# The tough question Microsoft asked to see how smart candidates were

Microsoft is one of the world’s most successful companies and it has some of the best software engineers in the industry.

There is a battle among top tech companies to attract the best talent, which is why companies like Microsoft, Google, and Facebook have multi-million-dollar campaigns to achieve this goal.

Attracting talent is, however, only part of the process. A good interview is necessary to ensure the candidate is suitable for the position.

Speaking to Bloomberg’s Emily Chang, former Microsoft CEO Steve Ballmer said he had confidence in their interview process.

He said the aim of their interviews, especially for technical staff, was focused on three things:

1. You want to understand the person and their motivation.
2. You want to understand their academic records as many of them come from college campuses.
3. You want to understand people’s ability to problem-solve.

To test candidates’ ability to solve problems, Balmer said he gave applicants the problem below during the interview process:

I’m thinking of a number between 1 and 100. You guess the number, and after each guess I tell you if it is high or low.

If you get it right on the first guess, you get five dollars. If you get it correct on the second guess you get four dollars, and so on. After five guesses, the candidate would lose one dollar incrementally for each passing guess.

The question is – do you want to play or not? Hence, will you be likely to make money?

Ballmer highlighted that the problem requires multi-dimensional thinking and tests the candidates on programming and computer science knowledge.

### The solution

Ballmer said candidates with a strong problem-solving ability will not play the game because they are likely to lose money.

“There are far more numbers on which you would lose, than you would win. Secondly, I can pick numbers that are specifically hard for you to get,” Ballmer said.

Yaniv Sadka explained the solution in more depth, courtesy of The Art of Strategy by Avinash Dixit and Barry Nalebuff.

I’m thinking of a number between 1 and 100. What’s your first guess?

Was your first guess 50? Smart. My number is higher, but you just increased your odds from 1/100 to 1/50. What’s your second guess? 75? Sorry, that’s too high. What did you guess next? 63? Sorry, that’s too high again. What about for your fourth guess? 57? That’s getting warm, but my number is lower than 57.

Thus far, you’ve been using the most logical method by implementing the median method. The median method maximizes the amount of information you learn from each guess by minimizing the number of possible choices left.

Speaking technically, the median method “minimizes the entropy.” Expectedly, the answer that Steve Ballmer was looking for in his interviews was 50, 75, 63, then 57. It is the most efficient way to play the game and will get you to the number I picked as quickly as possible. As Nalebuff and Dixit explain, “this is the ideal strategy in a game where the number has been chosen at random.”

Unfortunately for you, I didn’t pick the number at random. This is a game where I lose money if you pick the right number. Naturally, I want to make it as unlikely as possible that you pick the right number. I anticipated you would answer 50, 75, 63, and 57, so I purposely didn’t pick those numbers. Your guesses using the median method failed to account for my perspective on the game.

So here we are. You only have one more chance to take my money. 51, 52, 53, 54, 55, and 56 are left to pick from. What is your final guess?

Was your last guess 51? Close, but no cigar. I picked 52. Yet again, I was one step ahead of you. The section above on how the median method only works for a number picked at random was set up to mislead you further.