,000 field into three chunks. "I'm thinking of a number between 333 and 666. Is your number smaller than mine?" Osman says no, so we instantly know his number lives in the 667 to 1,000 zone. That single exchange chopped the possibilities from a thousand down to roughly three hundred and thirty-four.

Next, we aim at the middle third of that new range. "I'm thinking of a number between 778 and 890. Is your number smaller than mine?" Yes. The number is in there. Now we're working with a span of about 113 numbers, and we've only asked two questions. The next probe, "I'm thinking of a number between 704 and 738. Is your number smaller than mine?" gets another yes, narrowing us to a 35-number window.

Here's where the "I don't know" answer earns its keep. We ask, "I'm thinking of a number between 679 and 691. Is your number smaller than mine?" Osman pauses and says, "I don't know." That's not a dodge. It's a precise piece of information. If his number were below 679, he'd know it was smaller than our imagined number and say yes. If it were above 691, he'd know it was larger and say no. Only because his number actually sits inside that 679 to 691 band does he genuinely lack the information to answer. We just cut the field to thirteen numbers with one question.

The fifth question, "I'm thinking of a number between 683 and 687. Is your number smaller than mine?" gets a yes, leaving five candidates. Then "I'm thinking of a number between 680 and 681. Is your number smaller than mine?" brings another "I don't know," which pins the number to that tiny pair. One final question, "Is it 680?" and Osman confirms. Seven questions, done.

Why does seven work when the plain yes-or-no version needs ten? The math is clean. With two answers per question, ten questions cover up to 2^10, or 1,024 possibilities, and nine questions only reach 512, which isn't enough for 1,000 numbers. But with three possible answers, we get 3^7, which equals 2,187, comfortably above 1,000. Six questions would only give us 3^6, or 729, which falls short. So seven questions are both necessary and sufficient for this puzzle, and the middle-third strategy is exactly how you spend that budget.

## Why 7 Beats 10: The Math Behind the Magic

Six questions cannot possibly get you there, and here is the cold arithmetic: 3 raised to the 6th power equals 729, which falls short of 1,000. That means six rounds of questioning, with three possible answers each time, only cover 729 distinct outcomes. You need at least 1,000 possible paths to guarantee you land on the right number between 1 and 1,000.

Seven questions, however, change everything. Three to the 7th power gives you 2,187 possibilities, more than enough room to pinpoint any number in that range. This is the engine behind the puzzle: each question you ask with three possible responses triples your reach, so you only need the smallest number of rounds where 3^k equals or exceeds the size of your range.

The general rule works for any set of numbers. For a range of N possibilities, the minimum questions is the smallest k such that 3^k is greater than or equal to N. For 1 to 1,000, that k is 7. For a smaller set like 1 to 100, you would need 5 questions, since 3 to the 4th is 81 (too small) and 3 to the 5th is 243 (plenty).

This is where the trick diverges from the classic approach most people learn first. A standard yes/no question only splits the possibilities in half, what computer scientists call a binary search. With two outcomes per question, you need 10 questions to cover 1,000 numbers, because 2 to the 9th is 512 and 2 to the 10th is 1,024. That is why the straightforward "is it higher or lower?" method always lands at 10.

The clever twist comes from letting Osman answer "I don't know." That third option turns each question into a ternary split, dividing the remaining range into three roughly equal chunks instead of two. Every question now eliminates about two-thirds of the candidates rather than half. The savings compound quickly: 10 questions shrink to 7, and the same logic scales to any range you can imagine.


## Frequently asked questions

**What is the minimum number of questions to guess a number between 1 and 1000 with yes/no/I don't know answers?**

You can guarantee finding the number in just 7 questions when you allow "I don't know" as a third possible answer. Each question splits the remaining possibilities into three roughly equal groups instead of two. Since 3 to the power of 6 equals 729, six questions can cover at most 729 numbers, which is not enough for 1,000 possibilities.

**How does the three-answer question strategy work?**

The trick is crafting each question so that all three responses point to a distinct third of the remaining range. You ask about a middle segment, and depending on whether the answer is yes, no, or I don't know, you narrow the field to one specific block. This way, every single question eliminates about two-thirds of the remaining candidates.

**Why can't you guarantee finding the number in fewer than 7 questions?**

With only 6 questions and three possible answers each, the total number of distinct response patterns is 3 to the 6th power, which equals 729. Since there are 1,000 possible numbers, at least two numbers would share the same response pattern, making it impossible to tell them apart. Seven questions give you 2,187 patterns, which comfortably covers all 1,000 possibilities.

**What is the difference between binary search and the three-answer method?**

Binary search asks questions with only two possible answers, splitting the range in half each time. That approach requires 10 questions for numbers between 1 and 1,000. The three-answer method takes advantage of the "I don't know" option to split the range into thirds, cutting the needed questions down to 7.

**How do you formulate the question to split the range into three equal parts?**

You imagine a number in the middle third of the current range and ask whether your friend's number is smaller than it. If they say yes, the number lies in the lower third. If they say no, it lies in the upper third. If they say they don't know, it must be in the middle third where your imagined number sits.

**Can the same strategy be used for other ranges?**

Yes, the logic scales to any range of numbers. You simply keep dividing the current interval into three equal parts at each step. The number of questions needed is the smallest whole number where 3 raised to that power exceeds the size of your range. For 1 to 1,000, that number happens to be 7.

**What is the origin of this puzzle?**

According to the original account, this puzzle appeared in a column by Alex Bellos published on February 8, 2021, in The Guardian. Bellos is a well-known mathematics writer who regularly shares recreational math challenges with his readers. The specific attribution and publication date come from the source material, though the puzzle itself has roots in classic information theory. Editorial note: Any facts that remain unconfirmed, such as the exact publication date of the original column, are presented as approximate or hedged in the text above.

Canonical: https://curio.madeda.com/minimum-questions-puzzle-beat-binary-search-in-7/