D&D 5E Roll20's RNG - or Evil Electronic Dice Gods?

[Saeviomagy]

Every programming language has its own random number generator, and it is very, very hard to use these things incorrectly.

The top programming languages have easy mistakes that can be made by first time users, and that's before you even start trying to do things like simulate dice with them, or assign aggro based on them.

Yes, you could seed with the same number every time. That would be a mistake. And yes, the algorithms used have some limitations at the extreme edges of statistical consideration. But all in all, silicon-based pseudo-randomness is entirely adequate to roll a few fake d20's, and any pattern the OP is seeing is just, well, random.

.. assuming there is no bug in the program using the random number, which is a pretty big call to make, especially with a new piece of code.

 

[redrick]

The top programming languages have easy mistakes that can be made by first time users, and that's before you even start trying to do things like simulate dice with them, or assign aggro based on them.

.. assuming there is no bug in the program using the random number, which is a pretty big call to make, especially with a new piece of code.

The Roll20 dice roller isn't a new piece of code. They switched some of their back end functionality to use the same dice engine that gets used for dice calls by the players. This means that, if you write a script that performs a number of dice rolls in the background (say rolling saving throws for several creatures at once) those rolls will now use the existing library.

That doesn't mean that there isn't a bug. Bugs happen even in mature code. But what is being described here sounds, well, random.

These are not "first time users."

 

[Coredump]

eh... I have only been with my current DM for a few months. And I have already seen her roll 2 Crits... with disadvantage! Yeah, when you see double 20... y'know the gods have it in for you...

 

[Unwise]

My experience with programming and RNGs is basically that they are foolproof. They are so complex that they are not programmed at the providers end. Basically it is like writing a book and just photocopying that entire chapter off somebody else. It is just a snap-in that works. The things that can go wrong with it are on the integration side, meaning if something is wrong they just don't work at all. Nothing a provider does is likely to stuff up the generic RNG they have copied to make it spit out non-random numbers.

In my limited experience in programming, you can trust really simple stuff and stuff that is so complex nobody screws with. It is the stuff in between the two that people break.

So in short, I trust the RNG. I don't necessarily trust the program to not repeat the same result twice. I can see how somebody could accidentally stuff that up quiet easily. However, years of listening to complaining in Bloodbowl matches has made me inoculated to complaints of system bias. I'm going to trust the code over the observer.

 

[wedgeski]

.. assuming there is no bug in the program using the random number, which is a pretty big call to make, especially with a new piece of code.

Can't speak to that. I was just countering the popular notion that random number generators on computers "are not truly random" (not a quote from the OP, just a common sound-bite), and are therefore incapable of generating randomness, and thus that the bad luck you can have when rolling d20's, rolling for loot, etc., is not bad luck at all, but some fault inherent in the silicon.

 

[Malovaan]

Its human nature to see patterns where there aren't any.

Would people complain less if online 'random number generators' had some implementation to take away the random and produce something that people would perceive as looking more 'random'?

 

[wedgeski]

Its human nature to see patterns where there aren't any.

Would people complain less if online 'random number generators' had some implementation to take away the random and produce something that people would perceive as looking more 'random'?

Roll20's "QuantumRoll" system seems designed to do just that - reassure suspicious human brains that random doesn't mean you can't roll three natural 1's in a row.

 

[GMMichael]

What the hell roll20? Is this the trick of some cruel dice god?

Easy solution. You're using computer dice, which do the math for you. So - switch to 3d6 instead of 1d20. Heck, 3d6 is even faster to type than 1d20. Let's see the RNG get too many crits in THAT.

Its human nature to see patterns where there aren't any.

Amen. I'm off to go worship the Face-on-Mars god.

 

[delericho]

eh... I have only been with my current DM for a few months. And I have already seen her roll 2 Crits... with disadvantage! Yeah, when you see double 20... y'know the gods have it in for you...

Heh.

Once, long ago (3e days), I had a player choose to have his character run past an enemy creature, because he'd calculated that the monster needed a 20 to hit with its AoO, and anyway he had enough hit points to survive a hit. Two natural 20s later, and his character was down for the count.

That was with actual dice, though.

As for the randomness of electronic dice rollers: yeah, I don't trust them either. Which is probably irrational - there's actually very little guarantee that any 'real' dice are actually truly random either. Anyway, fortunately that's not an argument I have ever had to have with my players, none of whom are keen to give up our dice.

 

[Kalshane]

As others have said, the RNG code isn't new. They've just moved to using it for all dice rolls within the system. (Apparently the animated 3D dice previously used a different system.)

I've been playing and running games on Roll20 for two years now, and there's nights when it feels like the RNG is out to get you, but there's nights it feels like your dice (or the DM's dice) are out to get you.

The only difference with is the RNG you can't do the superstitious ritual of switching out your dice when they're rolling poorly.