Tutorial 1

Feedback loops in games
(Note: it might take a while for all the dynamic content to load..)

Introduction
During his 1999 lecture at The Game Developers Conference Marc LeBlanc introduced feedback loops to the game design world (LeBlanc, 1999). Since then, feedback loops have been discussed by a number of influential designers, including Salen & Zimmerman (2003), Adams & Rollings (2007) and Fullerton (2008). From my own experience as a game designer and researcher I can testify for the importance of feedback loops. Understanding the feedback structure of a game can help to understand the dynamic behavior of the game and sheds some light on the ever illusive concept of gameplay. Especially for games that are mechanic driven (as opposed to games that are story or level driven). This tutorial aims to explain the concept and its relevance to game design. To this end it will make use of Machinations diagrams. This is a type of diagram designed to expose a game’s feedback structure. The Machinations diagrams and framework is part of ongoing research conducted by myself (Dormans, 2009). Do not worry if you do not immediately understand these diagrams. This tutorial was written to introduce both the concept of feedback and the Machinations diagrams.

Feedback Loops
The first thing you need to know about feedback loops, is that with feedback I am not referring to feedback aimed at a player to inform him or her about the state of the game. Although this is commonly also called feedback, it is not the same thing. Originally feedback was used in biology to describe relations between predator and prey populations. Since then it has played an important roll in control theory that concerns it self with cybernetic (or electronic) systems. A good example of such a system can be found in your refrigerator to control the temperature. There the cooler is controlled by a temperature sensor. As soon as the temperature climbs above a certain point, it starts cooling, only to stop when it drops below another fixed temperature. This way the system keeps the temperature of your refrigerator under control.

What, you might ask, does this have to do with games? Well, a lot. Feedback loops such as the one in your refrigerator can be found in many games. In fact, I still have to encounter the first enjoyable game that has no feedback loops. Games without feedback loops exist, but they are usually little fun or very, very simple.

An example of a feedback loop can be found in the well-know game of Monopoly. In Monopoly there are two important resources: money and property. You use money to buy property, and, with some luck, property earns you more money. Figure 1 below, is a diagram of this feedback loop found in Monopoly. The resources are represented by small round tokens collected on the two 'pools' (the two large circles). Money enters the game through the 'source' (the triangle pointing upwards) represents your income. The 'converter' (the triangle pointing to the right) represents the player action of buying property. The arrows indicate the flow of resources from one element in the diagram to another. The dotted arrow, finally indicates that the number of resources in the pool has a positive effect on the chance (indicated by the die symbol) the income source will produce more money.



You might notice the run button on the lower left corner of the diagram. One of the neat things about the Machinations diagrams is that the on-line versions are dynamic. Click run and see the income source actually produce money. Click on the converter to buy property. Playing around with the diagram you should quickly discover that buying property quickly increases your amount of money.

One of the interesting characteristics of Machinations diagram is that it visualizes the feedback loop. The four arrows (including the dotted) arrow form a closed circuit. When using Machinations diagrams to design or analyze games, similar closed circuits are important structural features indicating feedback loops. Actually the diagrams where designed with this purpose in mind.

Positive and Negative Feedback
What then, is the effect of a feedback loop on game? The effect of the feedback loop in Monopoly is already very different from the effect of the feedback loop in your refrigerator. Where feedback in your refrigerator keeps the temperature under control, the feedback loop in Monopoly does do the exact opposite: the amount of money quickly spirals out of control. This is because both feedback loops are of a different type; the feedback in your refrigerator is negative, the feedback in Monopoly is positive.

To appreciate both types of feedback lets look at a diagram that represents a simple game with no feedback: basketball (see figure 2). Basketball served as an example in Marc LeBlanc original presentation. In this diagram two teams of five players create opportunities to score. This is represented by the the two sources on the left side. However, team 1 is much better at playing basketball than team 2 is. This is represented by a 'random gate', the little diamond with the die symbol. The chance next to the connection leading out of this gate indicates the chance each token has to pass the gate, and become a point (see figure 2). The chart shows the development over time of the game of a few games. Finally, the construction on the bottom is a timer. Time is slowly draining away, and when it is gone completely the game stops.



A quick glance at the chart tells us that results of this game are obvious: team 1 wins... Most of the time anyway. There is small chance that team 2 wins, but is chance is so small it is almost neglectable. The pace at which both teams score, stay steady over time, and are more or less independent of each other. For a sport this might be okay, it is a fair measure of the abilities of both teams, but it is also somewhat uninteresting. There are no dramatic reversals in fortune, no opportunity for risk taking, and strategy does not come into play. There is nothing you can do to defeat your opponent. Obviously, this is a gross simplification of the tactics and strategies of involved in a real game of basketball, yet if the difference between the two teams is large, this is what happens. However, realism is not our concern here. The effects of feedback are and this model of Basketball is an illustration to show those effects. For example, we might introduce a positive feedback by letting the team that takes a lead of more than five points field an extra player for every five points they are leading. In a diagram this looks like see figure 3). Note that the difference between the two teams is calculated by adding 1 for every point the blue team and subtracting one for each point of the red team. These values are then used to modify the team size (but the teams cannot drop below 5 players).



The effect of this feedback is obvious, it helps the stronger team, there score goes up exponentially and they end up with a lot players on the field. Their score pace increases exponentially over time, their line curves upwards (and off the chart). When the two teams are of equal skill, the effects are also quite interesting (see figure 4) In this case the teams score points at roughly the same pace. Until one team takes a decisive lead and the feedback kicks in, at which point they will win the game for sure. For what team will benefit from this effect, or even if the effect is going to kick in at all is uncertain.



Negative feedback does something different. By giving the the trailing team an extra player for every five points they are behind the feedback is reversed (see figure 5). The performance of the trailing team is boosted when they are falling behind, but they will never be able to catch up or beat the other team (if there is a difference in skill). This is because the difference in points is what is actually causing the feedback and the boost to the trailing team. The system balances out to a steady difference in points. The chart in figure 5 indicates the effect quite clearly. The scoring pace of team 1 is unaffected, whereas the scoring pace of team 2 is boosted to match that of team 1; the two lines slope upward in parallel, but team 2 will always be behind.



Feedback in Tetris
Nobody plays positive or negative feedback basketball. The above examples are illustrative, but they are nothing more than theoretical spielerei. However, feedback loops just like the ones in positive and negative feedback basketball crop up in many games. Tetris, for example, has an interesting positive feedback loop. In Tetris you need to work hard to keep clearing the falling blocks. As you progress the game speeds up, making it more difficult to keep up with the falling blocks. So far, there is little feedback in the game. But to make matters just a little bit more complicated, the more blocks there are on the screen the more difficult it becomes to clear them: the player simply will have less time to find the right position. Figure 6 models this effect.



What happens is that, at first, the player is able to keep the number of blocks under control. But at a certain moment, the pace of falling the blocks grows too fast and the difficulty spirals up ending the game quickly. In figure 6 the feedback loop that causes this effect is highlighted. A similar feedback loops can be found in many arcade and casual games: in Space Invaders the pace of the aliens speeds up as the player shoots more aliens, in Puzzle Bobble the player is put under increasing pressure as the wall of colored bobbles slowly makes it way down.

Usually, one feedback loop does not do the trick. Many games rely on a few interacting feedback loops. Not too many, either. Too many feedback loops make a game complicated beyond comprehension and too unpredictable. There is more to feedback in Tetris, too.

In Tetris a player can score more points by clearing multiple lines at once. A good player will let the blocks stack up and look for opportunities to clear many blocks at the same time. This spices up the game, the mechanic that causes the feedback that will eventually finish the game also rewards the player with more points. Now the player has do a balancing act. How much risk does she take? And what is the best strategy? The keep clearing the blocks as quick as you can to prolong the game, or to ride the wave of probability and take some chances for a higher score? Figure 7 reflects this new structure. This time you need to clear the blocks yourself, allowing you to try out both strategies.



Both mechanics might look similar in figure 7. How does one tell one tell one constitute a feedback loop and the other does not? The answer is that the first mechanic by affecting the rate at which the converter is activated and thus the rate at which tokens drain from the pool that causes the feedback; it actually affects itself. The other mechanism does affect only the output of converter. In a Machinations diagram feedback can only occur when there is a closed circuit of correctly directed connections or when a mechanism affects the rate or input value of a converter or drain (an element where tokens disappear, represented by a triangle pointing down).

Fixing Monopoly
The model of Monopoly (figure 1) does not do the game justice. There is more to the game than just this feedback loop. In Monopoly there is an additional source of income from passing ’Go’ and through chance cards, there are other players competing for the same property and paying rent to each other. Figure 8 is a better model of the game. In this diagram there are two players who buy property from the same pool. The arrows that end in a star (*) are activators. When a resource travels along these arrows it activates the element it finds. Pools that are activated in this way pull resources from other pools they are connected with. This way activating the pool with money of the orange player pulls one resource from the pool of money of the blue player. In a similar fashion the buy action will cause a property resource to be pulled from the bank to the pool of a player. The two boxes marked AI indicate there is two artificial players active. All these do is to activate the 'buy' converter once in a while. The chance this happens decreases as the Bank has fewer proprty available. However, the chance this will happen is twice as large for blue than it is for orange.



The chart in figure 8 keeps track of the player money supply (fat lines) and property (thin lines). From the diagram you can see that in most cases blue will yake a lead in property but it takes a while for this lead to be reflected in the difference in wealth. The feedback is delayed for a while, making the game appear tighter than it really is. Of course, experienced players of Monopoly know that the person with the best property is going to win in the end.

The reason why the effects of the positive feedback appear delayed is because a player needs to invest the same resource the feedback loop is producing. The result is more or less the same all the time. If we take a way some of the randomness the typical shape becomes even more clear: a steady line that suddenly slopes upwards at a steeper angle (see figure 9, go on run it 100 times!). This a shape that is a direct consequence of a particular construction pattern in the mechanics of Monopoly. It is caused by an (Unpredictable) Dynamic Engine.



There is also another shape that is quite common if you run this diagram a lot (remember to clear the chart after running it 100 times, otherwise the perfomance is ruined). This pattern has both lines sloping upwards. In these cases, the steady income form passing go compensates for the difference between property and the game drags on forever.

More to come later...