ForumSections

help needed for random numbers within a range

New comment

How do you generate a random number within a range of 1 to 10?

[Deleted]
https://docs.mql4.com/math/MathRand and why not use the search function believe me it's free!
farhang wrote >>
https://docs.mql4.com/math/MathRandand why not use the search function believe me it's free!

The MathRand function returns larger integers like 32767 and i have to then start using mod to reduce and reduce. not what i want.
thks.

Use MathFloor(MathRand(TimeLocal())/3276)

Roger:

Use MathFloor(MathRand(TimeLocal())/3276)

MathRand() has no inputs; I guess u meant calling MathSrand(TimeLocal()) before calling MathRand(). Anyway, an alternative way (and arguably a more 'standard' way of using MathRand()):

MathSrand(TimeLocal());
int num = MathRand()%10 + 1;     // random number between 1 to 10 (even distribution)

You can find an excellent explanation in the section about MathRand() and MathSrand() here -> Articles -> Features -> MQL4 Language for Newbies. Technical Indicators and Built-In Functions.
No! Never take the least significant bits of a random number, they are not very random. Always use the most significant bits.
int num = 1 + 10*MathRand()/32768; // 1-10
Also note you must divide by one more than MathRand's max value (32767) to get the proper range.
WHRoeder:
No! Never take the least significant bits of a random number, they are not very random. Always use the most significant bits.

Interesting! Please explain.

[Deleted]
Also look at the MathSrand() seeding function and be aware that the sequence of random numbers that you generate will repeat itself unless you change the seed.
And generating a series of unique seeds may cause you a problem when performing repeated backtests on the same data in the tester.

CB
I was referring to "Never take the least significant bits of a random number, they are not very random". I studied EE and not computer science, so I might have some theoretical gaps in my knowledge in this regard. Googling this hasn't brought any relevant results. I thought those sort of problems were relevant years ago, but not with modern pseudo-random number systems.
New comment