srand(time(NULL)) in C: Why Random Numbers Repeat and How to Seed Them

srand(time(NULL)) feeds the clock into the generator, and rand() then walks a fixed sequence from that starting point.

The line srand(time(NULL)); appears at the top of almost every C program that needs random numbers, such as the die-rolling example. It is often copied without explanation. Let’s see what each part does and what happens without it.

The Surprise: rand() Is Not Random

rand() returns pseudo-random integers. “Pseudo” means they come from a formula: each number is computed from the previous one, starting at a seed. Same seed, same sequence, every time.

Let’s roll five numbers from 1 to 100 without setting a seed:

#include <stdio.h>
#include <stdlib.h>

int main(void) {
    for (int i = 0; i < 5; i++) {
        printf("%d ", rand() % 100 + 1);
    }
    printf("\n");
    return 0;
}

Running it twice:

84 87 78 16 94
84 87 78 16 94

Identical. When we never call srand(), C behaves as if we called srand(1), so the program starts from the same point on every run.

Note

The exact numbers depend on the C library. The ones in this post come from glibc (gcc on Linux). Another platform gives a different sequence, but it will still repeat from run to run.

The Parts, One at a Time

1. rand() returns the next number in the sequence. It has no memory of the clock; it only continues from its current state.
2. srand(seed) sets that starting state. It takes an unsigned int.
3. time(NULL) (from <time.h>) returns the current time as the number of seconds since 1 January 1970, the Unix epoch. This value changes every second, so each run starts from a different seed.

Together, srand(time(NULL)) means “start the sequence from a point that depends on the clock.”

A Fixed Seed Is Useful Too

Passing a constant makes the sequence repeatable on purpose:

srand(42);

Every run now prints 67 41 82 42 13. That is exactly what we want when testing or debugging, or when a simulation must be reproducible. Print the seed, and anyone can replay the same run.

Seeding with the Clock

Here is the usual form:

#include <stdio.h>
#include <stdlib.h>
#include <time.h>

int main(void) {
    srand((unsigned int)time(NULL));

    for (int i = 0; i < 5; i++) {
        printf("%d ", rand() % 100 + 1);
    }
    printf("\n");
    return 0;
}

Two runs a second or more apart:

20 24 59 67 77
15 82 18 77 34

Why the (unsigned int) cast?

time(NULL) returns a time_t, usually a 64-bit integer, while srand() takes a 32-bit unsigned int. Passing one to the other narrows the value. With default warnings gcc stays quiet, but with -Wconversion it says:

warning: conversion from 'time_t' {aka 'long int'} to 'unsigned int' may change value [-Wconversion]

The cast says “yes, we know” and silences it. Losing the high bits is harmless here, since all we need is a number that differs from run to run.

Pitfall 1: The Same-Second Trap

The clock only changes once per second. Two runs that start within the same second get the same seed, and therefore the same numbers. Running the program twice in a row from a script printed this:

20 24 59 67 77
20 24 59 67 77

For a game or a quick demo this rarely matters. For a program launched many times in a loop (a test harness, a batch job), it does.

Pitfall 2: Seeding Inside a Loop

srand() belongs once, at the start of main(). Putting it inside the loop re-seeds with the same second again and again:

for (int i = 0; i < 5; i++) {
    srand(time(NULL));              // wrong: restarts the sequence each pass
    printf("%d ", rand() % 100 + 1);
}

Output:

15 15 15 15 15

Every pass restarts the sequence from the same seed and takes its first number, so we get the same value five times.

What srand(time(NULL)) Cannot Do

• It is not secure. Anyone who knows roughly when the program ran can guess the seed. Never use rand() for passwords, tokens or cryptography; use the operating system’s secure source instead (for example getrandom() on Linux or arc4random() on BSD and macOS).
• It is not very uniform. rand() % 100 slightly favours small values, because RAND_MAX + 1 is rarely a multiple of 100. For teaching and games this is fine.
• In C++, prefer <random> (std::mt19937 with std::random_device) over rand()/srand().

Quick Summary

Question	Answer
Why does rand() repeat across runs?	No srand() means the default seed 1 every time.
What does time(NULL) give?	Seconds since 1970, so a new seed each second.
Why cast to unsigned int?	time_t is wider; the cast avoids a -Wconversion warning.
How often do we call srand()?	Once, at the start of the program.
When do we use a fixed seed?	Testing, debugging, reproducible simulations.