Feedback Loops

One kind of dynamic that is often seen in games and deserves special attention is known as the feedback loop. There are two types, positive feedback loops and negative feedback loops. These terms are borrowed from other fields such as control systems and biology, and they mean the same thing in games that they mean elsewhere.

A positive feedback loop can be thought of as a reinforcing relationship. Something happens that causes the same thing to happen again, which causes it to happen yet again, getting stronger in each iteration – like a snowball that starts out small at the top of the hill and gets larger and faster as it rolls and collects more snow.

As an example, there is a relatively obscure shooting game for the NES called The Guardian Legend. Once you beat the game, you got access to a special extra gameplay mode. In this mode, you got rewarded with power-ups at the end of each level based on your score: the higher your score, the more power-ups you got for the next level. This is a positive feedback loop: if you get a high score, it gives you more power-ups, which make it easier to get an even higher score in the next level, which gives you even more power-ups, and so on.

Note that in this case, the reverse is also true. Suppose you get a low score. Then you get fewer power-ups at the end of that level, which makes it harder for you to do well on the next level, which means you will probably get an even lower score, and so on until you are so far behind that it is nearly impossible for you to proceed at all.

The thing that is often confusing to people is that both of these scenarios are positive feedback loops. This seems counter intuitive; the second example seems very “negative,” as the player is doing poorly and getting fewer rewards. It is “positive” in the sense that the effects get stronger in magnitude on each iteration.

There are three properties of positive feedback loops that game designers should be aware of:

They tend to destabilize the game, as one player gets further and further ahead (or behind).
They cause the game to end faster.
The put emphasis on the early game, since the effects of early-game decisions are magnified over time.

Feedback loops usually have two steps (as in my The Guardian Legend example) but they can have more. For example, some Real-Time Strategy games have a positive feedback loop with four steps: players explore the map, which gives them access to more resources, which let them buy better technology, which let them build better units, which let them explore more effectively (which gives them access to more resources… and the cycle repeats). As such, detecting a positive feedback loop is not always easy.

Here are some other examples of positive feedback loops that you might be familiar with:

Most “4X” games, such as the Civilization and Master of Orion series, are usually built around positive feedback loops. As you grow your civilization, it lets you generate resources faster, which let you grow faster. By the time you begin conflict in earnest with your opponents, one player is usually so far ahead that it is not much of a contest, because the core positive feedback loop driving the game means that someone who got ahead of the curve early on is going to be much farther ahead in the late game.
Board games that feature building up as their primary mechanic, such as Settlers of Catan. In these games, players use resources to improve their resource production, which gets them more resources.
The physical sport Rugby has a minor positive feedback loop: when a team scores points, they start with the ball again, which makes it slightly more likely that they will score again. The advantage is thus given to the team who just gained an advantage. This is in contrast to most sports, which give the ball to the opposing team after a successful score.

Negative feedback loops are, predictably, the opposite of positive feedback loops in just about every way. A negative feedback loop is a balancing relationship. When something happens in the game (such as one player gaining an advantage over the others), a negative feedback loop makes it harder for that same thing to happen again. If one player gets in the lead, a negative feedback loop makes it easier for the opponents to catch up (and harder for a winning player to extend their lead).

As an example, consider a “Kart-style” racing game like Mario Kart. In racing games, play is more interesting if the player is in the middle of a pack of cars rather than if they are way out in front or lagging way behind on their own (after all, there is more interaction if your opponents are close by). As a result, the de facto standard in that genre of play is to add a negative feedback loop: as the player gets ahead of the pack, the opponents start cheating, finding better power-ups and getting impossible bursts of speed to help them catch up. This makes it more difficult for the player to maintain or extend a lead. This particular feedback loop is sometimes referred to as “rubber-banding” because the cars behave as if they are connected by rubber bands, pulling the leaders and losers back to the center of the pack.

Likewise, the reverse is true. If the player falls behind, they will find better power-ups and the opponents will slow down to allow the player to catch up. This makes it more difficult for a player who is behind to fall further behind. Again, both of these are examples of negative feedback loops; “negative” refers to the fact that a dynamic becomes weaker with iteration, and has nothing to do with whether it has a positive or negative effect on the player’s standing in the game.

Negative feedback loops also have three important properties:

They tend to stabilize the game, causing players to tend towards the center of the pack.
They cause the game to take longer.
They put emphasis on the late game, since early-game decisions are reduced in their impact over time.

Some examples of negative feedback loops:

Most physical sports like Football and Basketball, where after your team scores, the ball is given to the opposing team and they are then given a chance to score. This makes it less likely that a single team will keep scoring over and over.
The board game Starfarers of Catan has a negative feedback loop where every player with less than a certain number of victory points gets a free resource at the start of their turn. Early on, this affects all players and speeds up the early game. Later in the game, as some players get ahead and cross the victory point threshold, the players lagging behind continue to get bonus resources. This makes it easier for the trailing players to catch up.

My grandfather was a decent Chess player, generally better than his children who he taught to play. To make it more of a challenge, he invented a rule: if he won a game, next time they played, his opponent could remove a piece of his from the board at the start of the game (first a pawn, then two pawns, then a knight or bishop, and so on as the child continued to lose). Each time my grandfather won, the next game would be more challenging for him, making it more likely that he would eventually start losing.

Use of Feedback Loops

Are feedback loops good or bad? Should we strive to include them, or are they to be avoided? As with most aspects of game design, it depends on the situation. Sometimes, a designer will deliberately add mechanics that cause a feedback loop. Other times, a feedback loop is discovered during play and the designer must decide what (if anything) to do about it.

Positive feedback loops can be quite useful. They end the game quickly when a player starts to emerge as the winner, without having the end game be a long, drawn-out affair. On the other hand, positive feedback loops can be frustrating for players who are trying to catch up to the leader and start feeling like they no longer have a chance.

Negative feedback loops can also be useful, for example to prevent a dominant early strategy and to keep players feeling like they always have a chance to win. On the other hand, they can also be frustrating, as players who do well early on can feel like they are being punished for succeeding, while also feeling like the players who lag behind are seemingly rewarded for doing poorly.

What makes a particular feedback loop “good” or “bad” from a player perspective? This is debatable, but I think it is largely a matter of player perception of fairness. If it feels like the game is artificially intervening to help a player win when they don’t deserve it, it can be perceived negatively by players. How do you know how players will perceive the game? Playtest, of course.

Eliminating Feedback Loops

Suppose you identify a feedback loop in your game and you want to remove it. How do you do this? There are two ways.

The first is to shut off the feedback loop itself. All feedback loops (positive and negative) have three components:

A “sensor” that monitors the game state;
A “comparator” that decides whether to take action based on the value monitored by the sensor;
An “activator” that modifies the game state when the comparator decides to do so.

For example, in the earlier kart-racing negative feedback loop example, the “sensor” is how far ahead or behind the player is, relative to the rest of the pack; the “comparator” checks to see if the player is farther ahead or behind than a certain threshold value; and the “activator” causes the opposing cars to either speed up or slow down accordingly, if the player is too far ahead or behind. All of these may form a single mechanic (“If the player is more than 300 meters ahead of all opponents, multiply everyone else’s speed by 150%”). In other cases there may be three or more separate mechanics that cause the feedback loop, and changing any one of them will modify the nature of the loop.

By being aware of the mechanics causing a feedback loop, you can disrupt the effects by either removing the sensor, changing or removing the comparator, or modifying or removing the effect of the activator. Going back to our The Guardian Legend example (more points = more power-ups for the next level), you could deactivate the positive feedback loop by either modifying the sensor (measure something other than score… something that does not increase in proportion to how powered-up the player is), or changing the comparator (by changing the scores required so that later power-ups cost more and more, you can guarantee that even the best players will fall behind the curve eventually, leading to a more difficult end game), or changing the activator (maybe the player gets power-ups through a different method entirely, such as getting a specific set of power-ups at the end of each level, or finding them in the middle of levels).

If you do not want to remove the feedback loop from the game but you do want to reduce its effects, an alternative is to add another feedback loop of the opposing type. Again returning to the kart-racing example, if you wanted to keep the “rubber-banding” negative feedback loop, you could add a positive feedback loop to counteract it. For example, if the opposing cars get speed boosts when the player is ahead, perhaps the player can go faster as well, leading to a case where being in the lead makes the entire race go faster (but not giving an advantage or disadvantage to anyone). Or maybe the player in the lead can find better power-ups to compensate for the opponents’ new speed advantage.