The Suspense is Skilling Me: Punishment, Learning, and Tension

I.

Long ago, I wrote a post about the different roles of challenge and punishment in skill-based games and how they relate to flow and learning. My argument was that challenge should vary with player skill to maximize opportunities for flow while punishment should be flat-out minimized to prevent disruptions to learning. Doing things like kicking the player back to a distant checkpoint when they die inserts delays and distractions between attempts, making it much harder to learn. But there’s a significant difference between first learning a skill and mastering that skill, and this absolutely affects what kind of punishment is appropriate. I’ll explain, borrowing an example from commenters on that old post.

Imagine you are playing a new racing game. The tutorial teaches acceleration, braking, steering, and drifting, requiring you to perform each operation before advancing to the next. You hold the accelerate button, then the brake button, steer around some turns, and then try the drift but your timing is off and you fail to execute it. In this case, it would be counterproductive for the game to force you to start all the way over and pass the accelerate, brake, and steer tests again before giving you another chance to drift. The game is teaching the skill, not testing it. Failing to execute this skill should result in an immediate opportunity to try again. Additional punishment would just make it harder to learn, which is the exact opposite of the tutorial’s goal. A punishing tutorial is a bad tutorial.

But once you’re out of the tutorial and you start racing, the scenario is different. The game is done teaching new skills and starts testing them. You are no longer learning skills; you are practicing them. Your goals are larger in scope - not just “perform a drift” but “win this three-lap race.” And because the scope of punishment defines the scope of challenge, a challenge of this scope is not possible without real punishment. If losing the race results in just restarting, say, the final lap, the challenge becomes “win this lap” rather than “win this race.” In order to challenge you to perform well consistently enough to win an entire race, loss must cost you the entire race.

Scoping punishment to cover multiple or lengthy tests of skill effectively increases challenge by demanding greater consistency and endurance. This is true whether the skill tests are interconnected like the moments of a race or independent like a series of difficult jumps in a platformer - such as 2009’s ‘Splosion Man.

“In [‘Splosion Man’s] Hardcore mode the game is exactly the same game, testing exactly the same skills in exactly the same levels, the only differences are that all enemies kill you in a single hit . . . and there are no checkpoints at all, so every time you die you have to restart the whole level . . . .

[I]t is a lot more punishing, but it . . . becomes more challenging too. You don’t now have to get skilled enough at performing a jump until you can do it once, twice or three times in a row perfectly, now you have to make perhaps 50 perfect jumps in a row. . . . You have to get so skilled that ‘you can’t get it wrong’ not just skilled enough to ‘get it right once’."
—Remy77077, Is ‘Splosion Man Challenging or Punishing?

But once you’ve cleared the game and unlocked Hardcore mode, you’ve learned all the challenges and skills. Hardcore mode presents an opportunity to master them. Just like with the aforementioned multiple-lap race, in order to test that you’re good enough to complete a level flawlessly, making a mistake has to cost you the entire level. The disruption to learning isn’t a problem when the game isn’t teaching something new and each retry is instead practice so you can attain and demonstrate the required degree of mastery.

Punishment that undoes progress makes sense for mastery challenges, not for learning challenges.

II.

The different approaches to punishment make the challenge feel different as well. If punishment is scoped to an entire race or level, then winning one lap or clearing a few jumps doesn’t make you safe. Until you complete the race or the level, you aren’t finished and your progress can still be undone by a mistake. You can still fail. And thus, tension is still high.

Tension is a subjective thing and different people process it differently (and we’ll get to that), but for now let’s define tension as the emotional state caused by the possibility of failure. It’s a signal that you need to focus and do your best, because doing so will make a real difference. As such, tension requires uncertainty - if you know you will succeed or that you definitely can’t, there is no tension - and it’s at its highest when the odds of success are a coin flip. Tension also requires stakes - if there is no punishment for failure or reward for success, there’s no tension. Greater punishments or rewards thus increase tension (though due to loss aversion, the effect of punishment is often greater than that of equivalent reward).

Tension in games thus occurs during (and in the lead-up to) tests of skill, proportional to how closely the challenge level matches the player’s skill level (creating uncertainty) and to the scope of the associated punishment (creating stakes). Once an outcome is achieved - whether it’s success or failure - uncertainty vanishes, the player receives their punishment or reward, and the tension is relieved.

There is no single universally-correct level of tension, but if you’re looking to test your skills and get into a flow state, too much tension is exhausting and too little is boring. This is true for both the magnitude of the tension and its duration, so carefully curating the expected level of tension players will experience over time is an important facet of game design. Ideally, tension will rise and fall over time in a deliberate rhythm.

In a typical skill-based game, the ebb and flow of tension as a player progresses might look something like this:

Individual challenges create tension peaks that deflate once the challenge is complete and the player has a safe moment to catch their breath. However, the challenge level increases over time as does the player’s investment, so the overall tension level gradually rises. Periodically there are particularly significant challenges (such as a boss fight) that take the tension extra high but then reduce it even further. The first challenges after one of these local climaxes are often easier than the last ones before it - this gives the player more of a chance to catch their breath after ratcheting the tension up so high. Plus, climaxes provide obvious break points where players might put the game down for the night or until the next weekend, so providing a chance for the player to reacquaint themselves with the game’s challenges afterward increases reapproachability.

In my old post from 2009, my argument was something like: punishment should be scoped to a single peak of tension. In practical terms, this would mean that there should be a checkpoint after every challenge. The rejoinder, then, is that I had it backward - punishment shapes the peaks and defines what constitutes a single challenge. And while challenges should be kept tightly-scoped when teaching new skills, challenges of mastery can well have greater scopes.

Let’s go back to our ‘Splosion Man example. Normally, levels contain checkpoints and only individual sections between checkpoints must be completed without failure - pass one section and fail in the next, and you only need repeat the latter section. Hardcore mode removes the checkpoints and the entire level must be completed without failure. Compare a hypothetical tension graph for the same level on Normal and on Hardcore:

On Normal, each checkpoint provides a significant deflation of tension, as the potential punishment resets to zero every time you clear one. As a result, the tension only rises so high throughout the level. But on Hardcore, there’s much less deflation of tension. While there are still moments between individual jump challenges where the player can pause, relax, and catch their breath, the removal of checkpoints means a mistake can always destroy all progress in the level. This means that the potential punishment for failure continually increases the further the player progresses in the level, and thus so does the tension. Relief only comes once the entire level is completed, so tension rises quite high before that happens.

This is probably not what most first-time players want while they are still learning the game’s skills. But once a player has mastered the skills to the point where beating the level with checkpoints is a certainty, playing without them may be the only way to create enough uncertainty to raise tension to an enjoyable level.

Tension is a product of the both the relationship between skill and challenge level and the scope of punishment.

III.

However, there’s an important question here: what happens on repeat attempts? What’s the effect on tension when punishment is scoped greater than an individual challenge and failure means repeating others at which you’ve already succeeded? The answer depends completely on the nature of the game’s challenges.

I recently laid out four broad “phases of challenge” that can show up in games: preparation, strategy, tactics, and action. In most cases, asking you to repeat a challenge you’ve already completed is asking for a redo of the action phase. After all, if you’ve succeeded at the challenge then you probably don’t need further preparation or adjustments to your strategy or tactics. You know what to do, it’s just a matter of doing it again.

In an action game like ‘Splosion Man, much of the challenge is in the action phase - executing ‘Splosion Man’s abilities at the right time to overcome obstacles and proceed. So doing it again - even if it’s just like before - still has challenge, still has uncertainty, and can still have interesting levels of tension. It’s still valid as a test of mastery.

Compare this to, say, Phoenix Wright: Ace Attorney. When cross-examining witnesses in the trial segments, you must carefully consider their testimony while keeping in mind the available evidence - and when you spot a contradiction, you must indicate which statement is disproved by which piece of evidence. Doing so correctly will advance the trial; doing so incorrectly will incur a penalty and require you to try again. Earn too many penalties and your client is found guilty - game over, restart that portion of the trial.

This adds stakes to the act of presenting evidence, and in theory increases the tension. But here’s the problem - the challenge of cross-examination is completely about figuring out what evidence to present when. Once you know, you know - so when you replay trial segments, there is zero uncertainty and thus zero tension.

You’re just being tested on a single piece of information - the right evidence/testimony combination - and you already know the previous ones. There’s no reason to repeat them, no practicing to be done. Like the hypothetical racing tutorial that forces you to redo acceleration, braking, and steering before letting you drift again, it’s a learning challenge that’s being punished as a mastery challenge.

Due to the repetition and lack of tension, replaying trial segments can mean several minutes of tedium - a price high enough that many players (myself included) are extremely reluctant to pay it. When a player is deep into a trial and a high-uncertainty cross-examination occurs, that’s supposed to be a peak of tension - but a player unwilling to risk the punishment will likely consult a FAQ for the answer. This completely removes the uncertainty and thus still eliminates tension, saving time but drastically reducing the satisfaction of success. Either way, punishment in games like Ace Attorney is not a tension regulator but a tension destroyer.

To be clear, the difference isn’t just about action games versus non-action games. Punishment only makes sense when it provides a valid opportunity to practice mastery and maintains a good level of tension. Consider roguelikes and score attack games - these are effectively always played on Hardcore, since failure requires a full restart. But even though these come in both action and turn-based varieties (Enter the Gungeon versus Dungeons of Dredmor, Tetris versus Bejeweled) this level of punishment makes sense across the board. Due to the random or procedural generation involved in such games, each playthrough is unique and presents fresh challenges to navigate while practicing general skills rather than a tedious retread of identical challenge-free content. Tension is thus preserved.

Alternatively, consider action games that use punishment to lump together challenges testing different skills. My go-to example for this is classic Mega Man games, where each stage features platforming and shooting challenges and ends in a boss fight with a specific pattern and weakness to learn - and if you run out of lives at the boss you must replay the entire stage. Like Hardcore mode in ‘Splosion Man, this might make sense as a mastery challenge once you’ve learned all the stages and bosses, but the first time through when you are trying to learn the boss, it’s more like the broken racing tutorial or Ace Attorney punishing a learning challenge as though it were a mastery challenge. Every time you redo the stage leading up to the boss, skill level increases, uncertainty decreases, and tension decreases. The delay is disruptive to your ability to learn the boss’s patterns and weakness, increasing the uncertainty of that challenge but also making the punishment particularly tedious and increasing the stakes in an unpleasant way.

Tension is the key to proper scoping of punishment. It’s all well and good to say that punishment should be scoped to the particular challenge or skill being tested - but since the scope of the punishment defines the scope of the challenge and endurance and consistency are themselves skills, this doesn’t actually lead to clear advice in many cases. Instead, punishment should be scoped to keep tension at levels that are neither too high nor too low. And punishing a learning test like a mastery test not only inhibits learning but distorts the tension curve in the worst possible way, eliminating tension before the challenge and skyrocketing tension during it.

Punishing a learning challenge as though it were a mastery challenge replaces tension with frustration and tedium.

IV.

Different players process tension very differently. I would love to see some studies here, because what I’ve seen suggests a few possibilities and I’m not sure exactly what’s happening. It may be that sensitivity to tension varies from person to person just as does sensitivity to physical pain, meaning that a given amount of uncertainty and stakes will create a boringly-low tension level for one person, a perfectly-engaging level for another, and an exhaustingly-stressful one for a third. It may also be that some people find that tension increases the satisfaction of success more than it increases the frustration of failure and others find the reverse. It could be that some people find learning a new skill more enjoyable than honing that skill and thus like to quickly move on to the next, while others don’t much enjoy new skills unless they can master them at a high level before moving on. Or it could be some combination of these. The effect appears to be that some players get more out of learning challenges and others get more out of mastery challenges - and therefore prefer different levels of punishment.

While we recognize that different people have different skill levels and therefore are best served by different levels of challenge, learning/mastery preferences and their causes are invisible. It’s very easy to assume that your own experience is normal and thus fail to realize that different players - even ones with the same level of skill - are best served by different levels of punishment. And so we end up with some players claiming that punishment is necessary to make challenge meaningful, while others complain about how frustrating it is.

“[I]f you die in a game, you lose whatever time you put into the game from the last checkpoint or save spot. Maybe that’s 10 minutes, or maybe it’s 30 seconds. In exchange for that time, you presumably got a chance to practice your skills. You may do better on your next try. . . . I usually consider the loss of time due to my failure as a player as fair. It rarely even registers for me as a real-world cost to dying in a game. Yeah, I have to redo the last five minutes of this game, and if I hadn’t died, I could have spent that five minutes cleaning the house or watching TV. . . . But without failure, I wouldn’t learn and improve, and that cycle of failure -> learning -> improving -> success is such a fundamental part of so many video games that it barely registers as punishment."
—Kirk Hamilton, Five Ways Video Games Make Failure Matter

“If I’m playing Dark Souls and I can’t get the hang of the timing on a boss, then I end up dying and slogging through a bunch of trash mobs before I can try again. That will take a few minutes. By the time I get back to the point where I made the mistake, I’ll have lost track of what I did wrong. . . . This punishment is not enhancing my enjoyment of the game. It’s actually enraging. I become angry and frustrated and I resent every single moment wasted trying to fight my way back to where I left off.

For me (and I imagine a lot of people who dislike Dark Souls) the cycle of punishment doesn’t make the eventual victory ‘more rewarding’. It takes away the fun I would otherwise be having throwing myself at a challenge again and again until I master it. The thing that ruins the game for me is the same ingredient that makes the game so compelling for fans. You literally can’t please one group without ruining the game for the other. . . .

So I get why we so often end up in these absurdist arguments where a Dark Souls fan is explaining to you earnestly and with no hint of irony that you would enjoy the game if you just played it differently, or thought about it differently, or followed their specific build advice. Many people really do think I’m somehow ‘missing out’ by not playing this game."
—Shamus Young, Arkham City Part 3: A Difficult Discussion on Difficulty

I was guilty of this failure of empathy when I first wrote about punishment. It was so obvious to me that the stress of punishment far outweighed the satisfaction of eventual victory that the only reason designers kept including punishment and players kept defending it was that they didn’t understand it was the worst way to increase difficulty. But I see now that punishment legitimately has a different effect on different people, and to many it’s a satisfying way to raise the stakes high enough to create an interesting level of tension for a mastery challenge.

Some games - Dark Souls, Getting Over It with Bennett Foddy, the aforementioned roguelikes - wear their high punishment levels and frequent or extended mastery challenges as a badge of honor. Other games - The Sims, Animal Crossing, exploration/storytelling games (so-called “walking simulators”) - do away with failure modes entirely and thus omit mastery challenges completely. That’s great - if you know what to expect from a game, then you can easily self-select into your desired punishment levels and challenge type and avoid the games you’d find unpleasant. The trouble is the huge and poorly-defined space in between.

Most skill-based games don’t clearly advertise their punishment levels or strictly adhere to well-understood standards, making it hard for players to know ahead of time whether a game will actually suit their tastes or leave them disappointed. Worse, many games start with a low punishment level but then crank it up right alongside challenge as part of the difficulty curve. That’s certainly what some players want, but to others that means the game starts well but gradually becomes unplayable.

Super Meat Boy punishes mistakes with an instant restart, which is very low punishment for the early, short, and easy stages but becomes brutally high punishment for later, longer, and fiendishly difficult stages. While I loved the challenge level in that game, the punishment grew too frustrating and disruptive when I was trying to learn to beat the later stages and I abandoned the game. Had I been able to keep punishment low, such as by adding checkpoints to later stages, I’m confident I would have finished it.

It would be easy for someone who prefers mastery challenges to misinterpret my behavior as meaning I simply found the game too hard, and unless I understood where that person was coming from it would be very difficult for us to have a productive conversation on the matter. But the truth is that due to the increasing scope of its punishment, Super Meat Boy was presenting me with mastery challenges when I was still trying to learn.

Punishment levels aren’t normally chosen from an explicit menu. But Hardcode modes and other forms of optional checkpoints as well as things like speedruns and boss rush modes all provide ways to turn a game’s learning challenges into mastery challenges. The key is that they are under the player’s control. Most players would prefer to start with learning challenges, after which some will be done but some would like to proceed to mastery challenges. There’s no reason we can’t give all of these players what they’re looking for.

So. I no longer argue as I did in 2009 that we should have varying optional levels of challenge but only minimal punishment. Now I argue we should have separately varying optional levels of both challenge and punishment. Hard mode and Hardcore mode should both be a choice. Tying a game to a specific, uniform level of punishment would be like ‘Splosion Man requiring everyone to always play on Hardcore or omitting the mode completely. The number of people who could enjoy the game and how much fun they could get out of it would both be lower.

To increase audience size and replayability, let the player choose between learning and mastery challenges by choosing their own punishment level.