The Two Types of Random in Game Design

From critical hits to random encounters, and from loot boxes to procedural generation, video games are stuffed to bursting with randomness. In this episode, I look at the way randomness is used in games - and why some forms are more contentious than others.

=== Sources and Resources ===

- Sources

Uncapped Look-Ahead and the Information Horizon | Keith Burgun

A Study in Transparency: How Board Games Matter | GDC Vault

GameTek Classic 183 - Input Output Randomness | Ludology

Why revealing all is the secret of Slay The Spire's success | Rock Paper Shotgun

Crate | Spelunky Wiki

Random Generator | Tetris Wiki

Level Feeling | Spelunky Wiki

Plan Disruption | Etan Hoeppner

Fire Emblem True Hit | Serenes Forest

The Psychology of Game Design (Everything You Know Is Wrong) | GDC Vault

How Designers Engineer Luck Into Video Games | Nautilus

Roll for your life: Making randomness transparent in Tharsis | Gamasutra

12: Into the Breach with Justin Ma | The Spelunky Showlike

- Additional resources

Many faces of Procedural Generation: Determinism | Gamsutra

Why Our Brains Do Not Intuitively Grasp Probabilities | Scientific American

How classic games make smart use of random number generation | Gamasutra

=== Chapters ===

00:00 - Intro
01:28 - Why we use randomness
03:42 - The information horizon
06:06 - The two types of randomness
08:59 - How input randomness can fail
13:32 - The advantages of output randomness
17:50 - Conclusion

=== Games Shown ===

Cuphead (2017)
Enter the Gungeon (2016)
Octopath Traveler (2018)
Mario + Rabbids Kingdom Battle (2017)
Griftlands (In Early Access)
Dicey Dungeons (2019)
Hearthstone (2014)
The Binding of Isaac: Rebirth (2014)
Darkest Dungeon (2016)
Dead Cells (2018)
SteamWorld Quest: Hand of Gilgamech (2019)
Into the Breach (2018)
Spelunky (2012)
Armello (2015)
Minecraft (2011)
Chasm (2018)
Downwell (2015)
Middle-earth: Shadow of Mordor (2014)
No Man's Sky (2016)
Celeste (2018)
Fortnite (2017)
Mario Kart 8 (2014)
Super Smash Bros. for Wii U (2014)
Tekken 7 (2015)
Super Mario Party (2018)
Bloodstained: Ritual of the Night (2019)
Borderlands 3 (2019)
Call of Duty: WWII (2017)
Valkyria Chronicles 4 (2018)
Civilization V (2010)
Wargroove (2019)
Plants vs. Zombies (2009)
XCOM: Enemy Within (2013)
Chess Ultra (2017)
Mark of the Ninja (2012)
StarCraft II (2010)
Slay the Spire (2019)
Apex Legends (2019)
Civilization IV (2005)
XCOM 2 (2016)
Overwatch (2016)
FTL: Faster Than Light (2012)
Card of Darkness (2019)
Diablo III (2012)
Tetris 99 (2019)
Puyo Puyo Tetris (2017)
Phoenix Point (2019)
Fire Emblem: Three Houses (2019)
Tharsis (2016)

=== Credits ===

Super Mario Party - Luigi wins by doing absolutely nothing | Nintendo Unity

Fire Emblem: Three Houses - New Game Plus Maddening Walkthrough Part 43! | MrSOAP999

Deadpool 2 © 20th Century Fox

Pandemic Card Art © Z-Man Games

=== Subtitles ===

[rodgwr]

: Are you a gamer?
: No, I am a _risk-calculating_ _tactician_

[Robert]

In players' defence, it's not just the disconnect between probability and our broken understanding of probability that causes anger (although that's part of it); it's also the disconnect between our intuitive understanding of *the situation* and the outcome. Yes turn-based systems are always abstract but I'm still seeing a special forces soldier point a rifle at a stationary target 3ft away and somehow miss. It completely shatters the fantasy of controlling this awesome elite squad.

[PrimerBlobs]

I passed over this video many times figuring it was about true vs pseudo randomness, but glad I finally clicked. A really cool video. Thanks!

[peterwang]

"Do we just like luck when it lands in our favour, and hate it when we lose?"


Yeah.

[indigocactus]

"The risk was calculated, but man am I bad at math."

[kamikeserpentail]

The big thing about output randomness is it feels much better when the player makes a choice to take that risk, rather than just being forced into it.

If you have to choose between option A that has a 50% chance of success or option B that has a 50% chance of success, there's only an illusion of choice.
But if you have to choose between option A which has 100% chance of 1 success, or option B which has a 45% chance of 2 successes, and 5% chance of 3 successes then it becomes more of a decision.
Not only is it better to have actual choices, but it puts more emphasis on when you choose each one.
You might only gamble when it is hopeless anyway and that 5% chance sometimes saves the day.
Or you might gamble when you're so far ahead that a loss doesn't matter as much and success allows you to feel even more powerful.
It gives the players better control over how much they are willing to be hurt/helped by RNG.

[fussel]

"99% hitchance, still miss!!!"

It's not even a mood anymore, it's a lifestyle at this point.

[jacobtaylor]

I'm reminded of when Team Fortress 2 first released weapon drops. Basically, every time you died, you had a chance of getting a random weapon. The percentage chance was set so that you'd get something every dozen deaths or so, and that's how it worked for the vast majority of people, yet some people were going hundreds of deaths without a single drop. Valve's mistake was that the game "rolled the die" on every death. Across millions of players, this meant that some unlucky folks ended up on the far far far end of the probability distribution. Put another way, the odds of rolling 10 six-sided dice and getting all 1's is about 1 in 150 million. However, if you have 20 million people each rolling 10 dice several times a day, someone will eventually draw that shortest of all short straws. Valve's solution was to change their game so that, after each drop, it picked a random number from a range that determined how many times you needed to die before you got your next drop. This created a hard limit on how long you could go without a drop.

Similarly, I don't doubt that some (maybe many) of the people who take to forums to complain about things like the hit chances in X-COM are suffering from a psychological misunderstanding of the odds, but I also suspect that, in a game that sold millions of copies, it is not impossible that a few people did, in fact, miss a dozen or more 95-99% chance shots in a row. It's this sort of thing that makes me think developers using RNG should always include some sort of backstop to prevent the game from being ruined for whoever ends up on the highly improbable end of their probability curve.

[hallowizer]

"For those who are unfamiliar with the term RNG, RNG stands for Really Not Good."
- Ceave

[olivierdubois]

I think you should have mentionned the games where the player is in charge of the luck. In Armello, you can burn cards to choose the result of the dice, so the player decides how much luck they're willing to rely on, and what price they'd pay for it. That way, when the result isn't favorable, the payer blames themselves for not paying more, instead of the randomness.

Danganronpa also has a similar system with the MonoMono Machine, where you can pay more tokens to decrease the chance of obtaining an item you already own.

[fakjbf]

The main problem is that people remember negative events more easily than positive ones. Yeah you'll sometimes miss in XCOM with a 99%, and you will distinctly remember just a couple rounds ago when you also missed with 99%. But you will forget the 100 shots you took in between that hit, they were exactly what you were expecting so you just move on and forget about them. This is also why no one complains about when they have strings of good luck, it's not just that the good luck makes them happy it's also that they are less likely to even realize that they are having a string of good luck at all.

[shmooters]

I feel like the issue with % is that we generally don’t understand the idea of that % being rolled every time something happens. For example, just because flipping a coin has a 50/50 chance of landing heads, getting a tails on the first flip does not increase the odds for heads on the next one, the probability is still 50/50 and will always be that. It’s just that mathematically we *expect* it to land on each side equally, even though in practice that rarely ever happens

[Waitwhat]

My favorite moment in XCOM was playing with my friend.
We were into the late game, only one death early on, main squad pretty beefed up.
None were showing pyscikik promise, so we swapped out one with a promising member. Nickolas Cage was his name. He was getting up in skill as well.

We entered the downed alien ship, fighting a brutal battle. We make it to a long hall way, three Cyber Disks show up. All of our shots miss. They fire, kiliing two.
We fire again, not much effect.
They fire, missing only on Nickolas Cage. The team is dead.

We say F it, and charge with Mr. Cage. He kills one.
All of their shots miss.
He gets a reaction shot.
He kills another.
He takes a blast to the face, still lives.
He kills the last. Mission completes, huge haul of resources making research possible that lets us crush enemies with (relative) ease till the end of the game.

[greatparmesan]

10:36 - "These cards are terrifying, game changing events that can completely demolish your team."
As someone who has played Pandemic several times, that is so true it's not even funny. It's always really dramatic when one of those cards rears its ugly head.
If it were more dramatic, the lights would dim and the card would glow green upon reveal, illuminating the horrified face of the unlucky person who drew it. The other players would gasp a little, and one player would ask, "Is it an Epidemic Card?" even though they already know the answer. The poor lad who drew the card in the first place would look up at the rest of their team, drop the accursed card on the table for all to see, and answer solemnly, "Yes."

[WhitefoxSpace]

It's been talked about ad nauseum at this point, and still I want to draw your attention to 14:42 - something about X-Com's animation system is such a slap in the face. 99% chance to hit, gun is deadlocked on the target.
<shoot>
> Soldier moronically pivots to the floor to miss a full burst of machine gun fire at 2 yards.

It'd be a legit way better experience, with the *exact* same number stats, if the animation somehow made sense. Like bullets flying underneath arms, or hitting cover really close by. At least that way the feedback would be "shit happens" and not "your soldier who you have spent a lot of skill points to upgrade and has seen many victories in many different parts of our alien-infested world is an absolute mongrel, but only every now and then."

[daviddamasceno]

Reminds me of a D&D session we had. I was helping my friend to make the best thief ever, helping him choosing the right perks, the right equipments and powers and magic weapons. We gave the character the name Chad. With all that combined, his thief became so OP he could hit anything on a roll of 5 or above. Turns out that, that whole night, he couldn't roll anything above 4!

It has been years but to this day we still remember and laugh at the tale of Bad Luck Chad.

[Ramzuiv]

"I've never said the word Epidemic so many times"... March 2020 says hi

[elim]

"[Randomness] can be a cruel mistress..." *Darkest Dungeon footage*
Yep that tracks.

[Mitrofang]

Quoting someone from one podcast I listen to: “I get angry whenever I watch Mark Brown. You can be good at editing, good with words or being right, but not those three at the same time; yet he does all of them every time”.

[monkeydkfetus]

“Scientists have calculated that the chances of something so patently absurd actually existing are millions to one.
But magicians have calculated that million-to-one chances crop up nine times out of ten.”


― Terry Pratchett, Mort