A Defense of Randomness in Competitive Games

[This post was partially inspired by Keith Burgun's article Randomness and Game Design.  While I ultimately (and respectfully) disagree with his conclusion, he makes several insightful points along the way, and his article is very much worth reading.]

Competitive play in all mediums engages players and spectators, but digital games are uniquely positioned to take advantage of randomness in a way with which physical games would be ill-equipped to replicate.

Before I go any further, I want to establish my "Base Principles" of competitive play.  These are axioms and rules that I believe competitive games should follow to remain engaging; ideally, disagreeing with my conclusion would necessitate disagreeing with my Base Principles.  I will ultimately posit that these base principles require a degree of randomness in competitive games to make them more compelling.

1. The winner of a game should only be obvious after the game has ended.

This first principle guarantees tension throughout play.  For example, if a game has a time limit of 10 minutes, but after 5 minutes one player has positioned themselves to be the clear winner, the remaining 5 minutes are essentially wasted time.  Once a player has crossed the "winning event horizon", the game has effectively ended.  This Base Principle usually requires come-back mechanics, victory deceleration, or other integrated balancing.

2. No one strategy should dominate or be naturally apparent.

Most competitive games develop 'metagames', a fine-tuned interaction between a group of players where certain tactics reign supreme; however, in order to keep play from stagnating, the developers of the game maintain certain affordances to vary the dynamics of the game, whether it be a patch in an FPS changing damage rates or Magic: The Gathering cycling out an old block for a new one.  A game where any majority of the gameplay dynamics are disregarded as objectively worse than others is a failure of game design and play area.

3. A game should always be "just."

In matches with a reasonable skill disparity, the better player should win.  The worst case situation for a competitive game is where someone plays a technically better game but, through no fault of their own, loses.

4. Testing player mastery requires uncertainty in gameplay

Because of the second Base Principle, a skilled player requires a diverse skillset and an adaptive style in order to be a top player.  Otherwise, when the metagame changes, their specific skillset may become invalidated.  For example, a player only skilled with long-range weapons becomes less competitive in a match mode favoring short-range weapons.  This player was not able to face the uncertainty of level selection adequately.

The last Base Principle champions randomness in competitive game design.  Of course, even 100% deterministic games have uncertainty due to the unknown element of player choice, i.e. in Chess I do not know how my opponent will move, so I have to plan for all of his likely moves.  However, designers can use randomness as a tool to enhance uncertainty in compelling ways.

Keith Burgun argued for two main types of randomness in his article (linked at the top):

Randomness can be separated into two categories: input randomness, and output randomness.

He also argues that input randomness where the player has little time to respond is essentially the same as output randomness.

While I think we would agree that input randomness is necessary in testing player skill for things like setting the initial game state, I disagree with him on output randomness removing complexity in games and only creating the appearance of depth.

I will use a competitive shooter as an example for illustrating my points.  Imagine there are two guns in a game:

1. Long Range: this weapon is deterministic if you are more than X units away from your target.  As you get closer than X units to your target, the weapon has a "miss chance" that increases.

2. Short Range: this weapon is deterministic if you are less than X units away from your target.  As you get farther than X units from your target, the weapon has a "miss chance" that increases.

Let's assume that players may choose their weapon upon spawn and respawn, that the weapons are indistinguishable at a glance, and that the level designers made sure that both weapon strategies are equally viable.  This would encourage dynamic, skill-based gameplay that isn't pure "risk management" but still includes high-risk high-reward gameplay. Just because a Short Range player is within X units of a Long Range player doesn't mean that the Short Range player will win.  One clearly has the advantage without being guaranteed victory.

We could build a technically similar game completely with 100% deterministic guns, but randomness allows us to create high-tension moments where we are unsure if we will succeed or fail - something that dovetails with our first Base Principle. We could enhance our current game that has randomness with more determinism.  Let's add a third gun:

3) All Range: this weapon is deterministic at all ranges, but deals less damage than the Long and Short Range alternatives.

Generally in life, the lower risk, the lower the reward; this system mirrors that in an intuitive way.

Output randomness can help balance a game between two players, maintaining the first Base Principle and match tension.  Imagine in this game we are creating that every time you die, you increase your chance of spawning with a temporary Super gun (when you spawn with it, your chance resets to 0).  In casual play, this allows players with a higher skill gap to still play together without one completely beaten.  In competitive play, this acts as a come-back mechanic to decrease the gap between a score.  Since you do not know when a player will spawn with the Super gun, the tension mounts as your kill count rises, then releases after they spawn with the Super.

This level of uncertainty would not be possible without output randomness; having a statically appearing Super (e.g. after every 5 deaths) allows each player to plan for the eventuality, rather than react and adapt to an unknown situation.  Planning and reacting are two very different skillsets, both required by a competitive player.

Finally, let's tie input and output randomness.  At the beginning of a match, the X units dictating the accuracy of Short and Long range guns is randomly assigned between a certain range of numbers, shifting the balance of power and usability of the weapons.  The mechanics of gameplay shift, and the players with better skill and adaptability should benefit.

More examples of output randomness causing greater gameplay depth:

For further arguments in favor of uncertainty, look online for one of Richard Garfield's talks on "Luck Versus Skill" in games.