Game Soup’s RNG Vid

What do you you think about this video about RNG?

Aaaaaaagh! I watched like half of this guy’s videos like a month ago, and this one was the one that was bad enough I wrote up a long comment on it, then my browser crashed and I lost the whole comment, because even my comment saving extension can’t save fucking youtube comments because their text boxes are so fucked.

The entire early diatribe about true randomness is unnecessary! We should know that computers aren’t really random, getting true random numbers doesn’t actually matter that much. Pseudorandom numbers based on modern algorithms are entirely suitable for most purposes! If a person cannot determine why a particular thing is the way it is, then it’s effectively random from a game design standpoint, even perfectly deterministic things like something being purely based on framecount, like item drops in darkwing duck on NES, or the items you get in Mario kart 64, which are based on the frame you press the button to stop the roulette.

He then props up pseudorandom number generation as a better fit for what we want out of games than true randomness, which is misleading. Pseudorandom number distributions could still be based on a true random seed and be exactly as useful. I myself think that pseudorandom number distributions that are weighted to certain outcomes are preferable, however most games don’t use this, and the distinction between true random numbers and pseudorandom numbers isn’t actually important to discussing probability distributions. Pseudorandomness versus true randomness is practically a theological topic from the perspective of game design, it’s only important to cryptographers!

Conflating true nondeterminism with skewed probability distributions is really fucking dumb. The real topic should be even probability distributions versus skewed ones. Here’s a rundown of true random numbers versus pseudorandom numbers.

And he mentions pac man, which is deterministic, not random. And Extra Credits. Please.

I’m going to ignore his initial conflation for the sake of sanity.

He asks if fair randomness is oxymoronic, bringing up poker hands, saying that if your opponent pulls out a 17 of spades, you know something’s up. His claim is that you know the full range of possibilities and they’re constrained within a certain set. However this has nothing to do with whether a distribution of random numbers is fair or not. You might know all the possibilities, but your opponent might simply get a better draw than you all the time or at all the critical times.

Competitive Tetris, Super Puzzle Fighter II, and Chess 960 are all games that involve random elements, however all of them are fair because their randomness is equalized between the players.

And he continues to reiterate that games don’t want true randomness, despite them effectively having even probabilistic distributions in a vast majority of cases, including ones he brings up himself.

The differences in RNG structures between per-call RNG, or RNG using the system clock or framecount as a seed largely don’t matter, because they don’t operate in a predictable enough manner to be controlled by a live human. The fact that TASers can manipulate RNG, or even that speedrunners can manipulate RNG is largely useless from a game design standpoint. We aren’t building games in assembly anymore, you can use Math.random(); all you like and it’s going to return a satisfactory non-deterministic result in most languages.

Most of the video doesn’t cover specific game design applications of RNG, just methods of building pseudorandom number generators and how rarely they can be abused. In my opinion, this isn’t useful to players, even if the game is competitive.

Late into the video talks a bit about actual probability distributions finally, like the distribution method used by Valve for Dota 2, which has small odds on events occurring twice in a row, but larger odds for each time they fail to come up, until it becomes certain that they occur. This skews the distribution to conform better with gambler’s fallacy, the illusion that when something hasn’t happened for a long time it becomes “due” or that it’s unlikely for something to occur twice in a row.

Also, it doesn’t matter in my opinion if clever players are capable of manipulating RNG, even in competitive games. If a player is clever and skilled enough to exploit that, then they deserve it. Game systems don’t need to be obscured from the player for the sake of keeping up the illusion of randomness. Games are supposed to be challenges. Overcoming the RNG is another challenge. It only makes sense to keep odds even in gambling games to guarantee that the house makes the majority of the profit. Why do you think card counting is against policy?

That item in golden sun that’s a reward for only the most dedicated players, if someone can figure out how the algorithm works and create a deterministic setup, they ARE one of the most dedicated players.

You don’t need to balance drop rates for rare items, you need to stop having rare items that unbalance the game. If you have a rare item that is extremely powerful, then you’re not balancing it by making it a rare drop. Think about the lucky player who gets that item, their experience is ruined because the difficulty curve has been upset. I’ve heard that story a few times about players who have gotten the black knight sword in the undead burg in dark souls. Super powerful item drops, now they’re all-powerful and the game is easy. In a competitive setting this is also true. Think from the opposite perspective, you have a rare item that outclasses all the others, but players discover a way to deterministically grab it, so now everyone has this rare item. What’s actually wrong in this scenario? Everyone is of the same power level. The idea of a rare item dropping making a single player more powerful because of luck in a competitive game is even worse than the deterministic abuse of PRNG patterns. The problem isn’t that the game can be broken, the problem is that the game breaks itself. An additional problem of the original scenario is, you’re asking players to seriously grind for hours to get a stupid virtual item.

A key thing he doesn’t discuss but lightly alludes to is the use of random numbers to generate a percentage chance that something occurs, versus using a random number to determine when the next event occurs. The difference between a step counter and a pure random chance per-step. I first learned about the differences in this when TF2 idling was a big thing, because it was originally a percentage chance for every unit of playtime, and it became something that occurs regularly after a randomly generated length of playtime.

The thing this all comes down to is, RNG is frustrating, players find it frustrating, players keep trying to find patterns in RNG that don’t exist, making them frustrated, or harming their ability to judge situations accurately.

About the only interesting point in the entire video is that the RNG in hearthstone is crucial to the game’s longevity, essentially the claim that fully deterministic games exclude people, and people are more drawn into games which aren’t totally deterministic, where the distribution of favorable to non-favorable outcomes is less even against the same players in similar circumstances, but not so much that it completely defies expectations.

The trouble is, this claim has no backing evidence, and ignores RPS type interactions (or interactions with hidden information), which are not strictly deterministic or truly random. Chess is still popular the world over and has been for a long time despite being a deterministic game of perfect information. Hearthstone, or an equivalent like Magic could function perfectly fine if you allowed players to specifically arrange the order of their deck in advance, or even draw whatever card they wanted to at any time.

Though if you let people draw whatever card they wanted, it would limit the usefulness of cards that let you draw specific other cards.

Even if you make these random elements deterministic, then you’re unlikely to upset the balance of the game, because players still can’t see each other’s hands. Any way you can build a deck would have another deck that counters it. You might have to insulate the game against early win cheese tactics a bit, because they’re likely to become centralizing (meta’s likely to centralize on decks that have cards up front that either win the game early, or prevent an early win, and if there’s too many ways to win the game early, versus ways of preventing it, then it’ll become an all-out guessing game), but honestly they’d be broken with lucky draws in a non-deterministic format too. I think combo decks are based on that idea anyway, loading the whole deck with early wins, so you’re guaranteed to draw one.

In this scenario, players can’t afford to settle into patterns, or risk getting countered by opponents. Similar goes for every fighting game ever. Dash dancing in smash bros is a great way of lowering your commitment and avoiding attacks, but if I see you doing it and nothing else, I can just hit you with something that sweeps across the whole DD range. You’ll never see a fighting game match repeat itself in general (though one time a gamecube froze, and we decided to replay a tournament match from scratch and ended up on pokemon stadium with the same lives, same percentages, on the same transformation of pokemon stadium. I was in a losing scenario and trying to get a chance to turn it around by resetting that one game which technically is within my rights to request in that situation. it didn’t work.).

Maybe it’s true that by removing RNG, bad players will never win, and therefore your audience shrinks because you have no bad players. That’s the thing that really hits home for me, that maybe games need to compromise themselves, allowing worse players to win over better ones, or correct decisions with correct execution to not pay off. There’s no empirical evidence behind this, but it’s a scary thing to think about that this guy portrays as a positive value of RNG. And we do already have the highly skilled playerbases of fighting games, arena FPS, and RTS that all have dying communities.

n his closing statement, you could reverse each of his statements to mean the opposite thing and make an equally valid case for it. There are already wrenches in the gears of how people ideally play games, as an infinite number of completely deterministic games could demonstrate.

Overall, I don’t like this guy’s videos, and did not find much value in them. I know not all the information in them is common, but it is entirely information I already have, except poorly communicated with conflations of terms, and a large number of mistakes in explaining technical details. I have this nagging voice in my head at all his statements, “What if you did the opposite of this thing? How could you make that work? If it does work, what’s the real underlying principle?” A lot of these videos involve repeating conventional wisdom without questioning it, and in my opinion, a lot of conventional gaming wisdom is incorrect.

His invocation of artificial difficulty in the difficulty video is especially frustrating. I had to make a comment about that.

Better than extra credits though.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s