Sunday, 24 January 2010

Why are games difficult?

You may want to start this series with part one, two, three, four or five.

Demon's Souls was lauded as one of the best and most challenging games of 2009, for the difficulty of both the game play and the placement the apostrophe in the title. It is this difficulty that was seen as a breath of fresh air by many jaded critics and game players, who had supposedly grown fat and weak on the Facebook Farmvilles and amusement park ride MMOs, with their watered down game mechanics designed to hook in and succor what Gevlon refers to as morons and socials.

Demon's Souls was criticised as one of the most harsh and unforgiving games of 2009, for the difficulty of both the game play and the placement of the apostrophe in the title. It is this unnecessarily hostile environment that continued the tradition of poor design catering only to hard core critics and game players, who have the obsessive compulsive disorders necessary and basement bedrooms at their parents' place necessary for the multi-hundred hour play times required by treadmill MMOs, with their grindtastic game mechanics designed to hook in and succor those Tobold parodies in the Rise of the Leet King.

Difficulty in game design is something I hold dear to my heart - mostly because I have written a game which is too difficult for me, or many others to ever contemplate completing. Witness the crushed and torn bodies of the many Shadow Fairies that litter the latest Angband competition ladder and the self-evident pride of the masochists posting that they've made level 5 (out of 50 - and level 50 is much, much more challenging again). As Konijn puts it:

Anyway, I think at this point, I give up. You cannot win with this character, there's instakill from a number of sources even while playing very carefully. I number my savegame tries, this is number 22, so you cant say I didnt try ;)
and then keeps playing...

The current shining exemplar of the roguelike genre, Dwarf Fortress, has a motto 'failing is fun', but roguelikes have always been notoriously difficult...

Let's start at the beginning. Tobold poses the question succinctly in a follow up post to the Leet King letter:
When you say World of Warcraft is too easy, how exactly should a good, hard game differ from that?
(An aside: the only literary technique more misplaced than writing a humorous opening to a serious question on the Internet - the mistake Tobold admits to - is posing a rhetorical question. It works in the context of the original post, but as soon as you're quoted elsewhere, people just assume you're stupid and answer the question, rather than following the link and finding out you've already got some answers. So only an idiot someone trying to be difficult would title an article with a rhetorical question. So if you're going to link to this, please call it 'The Seventeen Ways Games can be Difficult' - or something like that).

He misses a trick: the question shouldn't be how can we make games harder? It should be why are games difficult at all? And it's this, more fundamental question, that I'll be attempting to answer here.

Why are games difficult? We have conquered the world as a species, we are better off than every previous generation, we are - you especially - inherently bad-ass. As Neal Stephenson points out, provided you accept evolution as a theory, you are currently expressing more badassedness than any creature that has ever lived and died ever, by definition. And yet we can be challenged and beaten by something as a simple as a game - some lines on a ground, and some differing coloured pebbles.

We can be beaten because of random factors. As soon as you introduce the possibility of randomness in a game, there exists the possibility of events outside of your direct control. Something as simple as a dice, or a spinner, or something less biased like entropic heat exchange, can be used to simulate externalities: the environment around you, other competitors - either because you don't have the players available to represent them directly, or to represent conflicts for which it would be inconvenient or dangerous to play out in reality - the want of a nail for the shoe of a horse. Reality is more outrageous and unbelievable than the most fantastical fiction, and randomness is a fair arbiter of unlikely consequences.

Even when you are able to directly resolve events without a random number generator, stochastic forces ensure that the outcome is not pre-determined. In fact, many games require that. You may immediately think of games of chance, gambling, horse racing, boxing, but the continuum of indeterminacy extends to all sporting events. If the outcome of any sprint is that Usain Bolt is the winner, then no one will sprint against him - unless there is sufficient incentive to change the race to be 'by how much is Usain Bolt the winner' or 'by what amount will the world record held by Usain Bolt be reduced by this week'. Even pursuits as sedentary as chess can be influenced by illness, distraction, the mindset and level of preparation of the competitors.

Other players are the next reason games can be difficult. Multi-player suddenly increases the variability of the challenge enormously, both in directly competitive games, and cooperative games (PvE). Should I choose the optimal strategy, which requires that my opponent or my partner is also playing to this optimal strategy? If there are suboptimal strategies with reduced risk of loss, which one of these is predictable, which one less so? The skill set required to directly read the mind of the opponent, or yomi, and the complexities and strategies of game theory can be expressed in games as simple as rock, paper, scissors, or as complex as the stock market.

Implicit with competition is the notion of different outcomes. Games can be difficult because we don't necessarily have a binary notion of risk vs. reward in all games. And even games which are differentiated by simple winning and losing, the consequences of the strategy employed can vary widely depending on whether we anticipate repeat play with the same opponent, or if we carry resources (star players, information about the opponent's preferred strategy or risk aversion, preponderance of particularly strategies in the environment of possible players) from one game to the next. Whether you should be a hawk or a dove depends on who you flock with, in the prisoner's dilemma, your choice matters more on whether you'll meet your potential betrayer as a cellmate.

Reward or risk over time is another important factor. Unlike the perfect world of game theory, we are finite beings with immediate wants and needs, and long term goals. Over a sufficiently long term, everyone is a loser - dead, as well. There is evidence to suggest a negative correlation between time spent (computer) gaming, and income - does the time you devote to winning this hundred hour RPG meaning you're losing out on the bigger game of life? Even in the game, if there is a strategy that earns you 50G per hour, why are you employing a tactic that only earns you 25G in the same time.

But that's unfair - I hear you say, complaining on the forums and asking Blizzard to reach for the nerf bat. What I want to do in the game should be just as rewarding as the actions those people who bothered to go out and do the research are doing? Strategy X is clearly cheating, because I can't or won't do it, or don't enjoy doing it. Welcome to the world, scrub. You're not playing the game if you're not playing to win. You're playing some weird made up in your own head game that no one else on the server is actually playing. Feel free to keep doing it, but get out of the way of the real players.

But sometimes you just don't have the abilities needed to play at the highest level. Or the twitch reflexes.
Or the Internet connection. Or the money. Or clean water and parasite free food and a source of income able to keep you above the poverty line with a diet rich in enough protein and other essential nutrients to avoid crippling chronic disease. It should be self-evident that life is unfair, from the moment of conception, and it is only the badass inheritance of your genes that keeps you floating atop the cesspool of the real world, fighting over the last scrap of food at the table.

And that's before you sit at the table. When you do, are you playing black or white? Do you move first or second? Are you Protoss against Zerg, or Terran against Terran? Did you spawn in the jungle, next to elephants, or in the hills next to copper? Does your starting planet have enough iron, Quinns? The rules of the game may start out asymmetrical, with a playing field tilted towards or against you.

There is one asymmetry that everyone starts out with when they first play a game - they don't know the rules of play. The process of learning a game can be simple or challenging, but it is clear that we can derive significant enjoyment simply from exploring the possible space set up by starting with simple rules and gradually introducing game play elements with greater complexity. Critically lauded puzzle games like Braid and World of Goo as well as block busters like Half Life 2 rely on this technique to propel narrative and game play forward, and games like Civilisation, Diablo and most MMORPGs dangle the carrot of newer and shinier toys to play with at almost every opportunity. The recent after adventure report of a Solium Infernum game starts out with games journalist Kieron Gillen tripping up on straight forward failing to read the fine manual, through staying awake at night writing angry letters in his head to the designer when an ambiguity in the rules interpretation puts him at a disadvantage, through to him rules lawyering with the best of them via email to try to bring down the front runner in the end game.

And even when you know what rules drive the game forward, and what ones to drive around, you may still not know what is happening in the game. Fog of war, hard to implement in board games, far easier in computer games, will shroud the map, letting you know a little of what is going on, but not enough. Incomplete information is an art lost these days to the single player game and almost entirely the province of multi-player, thanks to the rise of and game assistance tools in MMORPGs. Roguelikes are fighting a rear guard battle here, with randomly generated dungeons, permadeath and identify systems, but even some of the staunchest Angband players now advocate the elimination of identify, and full 'monster spoilers' available to all players. Which is a shame, because as poker and Demon's Souls, with its in game tip sharing, both demonstrate, incomplete information is a powerful and addictive game mechanic when done right.

Related to incomplete information, but important enough that it warrants separate discussion, is incorrect information - exemplified by the difficulty spike. Jeff Vogel argues strongly that you should never bushwhack your players; that is design a section of the game which they cannot complete at the point they reach it. I class this as incorrect information, because players will blithely waltz pasts the signs saying 'certain death here, do not enter!', the skulls, wet with fresh blood, the wall of helpful townsfolk blocking the entrance with prophecies of doom, with a notion in their head based on the difficulty of what the last challenge required, not what the next one will. The continued expenditure of millions on protecting us from the last terrorist threat, not the next one, shows us humans incapable of predicting terribly well, but very good at remembering what just happened. From a game theory point of view, we play using an optimisation strategy based on what we know, and when the strategy ends up depleting resources we had planned to keep for later, we find it difficult to continue.

Incorrect information is the most harmful when inappropriately in the user interface to the game. At some level, almost all computer games are about a user interface strategy. If we can design an AI to play the game perfectly, and simplify the UI to a single button press to start that AI, I would argue we still have a game, but the user interface to that game has been solved. At the other extreme, the complex button sequences and twitch reflexes of FPS or fighting games are a user interface hurdle that proves to be insurmountable for some potential players.

Moving beyond a discussion of simple user interface design, there are more subtle interface disparities between play at an intermediate level and advanced level, or between single player and multiplayer. I've argued elsewhere that Civilisation IV fails as a game on one level, because the elements of the game that it emphasizes in the user interface - a broad spread of technologies, boom or turtle strategies, building lots of interesting units - completely mismatch the strategies required at high level play - forest chopping, rushing strategies, the stack of death. If the optimal strategy is to have horse units built by turn 10, as you lower the difficulty level, you should be getting rushed by the AI with horse units at turn 11, then 12, and so on. Similarly, your single player game should usually be designed to make you better at multi-player, because the high level, interesting play typically only develops in multi-player some months after release. The newly introduced Dungeon Finder in World of Warcraft appears to be revolutionising the game, because it makes the tactics needed for the end game: the trinity of tank, healer, DPS - the easiest and most effective strategy for advancement at prior levels.

But what allows these differing levels of play in the first place? What makes for increasing levels of strategy, plays and counter plays? Even games with perfect information, minimised asymmetry and easily learned rules like Chess and Go, have almost unlimited scope for interesting play, because the rule set allows for a combinatorial explosion of possible game play spaces. These different from tic-tac-toe because there are too many possible combinations of moves for any one player to know the perfect strategy for any situation. Chess and Go differ from Starcraft and Streetfighter because the possible moves that have to been examined for exploits goes beyond the rock, paper, scissor choices of rush, turtle, boom or strike, block, throw.

Closely related to complexity is emergence, where behaviour that could not be predicted by the game designers, but can be discovered by the players appears in the game. This differs from complexity in the way that chess pieces don't change colour or float in the air when the board reaches a particular configuration, but it is possible to scale a building in Deus Ex or Half-Life using some well placed limpet mines. It is the provinces of speed runs and ROM programmers.

Whereas play is the province of mod designers and role players. We can always make a game more difficult, by choosing to make it more difficult, limiting ourselves artificially to a different set of standards, or changing the rules of the game and exploring the results. I'll be the nurse, you be the doctor this time - scrub as strength instead of weakness. Even for a seasoned play to winner like David Sirlin, play provides the opportunity for some well-needed research and development time. If I play as Blanka, can I find some bug or exploit against Guile, my favourite character, that an opponent in an upcoming tournament may try to use against me? What about if I blank on the day? Who should I choose instead if I don't make my crouch low punches reliably?

And finally, we can choose to extend play in the moment, for pleasure's sake. The day is dawning over Africa in Far Cry 2, so I'm going to walk to the next assassination, admiring the god rays filtering through the trees, the mist, the sniper lining up a bead on me... I know I should move on from Aeris, but I'm going to try to rescue her one last time. Yorda and I will sit here a moment, and watch the birds flock, even if it brings the darkness one step closer to taking her from me...

To summarise, games can be difficult because of:

1. Random number generators
2. Indeterminacy of outcomes due to unpredictable external forces
3. Other players
4. Complex risk vs. reward trade offs
5. Finite playing time
6. Self-limiting performance
7. Inequality
8. Asymmetry
9. Learning the rules
10. Incomplete information
11. Incorrect information
12. User interface
13. Disparity between beginning, mid and high level play
14. Complexity
15. Emergence
16. Play
17. Pleasure

So to ask how do we make a game harder, is to ask how we can vary these factors in an interesting way.

There's a whole lot here to take in, and I have skimmed over topic areas that warrant an article in themselves. This has been a diversion from the path we were taking, but an important one, and for part seven, I'll be getting back on track, and making a second attempt to talk about failure.


Popeye Doyle said...

"He misses a trick: the question shouldn't be how can we make games harder? It should be why are games difficult at all?"

Also: "why do we want games to be hard?" There's little point in making a game hard for the sake of being hard; if you can identify why someone wants to play a hard game it should be easier to tweak a game more effectively.

Andrew Doull said...

It's a lot more challenging answering the question why do we want games to be hard, because I suspect we only want them to be hard in particular ways.

Or more correctly: some people want some games to be hard in some ways, but not others...

John Doe said...

14. Complexity

This one. This is why I want games to be hard. Or rather, why I don't want them to be easy--a hard game is not necessarily complex, but an easy game is necessarily simple.

All those interesting items and actions and combinations and AI and architecture just become ignorable "flavor" if the player can waltz through the game with a basic tactic.

VRBones said...

"why do we want games to be hard?" Sounds like a great question:

- Greater sense of achievement.
Knowing that you have achieved a difficult task increases your self-worth. This can be enhanced by acknowledgement of the deed by others, or even by the game itself through highscores, etc.

- Higher level of learning.
There is a sweet spot for difficulty called "flow" that gives the greatest learning achievements. This sweet spot is not too difficult to lead to anxiety & frustration, and not too easy that it leads to boredom. Falstein's work also supports a lumpy difficulty path as long as it stays within the boundaries of the flow channel as it increases interest.

- Feeds competitive nature.
Some of us thrive under difficult circumstances, where determination and mental strength pushes out the boundary of the flow channel to achieve a steeper learning curve or sustained high performance. This can take games into the realms of competitive sport with similar endorphin-fuelled experiences.

There's also some great stuff on the fear of failing as a way to utilise difficulty in games.