This is a version of an article that first appeared in the March 2012 issue of Game Developer magazine.

The simple die roll is one of the single most vexing tools in the designer’s repertoire.  Rolling dice is fun.  Landing double-sixes is a lot of fun.  Seeing big critical hit numbers pop across big bad boss monster’s head is deeply satisfying.  And yet, the experienced designer knows that, much of the time, the dice are his enemy. Which is tragic, of course.  Most game designers, at the core, love absurd amounts of dice rolling.

There are few ways to get a room full of RPGers excited than to bring back the fond memories of the Rolemaster Critical Hit Tables, where simple acts like trying to climb a ladder might result in accidentally dropping your sword and dismembering your own arm.  The sheer magic and hilarity of events like this is transcendental – it’s a story that will be described to other gamers for years after the fact.  What is often is only fun in retrospect is passing that point and actually playing further as a one-armed Paladin.

Luck Vs Skill.

Randomness is considered a cornerstone of gaming – in fact, few items better symbolize gaming than dice, the history of which goes back at least 5000 years to an ancient Backgammon set found in the Burnt City excavation site in Iran.  But even the ancients could get past randomness.  Chess, a game that emerged from the sixth century as perhaps the finest game design of all time, has no randomness whatsoever beyond deciding which player should go first.

In the absence of randomness, games are entirely about skill, which means that the debate between skill and luck goes back at least 1500 years.  That game design distinction also happens to be one of the few that actually have a legal definition – whether or not your game is considered to be a game of skill or a game of chance determines whether or not players can bet money on it on a Vegas gaming floor.

A game without a significant luck element involved is almost entirely skill-based – and most of the video games played professionally, like Starcraft and Team Fortress II, have little or no luck component.  Both hardcore electronic and strategy board gamers tend to eschew too much luck in their games – two board game classics, Diplomacy and Puerto Rico, have extremely limited luck components.

But what may not be obvious beyond this bias is that the converse is not necessarily true: games with heavy chance elements may also have strong biases towards skill – provided that skill can be used to make sense out of the chaos being thrust upon the board game.  Magic: the Gathering has a huge luck component to it, primarily based on the order in which you draw cards – and this random component allows new players to have a fighting chance against more experienced players.  That being said, pro Magic players go to great lengths to assemble decks designed to reduce this element of chance as much as is humanly possible.

Random in Solo Play

Of course, in some games, randomness is unavoidable.  In particular, many one player games are particularly hard to design without some random element to it – the AI, or the challenges faced within. One of the most common examples is randomizing the initial board state – this is done in games as casual as Solitaire and Minesweeper, or as hardcore as Civilization.  The randomness is a huge part of why these games can be replayed hundreds or thousands of times (and conversely, many modern games that abandon randomness for tightly scripted experiences have little or no replayability at all).

But the randomness also can present the player with wildly variable problem sets.  Civilization on the hardest difficulty is (at least to me) pretty much impossible without a favorable initial world set up.  Conversely, a straight game of Klondike solitaire has a pretty good chance of being completely unwinnable from the moment cards are dealt.  Still, there are opportunities for the enterprising designer – there are versions of both Solitaire and Mah Jongg available online that promise to present a winnable problem to the player.

The psychology of a known solvable game of solitaire is, as a designer, interesting to contemplate.  Once you know the game is solvable, is the player more likely to replay a vexing board?  Does the game feel more like a true skill-based game at this point?  Or does removing most of the unwinnable boards make the game, on average, trivial to beat?

The Downside of Random

Imagine if you swung your sword and did 10 points of damage with every swing, every second.  This would very quickly start to get stale and monotonous, so game designers find ways to add variance to the system – making the damage range from 5 to 15 points instead, or adding in random hits and random misses, occasional chances for critical hits.

One streak of bad luck can make this all fall apart.  If random rolls are truly random, its entirely possible that our hypothetical swordsman might swing for 5 or 6 every time – hypothetically halving his DPS in this example.  Even worse, if his opponent is having the opposite luck, she might be tripling his DPS, even if they both have the same sword and skill.  Furthermore, on the next fight might have exactly the opposite results.  The end result would feel – well, random.  So much so that it would feel like player skill had no role in the fight whatsoever.

Streaky behavior constantly has to be controlled for in games.  In MMOs, its not at all unpopular to give a raid monster an enhanced critical attack that does so much damage (or has some other dehabilitating effect) that the tank has to do extra trickery, or the healing brigade has to drop non-stop healing in a never-ending chain.  These powers that are extremely devastating if they happen once become unsurvivable if the boss happens to land 2 or 3 of these in a roll.  Designers then have to go in and put in cooldowns, or otherwise manage the randomness to keep the fight survivable.

Random is Random

One casual takeaway from this is that randomness is something that is more fun when it breaks in the player’s favor.  Critical hits are fun – and streaks of them are giddy fun – but rolling misses is not.  This is a useful insight, but still an oversimplification – especially as these reward events get rarer.

As an example, it’s decidedly hard to make loot both very rare, and reliably drop.  Take the odds of dropping a phat purple from 1 to 1000 to 1 in 10000, and you’ve increased the odds that some player will get unlucky and never see the item.  On the flip side, a one in a million chance of hitting the jackpot ceases to be rare when you have more than a million players pulling the lever more than a billion times.  The jackpot is going to drop, and some lucky players are going to get multiple bonuses.

On top of it all, in a true random system, each player’s rolls are truly independent.  If I’ve flipped a coin 19 times and gotten heads every time, the odds that the 20th will also be a head is still exactly fifty-fifty.  Similarly, without artificially mucking with things under the covers, there is no guarantee that multiple players playing the game won’t get lucky and strike it rich, either flooding the economy or bankrupting the house as a result.  There’s also a chance that the event will fire so rarely that its mere existence is but a myth to the playerbase – which may or may not be a bad thing.

When Random Isn’t

Is true randomness a good idea at all?  Designers may find that true randomness is not well understood or mis-interpreted.  True randomness can, for example, give you the same loot 13 times on 15 boss attempts – it’s unlikely, but entirely possible.  And while designers love to roll dice, it’s also fairly important that we can actually provide something resembling a predictable player experience.

This is actually trickier than it seems.  The human brain is designed to seek patterns, and instill order, logic, or superstition on what it sees.  Need for greed rolls in MMOs perfectly show the failures in psychology here: the player won’t even notice a full night of loot drops being distributed evenly, but will glom onto the fact that his party mate managed to win 3 things in a row.  The player will utterly discount when he’s rolling well if his party mates are rolling better than him.  And the average player is only subtly aware that larger party sizes decreases the odds that he’ll win any given roll.

Still, there are ways to give the illusion of randomness, while still finding ways to keep the math under control.  In Star Wars: The Old Republic,which party member gets to speak each line is chosen randomly – but in beta, players were getting frustrated because they were losing conversation rolls too often.  We responded by putting in a bonus for the losers – every time you lost the roll, you got a cumulative bonus on future rolls, which evaporated when you finally won.  This not only helped solve the problem, but rewarded the losers with some fun, ludicrously large rolls to brag about as well.

Another way to approach the problem is by removing successful rolls from the probability space.  Imagine that your 1d100 die roll is represented by a randomly shuffled deck of cards with those numbers on them.  When you choose a number, remove that card from the deck.  Once the last number, all cards are shuffled back together and are selectable again.  This approach has several benefits to it, the least of which being that a 50% chance of success is, over time, pretty much guaranteed to be 50%, and win streaks will almost certainly be counterbalanced over time.

There are pitfalls, though, as mechanics such as the above can be leveraged and potentially exploited by a savvy player.  An experienced card counter knows to adjust his game once he knows the blackjack deck is disproportionately full of face cards – this is possible because, at that point, the deck is no longer truly random.  That being said, this may not be something that has to be stopped, and may actually be a gameplay opportunity.  Strategy in deckbuilding games like Magic the Gatheringand Dominion heavily leverage awareness of what remains in your deck, and the designers of both games have learned to embrace this and give players tools to further finesse these possibility spaces.

Random conclusion

There are many reasons why Texas Hold’em exploded onto the scene, became the dominant form of poker and along the way found its way to televised tournaments and worldwide popularity.  One of those reasons is that the game is much less random than the strains of poker that came before it.

In five card draw, each player’s hand is relatively isolated from each other.  To have a great hand at all, you need to be fairly lucky.  The impact of one player’s hand on another is relatively minimal, and two players having great hands the same turn is relatively unlikely.   Players are operating blind – beyond bluffing, the game is all luck.