How Games Use Feedback Loops

Hey. Mark here. Over Christmas, I played the latest game from Bastion and Transistor developer Supergiant.

The game’s called Pyre and I guess it’s a party-based RPG, but instead of your usual turn-based battles, you face off against other characters in these bonkers, basketball-style competitions called Rites. The game’s a lot of fun, and full of clever ideas and talking dogs. But what really stuck out to me, was how Supergiant did some very clever things with an important part of game design, called feedback loops.

Feedback loops are systems where the output is fed back into the system as an input. And when it comes to game design, this typically means that a player’s success or failure impacts the likelihood of future successes and failures. And feedback loops come in two delicious flavours - positive feedback loops, and negative feedback loops.

Let’s start with positive loops. These reinforce successes with more successes, and compound failures with more failures, and I think a good example of the first one is the kill streak system in the Call of Duty games, especially the earlier titles like Modern Warfare. In those games, getting a handful of kills in a single life grants you special “kill streak” bonuses, which might mean putting enemy locations on your radar, or allowing you to call in an airstrike or an attack helicopter.

Which ultimately means that those who are already doing well by racking up kills, are rewarded with the tools to do even better. This can have something of a snowball effect - a sort of “rich get richer” type outcome where successes are compounded. But, of course, positive feedback loops can also be about how losses lead to further losses.

Like in chess, where losing a piece from the board means that you now have fewer tactical options which could lead to even more losses down the line. On the opposite side of the coin are negative feedback loops, which balance out successes and failures. And there’s one example everyone always uses when talking about this stuff - but that’s just because it’s the perfect one.

it’s the item distribution in Mario Kart. So, players who manage to get to the front of the pack will find themselves getting rather weak items like green shells and banana peels. But those who are at the back of the crowd are handed really powerful weaponry like the bullet bill, the golden mushroom, and - of course - the infamous blue shell which specifically targets the person in first place.

So winners are punished, and losers are rewarded. If positive feedback loops work to compound changes in a game state, negative feedback loops strive to always counteract those changes and keep things in equilibrium. Now some games don’t have feedback loops at all.

In Tekken, for example, losing one round gives you no advantage or disadvantage in the second round. But loops do have their place. They might just be an inherent part of the game’s design, like how in Splatoon, the team with the most ground covered in ink will also naturally have the most hiding places and the fastest travel.

Or they can solve problems, like the issue of never being able to catch up to the first-place racer in a driving game. And both positive and negative loops have their advantages. Positive loops help push a game to its conclusion when one player or one team starts to win, which stops a game from getting stuck in a stalemate or having the ending be a long, drawn-out affair.

In Team Fortress 2, teams that control lots of capture points will see their respawn time reduced, allowing that team to compound their victories and push on for a quick, decisive win. And negative feedback loops can be great for party games where everyone should have a good time, and they create an interesting ebb and flow of comebacks and turnarounds which leads to good drama. But both need to be balanced carefully - positive loops are frustrating for weaker players who get stuck in a death spiral, or struggle to catch up to a runaway winner.

And negative feedback loops feel unfair to successful players, who constantly have their victories get taken away. And for an example of a feedback loop that can go very wrong, we should look to XCOM. Now, XCOM is primarily a singleplayer series, and so far I’ve been talking exclusively about multiplayer games.

And that makes sense: balance and fairness are extremely important factors in multiplayer design. But feedback loops exist in singleplayer games, too. In fact, some genres are built around positive feedback loops, like JRPGs where you kill monsters, which gives you experience points to level up and get stronger, which lets you kill monsters more quickly.

And so on. You get the addictive thrill of always getting stronger - just, with enemies getting harder as the game goes on to balance things out. Plus, you don’t get the other side of positive feedback loops - the compounding of losses - because you just retry when you die.

Your failures aren’t carried over. Except… in truly persistent games, like XCOM. Here, you don’t retry a lost mission but you just keep going.

And, on top of that, any soldiers who died during a failed expedition are gone forever, thanks to the game’s permadeath system. And so, Gunpoint designer Tom Francis noticed that XCOM has an unfortunate positive feedback loop - or a “snowball problem”, as he puts it. He says, “if your troops survive the mission, they get stronger, tougher and get more abilities, which makes them more likely to survive future missions and get tougher still.

If they die, they’re replaced by vulnerable, weak rookies, who are likely to die and be replaced by vulnerable, weak rookies. ” And that is a nasty issue, if you’re unlucky enough to fall into one of the two extremes on XCOM’s difficulty curve. Those struggling can’t stop and train up some new troops because you can’t go back to an easier area and fight low level baddies.

And with veteran troops constantly getting more health, they become less and less likely to die, so they become powerful steam rollers - which isn’t always that much fun. Now Tom’s got some super smart ideas for fixing this issue, like low profile missions that force you to use rookie soldiers, and making squad size a per-mission variable to stop you taking the maximum number of units into every fray. So go read his article for more on that.

But I found some really clever solutions to this problem in Pyre. So Pyre works a lot like XCOM. You have characters who level up and get more powerful when they win.

And the game always moves forward, so if you lose a match, the game just keeps going. And this could have been the source of some nasty, runaway positive feedback loops: you could use the same team members for every match, watching them get stronger and stronger as the game goes on to the point where you’re no longer challenged by anything the game has to throw at you. Or you could lose a few matches, and become so underpowered that you can never catch up.

But Pyre counteracts this issue, by pairing those positive feedback loops - with negative feedback loops. So, yes, losing matches sucks but - here’s loop number one - you still receive some experience points for failing. Meaning that you’re not punished too much for messing up.

Also, unlike XCOM, there’s no permadeath so you’re not massively screwed for future matches. And then on the other side of things, counteracting the issue of overpowered troops, there’s the liberation system. You see, Pyre is set in a sort of grim purgatory called the Downside - and the entire point of the game is to win these special liberation rites - because the prize is that the team’s leader gets to leave the Downside, have their crimes forgotten, and return to the Commonwealth.

And in the game, this means that at various intervals throughout the adventure, you permanently send away your most powerful soldiers, because you can only liberate those who are of a high enough rank So I guess Pyre does have permadeath. . .

but it’s a good thing. And all this means that the resulting experience - at least for me - is that the game is always a close and interesting challenge. Mistakes became a road bump on the way to the conclusion, not a tipping point that could send me into an inescapable death spiral.

And by constantly removing my best units from the roster, I was forced to carefully level up and use everyone on my team - and then fight the last, most difficult battles not with my best soldiers but the weaker units who had not yet been liberated. And that lead to some of the most tense, challenging, and downright enjoyable battles of the entire game. Oh, and here’s an important bit.

One big issue with negative feedback loops is that they can send mixed signals. As we talked about in this episode, games are carefully designed to entice and encourage players towards a certain way of playing - so punishing someone for successfully playing the game completely goes against those encouragements. This is one reason why is was so important for Capcom to never reveal that Resident Evil 4 was using a dynamic difficulty system - ultimately, a negative feedback loop.

Telling players that the game will get harder if they’re doing well, or easier if they’re struggling, is effectively telling good players to make mistakes. But this is never the case in Pyre. It is still better to win a match than to lose it.

And getting rid of people is the goal of the game, not to mention the fact that it's tied to character arcs and good drama. Call me soppy but I just had to send away my most effective and often-used characters, because I felt they truly deserved to get out of purgatory. There are actually more negative feedback loops in Pyre.

Like, during the basketball game, when someone scores a point they’ll have to sit out the next play - meaning that the losing team now has a pretty significant number advantage over their opponent. This is similar to real world basketball, because when a team scores, the other team starts with the ball. But I really like this idea of using negative feedback loops to cancel out the most destabilising properties of positive feedback loops.

Pyre definitely didn’t invent it - it’s just the game that helped me understand the dynamic. And now, hopefully, you too. So you can see the same thing happening in games like Civilization.

In that game you can wipe out another nation and immediately take their land, which could lead to a snowballing positive feedback loop where you massively increase the size of your empire every time you defeat another leader. But negative feedback loops, like unhappy citizens and the increased cost of running a larger empire, pushes back against over-expansion. Especially in the early game: because positive feedback loops often magnify over time, making those critical first few turns, kills, rounds, or decisions have an enormous impact on the eventual winner of the game.

Oh, and this is not to mention the fact that other nations will often team up and declare war on any nation that seems to be barrelling ahead. I’m pretty sure that the AI is programmed to do that - but it will naturally happen in multiplayer, too, as everyone gangs up on the leading player to stop them from winning. That’s like a naturally occurring negative feedback loop.

And this is also a big part of RPGs: the more you level up, the more it costs for you to level up, which stops you from standing around the opening area and boosting yourself to level 99 in 20 minutes. There are other ways to deal with runaway feedback loops, of course. A strong loop can simply have its effects dampened - which is what happened in Call of Duty.

In the earlier games, killstreak bonuses like the aistrike, used to count towards future killstreak bonuses - giving you this absolutely bonkers positive feedback loop where a predator missile would give you enough kills to give you an AC-130 which would give you enough kills to grant you a tactical nuke. That got changed in Call of Duty: Black Ops, where killstreak bonuses no longer count towards killstreak bonuses, so that players would need to actually shoot or stab other players themselves to get these rewards. And as for Mario Kart, a new item in Mario Kart 8, the super horn, does give you an effective strategy for destroying blue shells and helps fight back against the most frustrating part of the game’s negative feedback loop.

But sometimes it’s best to remove a feedback loop altogether. In Devil May Cry, getting a high score at the end of a level will grant a player more red orbs which are used to buy better weapons and health potions. This is a positive feedback loop that has the unfortunate effect of giving the best players a helping hand that they really don’t need, and means struggling players won’t have enough cash to get the items they require.

It can be a good idea to simply decouple rewards for doing well, from the progression of the game. Donkey Kong has these challenging-to-collect KONG letters but grabbing them all doesn’t earn you coins or extra lives or anything like that: instead, you get extra difficult bonus levels, which is a suitable reward for players who are already proving themselves to be good at the game. So, there we go.

Feedback loops are a key part of many types of game, from JRPGs and persistent strategy games, to online shooters and arcade-y racing games. But using them effectively - to keep multiplayer games fair and to keep singleplayer games challenging, requires some very careful consideration. Hey everyone!

Thanks for watching. I got a great response to my last video on design theory stuff, about encouraging different play styles, so I’m looking to do more stuff like that in 2018 - plus, more analysis videos, level design stuff, critiques, and, well, a bit of everything, really. Hit that subscribe button to make sure you never miss an episode.