Feature Request: Better Random Number Generators

[PartialCloning]

trafficstars

The current random number generators are not that random.

1. random.
2. random_num (Using the engine's RandomLong).

I have no data to back that claim but I'm not the only person to notice a pattern. A forum search shows many posts making the same claim.

In 2005, "Twilight Suzuka" released a better random number generator for AMXMODX, Module: RandomX using the Mersenne Twister PRNG.

RandomX uses a random number generator called the "Mersenne Twister", which is one of the fastest random number generators availible, and is light years ahead of the current ones implimented in AMXx. This means high quality and inexpensive randomized variables.

In 2009 "altex" reported a bug in the random number generator used in Sourcemod (Bug #3831). Mersenne Twister PRNG based random number generators GetURandomInt and GetURandomFloat were then added to SourceMod. Commit: Added better random number generator.

Possible Solutions:

• Adding RandomX to 1.9.0.
• Porting SourceMod's implementation into AMXx.

[ClaudiuHKS]

Tested Mersenne Twister (std::mt19937)

INT = -13, 0, 13, -7, 10, 6, -9, -14, 9, -11, -8, -5, 0, 7, 10, 10 REAL = -9.73, 10.94, -15.11, -14.70, -13.85, -3.49, -5.18, 2.96, -9.95, 4.09, 13.67, 3.87, -6.42, -6.23, -7.89, -6.28

So Mersenne Twister isn't much more better in my opinion.

[rsKliPPy]

Repeating numbers aren't a sign of a bad algorithm, that's not how random number generators are tested.

[ClaudiuHKS]

I'm not searching for some algorithm that isn't repeating numbers.

I know that M.T. method is better but the differences aren't huge though.

Besides, I would definetely agree implementing M.T. into AMXX.

[WPMGPRoSToTeMa]

random_num(0, 255): random_num(0, 1): random(256): random(2):

[justgo97]

You can try what HamletEagle suggested to try the RandomX module https://forums.alliedmods.net/showpost.php?p=2613505&postcount=11

[ClaudiuHKS]

AMXX can also add SM's M.T. generator but I strongly recommend that random, random_num and random_float functions will be the same as they actualy are.

[WPMGPRoSToTeMa]

@ClaudiuHKS as you can see from pictures random isn't so good.

[ClaudiuHKS]

Oh my.. I've always thought that..

#define random( x ) random_num ( 0, x )

You're right..

[IgnacioFDM]

It's probably best to just add a new function for a CSPRNG

[WPMGPRoSToTeMa]

@ClaudiuHKS

#define random( x ) random_num ( 0, x )

#define random( x ) random_num ( 0, ( x ) - 1 ) of course.

[PartialCloning]

@WPMGPRoSToTeMa can you post a picture showing how different RandomX's or SourceMod's PRNG looks? Do you think such a feature is needed or is random_num good enough?

[WPMGPRoSToTeMa]

@PartialCloning I think picture is only a minimal test, we need to make additional tests.

[toorisrael]

If you do something like this (just an example): public void OnMapStart() { for(int i = 1; i <= 3; fake++) { LogMessage("%i", GetRandomInt(0, 148)); } } and just keep restarting server, it will almost always log the same 3 numbers. (I've tested this only on one server). EDIT: sorry, now I see it's amx topic.

[Maxi605]

Will this ever be addressed?