Whats up with this?

[BamesJond]

So I wanted to test if the RNG in java is legit and came up with this:

 

 

 

 

 

 

Which basically shows the percentage of how many of each of 3 numbers were generated.

But if I add one more 0 to the counter, this happens:

 

 

 

 

 

Can anyone explain WHY this happens.

 

[Guest]

Are you sure that int in java can be that big?

[BamesJond]

Are you sure that int in java can be that big?

Well if I add one more 0 to the counter, it gives me an out of range error.

[mopman94]

Well if I add one more 0 to the counter, it gives me an out of range error.

Try and long instead of an int and see what that does.

[mopman94]

actually an int should hold 10 digits so it might not be that.

[mopman94]

I think the numbers are to big for a double. Try a long double assuming there is such thing in Java (there might not be)

[median555]

I think that AleksaNS is right. I believe that when you initialize your 'a' value Java treats it as:

double a = (double)(yks + kaks + kolm);

Meaning that the summation is done as integers.

Looping 100,000,000 times is fine, but taking it up to 1,000,000,000 takes the sum over the highest limit of what Java 32 bit signed integers can store (which is 2,147,483,647) and you end with an overflow, which causes the a value to be negative.

I haven't copied your code to test this, but try casting your 'yks', 'kaks' and 'kolm' values to doubles or longs as suggested by mopman, before doing your summation.

 

quick tip edit: To get the highest value of an int, double or long, just use Integer.MAX_VALUE, Double.MAX_VALUE or Long.MAX_VALUE 

[mikebald]

It also looks like, from reading the API, you'll end up getting a sequence of random numbers from doing one new Random(); and the sequence will loop. It's interesting that the Java implementation operates with 48-bit values but only returns 32 most-significant bits. Here's more info if you want to read about the algorithm used to create the pseudo-random numbers.