Maze Generation

Mazes are designed to be complex; otherwise, they would not be used to make people lost. However, the techniques for creating mazes can actually be quite straightforward. You can create a complex maze by following a simple algorithm. This makes mazes ideal for diving into procedural content generation, a field of study devoted to building complex structures and art from code with little or no input from humans during the process. This tutorial will not provide any actual code, but rather is designed to be an introduction to maze generation algorithms. Let’s first look at an example algorithm:

  1. Begin with an empty space with no walls.
  2. Divide the space in half with a vertical wall.
  3. Randomly place a single opening in the wall.
  4. Divide each half in half with a horizontal wall.
  5. Place an opening in each wall.
  6. Repeat 2-5 for each new section until the desired level of detail is achieved.

This is an example of a recursive division algorithm. Following this algorithm, we end up with this:

simple subdivision algorithm

There are obvious weaknesses. One is that there are more horizontal walls than vertical walls. This is inevitably the case for any square maze created with this algorithm. The other problem is that it is somewhat predictable. Given a view of the whole maze, you can see an obvious middle point you must reach to get to the other side of the maze. You could find similar points that connect each quadrant of the maze to another quadrant. You can easily find the red, then green, then blue dots that you must pass through to access other parts of the maze. Connecting those and finding any other point in the maze from those is trivial. We could modify the algorithm, but related algorithms will have similar issues. Instead, we need a different model for our mazes, one that will give us flexibility to create all sorts of algorithms.

Graph Theory and Mazes

We need to better define our goal. Let’s say that our ideal maze is a maze where every point in the maze is accessible to every other point in the maze in exactly one way. This means no loops and no inaccessible areas. We can see the recursive division algorithm achieved this, which makes sense. Since we only left one opening between halves at each level of detail, there could only be one path between them. Looking at mazes in this way though, of points connecting to other points, lets us use graph theory. For those not familiar with graph theory, it is a mathematical field of studying graphs, which are structures that have objects, or nodes, connected to each other through edges. If our nodes, for example, are cities, then we might consider the presence of a road between them to be an edge. In our maze example, we can set up various points that we assume will be empty, like a room, and connect them together in various ways.

Our definition of an ideal maze lets us do something interesting with our graph. If there is only one path to any node from any other node, we could use any node as the root node of a tree. Trees are a special type of graph, and a lot of algorithms have been invented for building and traversing tree structures.

Because mazes are random though, we do not know what the tree will look like until we generate it. This is okay. If we imagine our maze as a tree, we can traverse it to “discover” what that tree looks like, even if that is not determined until the traversal. There are two popular types of tree traversal/generation algorithms worth describing, and they each produce a very different kind of maze. After their description, I included another algorithm that is a sort of combination of the two.

Depth-First Search

Depth-first search (DFS) is a simple tree traversal algorithm that works the way it sounds. First, keep going deeper in the tree until you can’t go any further. Then back up and try another branch. More specifically, the process is as follows:

  1. Start at the root node, which can be arbitrarily chosen. Mark it as visited, and set it as the active node.
  2. Do the following:
    • If there are no unvisited children of the active node:
      • If the active node is the root, you are done.
      • Else, the parent of the active node becomes the new active node.
    • Else:
      • Pick a child, mark it as visited, and set it as active.
  3. Repeat Step 2.

Since we are using this for maze generation, this requires a few tweaks. To implement this, I used a 2D array, where a 0 represents empty space and a 1 represents a wall. For example:

1  1  1  1  1  1  1
1  0  0  0  1  0  1
1  1  1  0  1  0  1
1  0  1  0  0  0  1
1  0  1  1  1  0  1
1  0  0  0  0  0  1
1  1  1  1  1  1  1

The bold zeroes are our nodes. A zero between two nodes means the nodes are connected. A 1 means they are not connected, and there is instead a wall. We can begin with a map entirely filled with ones. Whenever we change a node to a zero, it means we have visited it. If we go back to thinking about this maze as a tree, then each node can have up to four children, one in each cardinal direction. This is not always the case, though, since sometimes, moving in a direction means leaving the edge of the maze or connecting to an already-visited node, creating a loop. If there is no direction to possibly connect to, then we back up to the previous node. Here is how we carved out the above maze. Red is the active node. Blue shows the possible next move. Orange is the path created so far.

Normally, DFS would have a consistent yet arbitrary way of selecting the next node in the path. For a maze, though, it is better to choose randomly. Whenever there are multiple blue paths, one is chosen completely at random. Note that at steps 5 and 10, there were no options for possible moves. In these situations, we back up until there are possible moves. We can keep track of the path using a stack. With each move, add something about the move, such as coordinates, on the top of the stack. When backtracking, take the top coordinates off the stack and set that as the active node. When we arrive back at the root node (which we know we did if the stack of moves is empty), and there are no possible moves, we know we are done. Following this, I created a maze that looks like this:

This does not have the same problem as recursive subdivision. There are no obvious ways to divide this maze in half. However, if you try following a path, you will notice that you are not actually faced with many choices when navigating this maze. There are a few dead ends, but most of the maze is just a couple long, winding paths. This might be preferred in many situations, but not always, so it is worth exploring other algorithms.

Simplified Prim’s Algorithm

Prim’s algorithm is in many ways opposite of DFS. It is breadth-first, so we will be looking very shallowly at a lot of paths before moving deeper. It is usually used for finding minimum spanning trees, or a tree with the fewest/shortest edges. The result is an optimal path that connects everything together. The algorithm looks something like this:

  1. Start at the root node, which can be arbitrarily chosen. Add it to the tree.
  2. Of all the edges that connect the tree to an unvisited node, choose the one with the smallest edge weight.
  3. Repeat until all nodes are visited.

For our purposes in making a maze, we can simplify this by removing all concept of edge weight. We don’t want to find an optimal solution; that would be a very boring maze. Instead, we will choose the edge randomly. Imagine maintaining a list of possible next moves. Every time you visit a new node, you can add another move to its adjacent unvisited nodes to the list. However, unlike DFS, the next move is selected from the entire list, not just the ones from the last visited node.

 

While this maze is also random, it is significantly different in appearance to a DFS maze. The modified Prim’s algorithm creates a maze with lots of dead ends, but very short paths. While this fixes the problem mentioned about DFS, it takes it to another extreme. There are so many dead ends that none of them go very far. You aren’t likely to be fooled by a dozen nearby dead ends that don’t even turn a corner.

Combination of the two

Both maze algorithms have weaknesses, so I figured a combination of the two ideas would create a better balance. For this algorithm, like our simplified Prim’s algorithm, we maintain a list of edges, but we tend to stick to the same path a little longer like in DFS. Do this by defining some maximum number of moves before a jump, such as 5. Now, when we pick an edge, we then do DFS for only 5 moves before quitting and picking another edge like in Prim’s. For simplicity, it may not be necessary to maintain a stack of move history like in DFS. Instead, you could simply say that if you get stuck, switch back to the Prim’s algorithm instead of DFS, and start working elsewhere.

  1. Start at an arbitrary root node. Add it to the tree.
  2. Add this nodes edges to a list of possible edges to choose from.
  3. Choose one of the edges to connect to. Add its edges to the list of edges. Also replace lastEdges with its edges.
  4. Do the following MaxMovesBeforeJump times or until lastEdges is empty:
    • Choose one edge from lastEdges and connect to it.
    • Replace lastEdges with just the edges available from the last move.
  5. Repeat 3 and 4 until there are no edges left.

After experimenting with different values for the maximum number of moves before a jump, I was able to get different mazes on the range of qualities we see from the DFS and Prim’s algorithm extremes.

2 step combo algorithm
2 steps max before jump
5 step combo algorithm
5 steps max before jump
10 step combo algorithm
10 steps max before jump

Conclusion

These algorithms form the basis for most maze generation, and they can be adapted in many ways. While these examples acted on a 2D array to produce a square maze as is most common, the graph theory behind the algorithms does not require these kinds of limitations. It can easily be adapted to 3 or more dimensions (I actually created a 4D maze one time). You could even build a maze from other types of graphs, such as a maze with hexagon nodes. I would have posted my implementation code, but it is too long for this post, and it is for sale on the Unity Asset Store in a package called Maze Creator. The package also allows the creation of maze templates, a tool for greater design flexibility outside the scope of this article. If anyone is doing anything cool with mazes or have any questions/suggestions, I’d love to hear about it!

3-Handed Laser Shuffle

I recently competed in a game jam on Itch.io. The requirements were a 128×128 resolution, four colors, one weekend, and a theme of “break out.” So I made this. The controls take some getting used to. It’s intentionally a bit awkward, and requires some practice switching among three controls with only two hands. WASD controls the red laser. TFGH controls the green laser. The arrow keys control blue. Try to destroy the matching blocks before they get too close. Every 100 blocks cleared gives you a power-up that you can use with SPACE. This pauses block movement for a few seconds.

3-Handed Laser Shuffle won first place in the Mini Jam, among 20 entries. Try it out!

Creating a Circuit in Unity

During a recent game jam, my team made a game called Elevator Circuit. The game is now available on for free on Itch: https://tykenn.itch.io/elevator-circuit. The objective is to complete a circuit by moving elevators holding circuit segments, aligning them into a complete path. It currently has five short puzzles with more to come. My biggest role in the project was to program the circuit system.

The idea was that any segment without a path to a generator would be red. A segment with a path from one side, but not a path from the other is yellow. When there is a path to a battery from both sides, current can actually pass through the segment, it turns green. I used Procedural Lightning to indicate that the connectors of two segments are close enough to be connected (It happens even when the two segments are not connected to a battery, but we needed some kind of feedback).

My first (failed) attempt

I made a script for each connector that checked for other connectors to enter and exit its trigger area. It then registered the segment of the other connector to current segment. The connector then told the segment to update its material by recursively checking its neighboring segments, and then its neighbor’s neighbors, until it either dead-ends, loops , or reaches a battery. All the segments along that path would update their materials accordingly. I soon ran into all sorts of race conditions and edge cases. The lightning would correctly connect segments, but the colors of the segments would be wrong half the time. So, I scrapped that and tried something new.

Working from the battery to the segment.

Since there would only ever be a handful of segments in any stage, it would not be too expensive to check for changes every frame. I kept what I had before with each connector registering other segments to its own segments, but instead of updating the materials only during a connection change, the battery would “pulse” every frame. Going recursively to each connected segment from the battery, it would mark the visited segments as “connected” either from the left or right, depending on where the pulse came from. The segments would then all update their materials accordingly, the frame would render, and then the connections would reset for the next pulse.

Here is the script that goes on the individual connectors:

using UnityEngine;

public class CircuitConnector : MonoBehaviour {

    public bool isLeft;
    CircuitSegment segment;

    private void Awake()
    {
        //Expect CircuitSegment script to be on parent.
        segment = transform.parent.GetComponent();
    }

    //Make a connection
    private void OnTriggerEnter(Collider other)
    {
        CircuitConnector otherCon = other.GetComponent();
        if (otherCon != null && otherCon != this)
        {
            //Register connection to segment
            if (isLeft)
            {
                segment.leftSegment = otherCon.segment;
            }
            else
            {
                segment.rightSegment = otherCon.segment;
            }
            //Code for anything else to do during a
            //connection, like instantiate lightning
        }
    }

    //Lose connection
    private void OnTriggerExit(Collider other)
    {
     
        CircuitConnector otherCon = other.GetComponent();
        if (otherCon != null && otherCon != this)
        {
            //Unregister segment
            if (isLeft && segment.leftSegment == otherCon.segment)
            {
                segment.leftSegment = null;
            }
            if (!isLeft && segment.rightSegment == otherCon.segment)
            {
                segment.rightSegment = null;
            }
        }
    }
}

And here is the script that goes on the circuit segment:

using System.Collections.Generic;
using UnityEngine;
using UnityEngine.Events;

public class CircuitSegment : MonoBehaviour {

    [HideInInspector]
    public CircuitSegment leftSegment;
    [HideInInspector]
    public CircuitSegment rightSegment;
    [HideInInspector]
    private List meshesToPaint;

    //Usually only one of these in the scene at a time
    public bool hasBattery;

    //I used a red, yellow, and green material
    public Material deadMaterial;
    public Material oneWayMaterial;
    public Material poweredMaterial;

    //Optional event for receiving/losing power,
    //like opening and closing a door
    public UnityEvent powerEvent;
    public UnityEvent losePowerEvent;

    //Resets every frame, set by pulses
    private bool leftPowered = false;
    private bool rightPowered = false;

    //Does not reset, decided after pulsing finishes
    private bool powered = false;

    //only true during a pulse for checking
    //for loops
    private bool pulsing;

    private void Awake()
    {
        //Register all the meshes I want to paint.
        //I tagged all of them with "Wire"
        meshesToPaint = new List();
        foreach (Transform child in transform)
        {
            if (child.tag == "Wire")
            {
                meshesToPaint.Add(child.GetComponent());
            }
        }
        UpdateMat();
    }

    public void Update()
    {
        //Handle events for gaining and losing power.
        if (!powered && leftPowered && rightPowered)
        {
            powered = true;
            powerEvent.Invoke();
        }
        else if (powered && !(leftPowered && rightPowered))
        {
            powered = false;
            losePowerEvent.Invoke();
        }

        //Reset power
        leftPowered = false;
        rightPowered = false;
    }

    //Don't pulse until everything is reset from Update()
    private void LateUpdate()
    {
        //Only segments with batteries start a pulse
        if (hasBattery)
        {
            if (leftSegment != null)
                leftSegment.Pulse(this);
            if (rightSegment != null)
                rightSegment.Pulse(this);
        }
    }

    //Recursive function for deciding connections to batteries
    public void Pulse(CircuitSegment from)
    {
        //If a segment has already been visited in a pulse, it
        //must have looped.
        if (!pulsing)
        {
            pulsing = true;
            if (from == leftSegment && from == rightSegment)
            {
                //Handle edge case of a circuit with only two segments
                leftPowered = true;
                rightPowered = true;
            }
            else if (from == leftSegment)
            {
                leftPowered = true;
                if (rightSegment != null)
                    rightSegment.Pulse(this);
            }
            else if (from == rightSegment)
            {
                rightPowered = true;
                if (leftSegment != null)
                    leftSegment.Pulse(this);
            }
        }
        UpdateMat();
        pulsing = false;
    }

    // Update is called once per frame
    public void UpdateMat() {
        Material mat;
        if (hasBattery)
        {
            //Batteries work a little different. Never red.
            if (leftSegment != null && leftSegment.leftPowered && 
                rightSegment != null && rightSegment.rightPowered)
            {
                mat = poweredMaterial;
            }
            else
            {
                mat = oneWayMaterial;
            }
        }
        else
        {
            if (leftPowered && rightPowered)
            {
               mat = poweredMaterial;
            }
            else if (leftPowered || rightPowered)
            {
                mat = oneWayMaterial;
            }
            else
            {
                mat = deadMaterial;
            }
        }

        //Apply the material to every assigned mesh
        foreach (var rend in meshesToPaint)
        {
            rend.material = mat;
        }
    }
}

Now, in Unity, make a GameObject and attach the CircuitSegment script. Then add some meshes that you want to change colors. Tag them as “Wire” and make the CircuitSegment object its parent. Then also make two connectors, attach the CircuitConnector script to each, and mark one as isLeft. Make them also children of the segment. Now, if you have some object you want a complete circuit to trigger, like opening and closing a door, make a new script with two public methods such as Open() and Close(). On the segment that powers the object, you can assign those methods in the inspector through the Power Event and Lose Power Event.

Ten Super Helpful Unity Assets

As I’ve been working on Nebula Gladiator VRFly Around and Zap Aliens,  and Fail to Win, I’ve relied heavily on the Unity Asset Store. It is a great way to avoid reinventing to wheel. Unity itself all the basic tools you need (physics, animation, etc.), but to get a good starting point for your game, it is pretty much essential to add some packages from the Asset Store. Many of the packages cost a small amount (I have five packages ranging between 5 and 20 dollars), but many are free as well. Anyway, I’ve played around with a lot of Unity packages and found a few to be particularly useful. I’ve limited this list to just code and effect packages, since those tend to be more generally applicable.

1. Ragdoll and Transition to Mecanim

This is what I’m using for the ragdoll transitions in Fail to Win. In a 3D game where your avatar can get hurt (which is almost all of them), using ragdolls is a lot easier than creating death animations. It also tends to respond better to forces than pre-set animations. This tool by BzSoft, makes it even easier by automatically turning a character into a ragdoll when, for example, you fall from a really high ledge. Then, when things calm down, the avatar get back onto his or her feet. Also, it’s free.

2. Procedural Lightning

I am using this to do the zapping part of Fly Around and Zap Aliens. I used to have a simpler procedural lightning effect, but this looks way better. It has a surprising amount of configuration, so if you need anything that even somewhat resembles lightning or sparks, you can probably use this. Despite that, it is super easy to set up. It only costs $9.

3. Log Viewer

This solved a frustrating problem. I tested everything in the Editor, and it worked great. Then I made a build and deployed it to my phone. I opened it up to find the game broken. I could dig around and find the log file, but finding problems that way takes more time, and you may have to dig through a lot more. I then added this to my project and could open the log from my phone with a simple circle swipe gesture. It looks like the Unity console. It also displays the frame rate, memory usage, and other stats. This could be super useful for collecting information from beta testers. This package is free.

4. ProBuilder

So, I’m not much of an artist, and even if I was, I wouldn’t want to go into Blender or Maya and design a level just to test a simple idea. So I spent a lot of time carefully aligning primitives (mostly cubes) into rooms. The Standard Assets Prototyping package helped a little, but it still lacked the kind of control I needed. ProBuilder is a free package that lets you model directly withing the Editor. Granted, it would not be ideal for anything as complex as character design, but it is fantastic for level design.

5. FinalIK

This is the priciest package I’ve listed so far as $90, but if your game involves a lot running around and climbing on stuff, this may make your animations look a thousand times better. I got tired of having floating feet when walking on slopes, so I was really excited to finally have a way to fit my animations seamlessly to the stage just like in AAA games. This package does a lot more than stick feet to floors. Climbing, opening doors, pressing buttons, and pretty much anything else that involves animation and an avatar’s surrounding are going to look a lot better with FinalIK. I’m even using it in Fly Around and Zap Aliens, although I had to be sure to turn it off every time you become a ragdoll; otherwise, weird stuff happens.

6. Mesh Slicer

Another BzSoft package that makes it easy to create a more fun game. It lets you cut a mesh along a plane, or even along the path of a knife (or sword). It is a super satisfying effect in VR. That’s why I added it to Nebula Gladiator VR. A VR sword isn’t going to meet any resistance except maybe when a family member or roommate walks through the living room at the wrong time and gets hit with a controller, so a VR sword should cut cleanly through anything it hits. It got even better when I wrote a simple script that forces the pieces to fly away from each other after being cut.

7. Head Look Controller

I’m amazed this package still works, since it hasn’t been updated since 2010. It is free, and simply makes people look at stuff. There may be some other similar packages that do the same thing and are more up to date, but most of the time, this package should be all you need. It is super easy to set up and is another subtle effect that really makes a game look so much better.

8. MK Glow

It is a nice-looking glow effect that can be added either to individual objects or to a whole scene. I’m probably going to add it to Fly Around and Zap Aliens soon. I’ve played around with it before, and it is really easy to set up. They even have a free version. The upgraded version is only $10, and offers some extra control, but the free version is already super useful.

9. Unity Particle Pack

This is a Unity Essentials pack, so there’s a good chance you have already been using it, but I had to list it here because of how useful it is. I am using its explosions in Fail to Win, but a lot of its more subtle effects like dust and sparks can add the right kind of detail to a game. Most games are going to have a good reason to use at least one of those effects.

10. Post Processing Stack

Another Unity Essentials package, but worth mentioning. When used right, this could help make your game look less like an obvious computer rendering and more like a cinematic masterpiece. I discovered it doesn’t work so well on mobile, but on other platforms, it helps give the finishing touches.

A Final Note

Many of these tools were things I discovered when I was noticing details in my games that made them seem like ugly amateur indie games instead of modern quality masterpieces. Thank you to all the creators of these tools. The asset store is a fantastic resource. I’ve both sold and purchased tools on the asset store, and the system is great. I, of course, haven’t tried everything on the asset store (there are thousands!), so I’m curious if any of my Unity gamedev readers here have used a Unity package they consider a must-have. If so, please share in the comments below.

Fly Around and Zap Aliens Beta

I recently released a game on the Google Play store called Fly Around and Zap Aliens. It is currently in an open beta, so you are welcome to try it out, although I plan change some things in future. You can download it here: https://play.google.com/store/apps/details?id=com.kenningtongames.spacegame.

History

This game actually started back in high school. I had previously made a simple space game in Flash where you and another player orbit a planet, trying to knock each other out of orbit. When I was learning Unity, I thought I would try making a 3D version of the same game. However, I quickly ran into some serious design issues:

  • There’s a lot more space in three dimensions. Traveling from one side of the planet to the other felt like it took way too long, and it was hard to tell how fast you were going until you crashed into something, unable to slow down fast enough.
  • Your field of view is limited, since you are following a spaceship instead of getting a full view of the arena. This made the pull of gravity really confusing. Since there is no up or down in space, the pull of the planet onto the ship just made uncomfortable controls.
  • Multiplayer would be trickier than simply sharing a keyboard with a friend.
  • Boundaries are unclear. In 2D, it was simply the edge of the screen. In 3D, I needed a way to box players in and still have them experience the openness of space.

So I made an entirely new game with a similar-looking stage. There is still a planet in the center, and you still control a spaceship, but now you are defending the planet from aliens. You zap them instead of knocking them backward. There is no real gravity from the planet, but you still fly around it. I originally had a force-field for boundaries, but my force-field effect looked awful. The game was pretty much playable, though. I never could settle on a name. I then took a two-year break from game development and forgot all about it. More recently, I was looking through my old projects and re-discovered it. I had learned a lot since made it (it’s been about five years now), so I decided to start fixing it up.

The Game Now

I chose to port it to mobile because of the availability of tilt controls. Really, a platform like the Nintendo Switch would be more ideal, since it has a little more graphics power than a mobile phone while stile having the same tilt control, but Android is an easier platform to begin publishing for. The Asset Store has come a long way since I started this project, so I was able to make things look a lot nicer, but the game works in pretty much the same way. I did, however, change the boundaries, using a portal instead of a force-field. If you fly too far away from the planet, a portal appears, taking you to the opposite side of the planet, flying toward it instead of away from it. I released it recently to the Play Store, but it still has a few problems. For example, it is possible to dodge the portal and keep flying away. Most people I show the game are initially confused by the controls. It doesn’t take long to get used to them, though. Also, the end screen is kind of boring. I hope to fix these things and more. Still unable to choose a name, I chose the most concise way I could explain it.

The Plan

Between now and the full release, I plan to add the following:

  • Appearance tweaks (Already swapped out the spaceship models; just need to publish that change).
  • Fix the problem of dodging the portal and venturing into the void of space.
  • Add a leader board
  • Have the aliens come in waves instead of a steadily increasing rate.
  • A tutorial
  • Optimizations and device support (If you try the beta, and it does not work on your phone, let me know!)
  • An iOS port

These updates will come gradually as I make them, leading up to the official release. Please try it out and let me know what you think!

Player Motivations: for Novelty or for Sport?

There are many ways to classify players. I most commonly hear casual versus hardcore. This classification is not very useful, since it seems to describe the level of commitment to a game, not the type of commitment. I struggled to classify myself, since I enjoy both casual and hardcore games. I noticed, however, that I had a very different play style compared to many other players, in both casual and hardcore games. I realized that these differences in style could be explained by a difference in motivation. So I propose a new classification based on two kinds of player motivations: for novelty versus for sport. These two groups can be described like this:

Players for novelty (adventurers):

  • Believe games are about trying new things
  • Prioritize accessing new content
  • Care very little about metrics
  • Play recklessly
  • Embrace randomness and imbalance
  • Like things to change

Players for sport (competitors):

  • Believe games are about overcoming challenges
  • Prioritize winning
  • Focus on metrics
  • Play cautiously
  • Get outraged at randomness and imbalance
  • Like stability

This is a spectrum, and while many players will fall somewhere between the two, the direction a player leans will greatly impact the decisions that player makes during the game. I, for example, lean far on the novelty side, meaning I am more of an adventurer than a competitor. Even in physical sports, which obviously are designed more with competitors in mind, I would try to make the game more novel. Because of this, I wasn’t the most competitive. I remember one time in elementary school when I got bored during a game of dodge-ball and thought, “How long could I last without moving my feet?” The children on the other side of the spectrum thought I was being an idiot, since not moving my feet would obviously put me at a disadvantage. I didn’t care about the disadvantage; I just wanted to see what would happen.

Another example, this time with video games, is Super Smash Bros. I like to play timed battles with a random character, on a random stage (no stages removed from the random selection), with all items turned on and with the maximum number of players. Many Smash players probably cringed while reading that last sentence. This introduces tons of luck. You could end up winning a round, not because you have better tactics and reflexes than another player, but because the perfect item happened to drop right in front of you while your opponent is distracted by a random stage hazard, allowing you to deal a final blow a second before the clock hits zero. Many players try to avoid these situations by only playing stock mode on the simple Final Destination stage with all items turned off.  Such a player will feel satisfied that any victory was a result of his or her own skill, and nothing else. It is the pleasure of learning to overcome a difficult problem. Players like me, though, would look at this setup and think, “Why spend all your time playing only half a game?” Sure, the game is unfair and unpredictable with my preferred setup, but each match has a higher chance of seeing a combination of events I have not seen before, and for players like me, that means more fun.

You have two types of players who will judge your game in very different ways. One type wants more options. They are forgiving of bugs and glitches, and may actually appreciate them if if they do not completely block game progress. The other type wants games to be fair and predictable. They are more willing to replay the same or similar content without getting bored, but also more likely to complain about glitches, imbalance, and random number generators. They will prefer existing game mechanics to be deep rather than be introduced to new game mechanics.

Designing for Player Motivation

It is important to consider both of these kinds of players in game design. Many games support play styles preferred by both groups. In farming simulators, for example, players for novelty can focus their attention befriending villagers, collecting rare items, etc., while players for sport can focus on becoming millionaires. These differences can often make interesting gaming communities. An adventurer playing for novelty might discover a glitch in a game that is then used by a competitor playing for sport to set a new speed-running record.

Some game mechanics, however, tend to appeal to one side of the spectrum more than the other. Sandbox and open world games tend to attract adventures playing for novelty since they encourage discovery and do not have as clear of goals. Competitive and linear games attract players for sport because they are more controlled, more fair, and more clear about objectives. By considering these two kinds of audiences, we can tailor the details of a game’s design to the things our players most hope to find.

Of course every player is unique, so no classification is perfect. Most adventurers need at least some goals to get started. Most competitors will appreciate the occasional game changer. Nearly every game design will need to account for both motivations, regardless of the expected audience. Many game design decisions are highly controversial because they appeal more to one side of the spectrum than the other, earning game developers both praise and criticism. By knowing our players’ motivations and by designing for the different resulting play styles, we can design games that both can enjoy.

Where are you on the player motivation spectrum, and how has that affected how you play games? Do have any other ways you like classify play styles? I’d love to hear about it in the comments below.