I am trying to simulate a random dice roll as a basis for a chutes and ladders game
Code:int main { int i = 0, diceroll; while (i>5) { diceroll = 1+rand()%6; i++; } return 0; }
however the numbers get don't seem to be random, it always starts off with 6,6,5.. then its random.
That's invalid syntax, main is declared incorrectly. You wont get any numbers. Post real code next time.
EDIT: And make sure it's properly indented and actually readable.
You need to call srand(), to "seed" the PRNG. Call it once, near the beginning of your program (i.e. before you call rand). A common way to get different random numbers on every roll is to use the current time for the seed, though if you run your program in very quick succession, this doesn't work so well (since time() only reports back in whole seconds, and if you run your program twice within the same second, you get the same seed, and same PRNG sequence). Also, #include the right headers
Code:#include <stdlib.h> #include <time.h> int main() { srand(time(NULL));
Last edited by anduril462; 01-24-2014 at 10:55 AM.
There is then the problem that, depending on how srand()/rand() themselves are implemented, that rand()%6 may not give results uniformly distributed between 0 and 5. For example, they may be biased more toward some of the values.
If you want particular distribution or other properties of your random values, you need to find a generator which supports those properties.
I believe what the OP doesn't know, is that rand() is SUPPOSED to return the exact same numbers, with every run.
That is meant for statistical testing and debugging purposes.
When you are ready for more random values, just seed the random number generator, as explained above.
And be aware as Grumpy noted, that the values you receive, from a very small range of possible values, will be less than perfectly random - sometimes a LOT less.
Perfectly random number generation is much trickier than it seems, at first.
Just for fun - as mentioned the particular implementation of rand()/srand() may produce better, or worse, results.
For 20 samples (absurdly low, I know) I get
Code:Value Method 1 Method 2 0 8 (0.40000) 7 (0.35000) 1 12 (0.60000) 13 (0.65000)Code:#include <stdio.h> #include <stdlib.h> #include <string.h> #include <time.h> #define DICE_SIDES 2 #define NUMBER_ROLLS 20 int main(void) { unsigned dist1[DICE_SIDES], dist2[DICE_SIDES], i; /* srand(time(NULL)); */ memset(dist1, 0, sizeof dist1); memset(dist2, 0, sizeof dist2); for (i = 0; i < NUMBER_ROLLS; i++) { dist1[rand() % DICE_SIDES]++; dist2[(int)((double)rand() / RAND_MAX * DICE_SIDES)]++; } printf("%5s %25s %25s\n", "Value", "Method 1", "Method 2"); for (i = 0; i < DICE_SIDES; i++) printf("%5u %15u (%.5f) %15u (%.5f)\n", i, dist1[i], dist1[i] / (double)NUMBER_ROLLS, dist2[i], dist2[i] / (double)NUMBER_ROLLS); return 0; }
Hodor: a die with two sides might be better known as a coin, and will only roll if it lands on neither side.
It had six sides originally but I stood on it.Originally Posted by grumpy
In particular, if RAND_MAX is not evenly divisible by 6 (and on any computer you're likely to encounter, that will be true), there will be a bias towards some values. E.g., if you're using the modulus method and if RAND_MAX is 32767 (a depressingly common value), then 6 goes into it 5461 times with 1 remainder. Hence, 1 is 1/5462 (or about 0.000183) more likely than the other values, assuming an even distribution. The multiplication method still has this problem, but the biased value will probably be different.
And historically there was a problem with the low bits, which makes the modulus method especially problematic.
Basically, rand() is the poor man's RNG and needs to be replaced for anything serious.
And of course, there's no such thing as "perfect random number generation" with pseudo-random number generators. You'd need to access quantum effects for perfect randomness.
The cost of software maintenance increases with the square of the programmer's creativity. - Robert D. Bliss
Specifically, it will roll to the air conditioning vent.
The approved way to do it is
// uniform random number on 0 -1 - epsilon
#define uniform (rand()/(RAND_MAX+1.0))
// uniform random number on 0 - N-1.
#define rnd(N) ((int) ( (N) * uniform()))
// typical six sided dice
#define diceroll (rnd(6) + 1)
That way you use all the bits of the random number, so you'll get better results. If you're serious about random number generation, you'll have to use something more sophisticated than rand() however.
I'm the author of MiniBasic: How to write a script interpreter and Basic Algorithms
Visit my website for lots of associated C programming resources.
https://github.com/MalcolmMcLean
Approved by who?
Your "approved" solution makes specific assumptions about the properties of floating point types and operations, for which there is no guarantee.
