Saturday 31 October 2009

The Quest for Quests - Part Four (Assumptions)

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

Quest content design is hard: which is why so many quests are either Bounties or Fed-Ex quests. In their simplest form, a Bounty is a 'go get me x of y', and a Fed-Ex is 'take a to b'. But even these simple forms raise many interesting questions and hidden assumptions - and are worth examining further.

Bounty Quests

Bounty quests traditionally involve killing n rats where n > 0, but can equally be seen as collecting n items, for similarly sized n. But rats are traditionally difficult foes in quest design, because they make the following assumptions:

i. n rats exist to be killed
ii. the rats to be killed move slower than the player and remain in accessible locations which are known to the player
iii. the player always has access to the appropriate rat-killing equipment
iv. rats in the boundaries of the agreed quest location are only killable by the player, never by anything else including natural causes, or are else recreated through spontaneous generation if they are killed by other means despite the absence of a procreative rat population due to the ongoing rat holocaust
v. by rats, the quest giver and player automatically agree that this includes or excludes rats of a particular age, gender or profession, completely distinguishable and agreed postmortem, and that the moment of gestation of a rat is definitively known
vi. the quest giver has the omniscience to know when the player has killed n rats, or the rat evidentiary material required by the quest giver only ever exists on a rat in a singular form and cannot be subdivided and is only produced due to the specific actions of the player, or there is a thriving black market in rat body parts to the exact value of the quest reward
vii. if rat evidentiary material is required, this material is perfectly preserved in perpetuity and contained within an immaculate boundary which extends beyond the player's body sufficiently to interact with this material in a defined way, or if it should decay or be lost, the rat population should replenish itself miraculously
viii. the killing of rat n, is more interesting and challenge an exercise to the player as rat n-1, which was as fulfilling as rat n-2, and so on

Assumption viii strongly suggests that there should be an increasing challenge as the player progresses through the quest. One approach may be to simply have the initial encounter be with 1 rat, the next with 2, then 3, and so on, which will guarantee that nth rat will be as interested, provide n is on the Fibonacci sequence, or the remainder included in the final encounter. Another approach may be to scale the difficulty of locating or fighting each rat higher than previously, which could be achieved by placing them randomly across a location of varying difficulty (the player will naturally migrate to the easier rats first).

As you can see, the amount of effort required to ensure that Bounty Quests cannot be failed can be considerable, and are often diametrically opposed to suspension of disbelief. If we simplify a Bounty Quest so that we just count the number of times we perform an action, instead of counting the successful outcomes of an action, we end up with a Do Something quest. The Do Something quest is an instruction to do an action so many times: such as chin up exercises in Metal Gear Solid 2, or the fight tutorials in Zelda: the Wind Waker. The Do Something quest is far easier to measure, but does not necessarily gauge skill, just determination.

If you are interested in measuring skill, you could get the same effect as a Bounty Quest but with easier to control the outcomes by using an Arena, where we place the player in a defined area, and measure their actions. Similarly, puzzles, skirmishes and mini-games can be used to measure player mastery of a set of skills without the complex boundary conditions that a Bounty Quest demands. The benefit of the Bounty Quest is that we allow the player to suspend the quest while having achieved partial completion, and potentially replenish the resources they require from elsewhere in the game world. I would argue that this benefit comes at considerable complexity and inflexibility.

Fed Ex Quests

Let's look at the set of assumptions required to perform the minimal Fed Ex quest for which the game designer has made it absolutely trivial for the player to complete:

1. The item is provided by the quest giver
2. The player can carry the item easily and safely without impact their existing abilities
3. The player cannot drop, lose, consume, sell or destroy the item in question
4. The item has no utility value to the player
5. The player cannot find the item elsewhere
6. The player can reach the requested destination
7. The destination is at a fixed point in space
8. The destination is known in advance
9. There is no time limit
10. The shortest path to the destination is clearly indicated at all times
11. The shortest path to the destination requires no expendable resources be consumed
12. When the player reaches the destination, they automatically complete the quest
13. There are no other quests which it would be useful to be closer to the destination than the start of this quest
14. The reward for completing the quest is greater than any other bonus that the player could acquire with the same time and effort elsewhere
15. The completing the quest does not change the accessibility of any other game content

Some of these assumptions are involve complex logic, such as 1, 3, 4, 5, 6 and 12, where the wrong design can result in the player not being able to complete the quest at all. For instance, if the game designer has chosen to make quest items act like other objects in the game, which results in them sharing common code (command interaction, menus, physics engines and so on), the presence and reach-ability of quest items must be guaranteed at all times.

Taking a simple example of being able to drop an quest item from the player inventory. In this instance, the player must:
a. Be able to pick the item up again
b. Not be able to drop the item in an inaccessible area
c. Not be able to prevent themselves from reaching the area they dropped the item again
d. Any other actor which picks up the item must not be able to make it inaccessible
e. Any other interaction in the game must not be able to destroy or move the item so it becomes inaccessible
f. If the location containing the item is freed from memory, it must first be saved to non-volitile in a form which it can be retrieved in subsequently
g. Data structures and code must be bug free so that the item cannot be duplicated or lost

And so on. Note that some of the above assumptions may not be achievable at all within a reasonable development time frame.

Given the complexities of assumptions which can result in a game state where a player cannot complete a quest, it is amazing the lack of effort spent on varying the assumptions for which the failure states are far simpler to control, or impossible to achieve at all. Consider assumption 2. What if there were multiple parts to quest item, of which the player could only carry some of at one time. There is the classic puzzle example of crossing a river with a fox, a hen and grain, of which only two fit in the boat at any one time, and where the the fox cannot be left with the hen on either bank, and the hen cannot be left with the grain. These sorts of simple logic problems can be woven into a quest fabric where failure states can be controlled in a much more straight forward way.

Another approach to assumption 2 would be to turn the quest into an optimisation problem. Consider an example where the whole quest item consumes multiple valuable inventory slots, but parts of the quest item can be carried in single slots. The choice is then: do I give up equipment already in these slots (which presumably gives me greater capability to reach the quest destination), or do I make multiple trips, and so take longer to complete the quest? In these types of optimisation problems, there is no failure state, only a clearly delineated series of trade offs that the player must make. Assumptions 13 and 14 similarly encourage the player to think in terms of optimal route selection.

Take care when designing these problems that the process of solving it is interesting for the player. Inventory optimisation is already an interesting decision in Angband, so the inventory slot example is appropriate for that game, but in a game where the player never reaches their inventory capacity, this mechanic would be a waste of effort .

Let's turn the notion of Fed-Ex quest on it's head for a moment. If we start by violating assumption 14, and then progressively replace unique quest items with commodities through out the other assumptions, one natural conclusion is that the counterpart to the Fed-Ex quest is a trading game, such as Elite. In this genre, fungible, easily replaced items can be purchased and sold in multiple locations, and the player focuses on the strategy of buying low and selling high. It is possible to simulate this trading game using multiple Fed Ex quests to send the player along the optimal path, but this makes the assumption there is an easily discoverable path - which may not be the case in a game which simulates market conditions by varying prices over time. You may argue that the code complexity for an interesting trading game is high, but it is no more complex than some of the challenges I highlighted in simply allowing the player to drop quest objects. The fact that Fed Ex quests are more prevalent than trading games is because most games featuring Fed Ex quests don't allow you the luxury of dropping quest items.

If no tangible item is involved in a Fed Ex quest - e.g. no item that has any impact on game play other than a UI representation, we ensure that the quest meets assumptions 1 to 5 by default, and greatly simplify the implementation of the quest code. These simpler quests usually have the option not to deliver the quest item, so that the item has nominal meaning: this is usually a narrative fig leaf covering the fact that the player is simply going to a particular destination for the quest, where not delivering the item is the equivalent of 'I don't want to advance the story yet'. In any event, whether a player is given the choice or delivering the item or not, the result of this simpler Fed Ex quest is the Go To quest, where the player is told to go to a location, and does so. Assumptions 6 -15 then fall into two categories: violating assumptions 6 - 14 are ways of keeping the journey phase of the Go To quest interesting, violating assumption 15 is a way of keeping the consequences of the Go To quest interesting. The Go To quest can be seen as the atomic counter-part to the Do Something quest I discussed above. All quests consist of these two actions (Go To and Do Something).

Keeping the journey interesting is a function of ensuring the moment to moment game play is interesting - important in a game even without quests - while keeping the consequences of completing a quest interesting is traditionally seen as a much a function of narrative as it is of game play. It is consequences I want to talk about in the next part of this series, of success, of choice and of failure.

Friday 23 October 2009

The Quest for Quests - Part Three (The Map)

You may want to start this series with part one and part two.

The traditional approach to quests in games is hand written design, followed by scripted implementation. This is not to disparage hand-written design: the map for Shadow Complex is a beautiful, complex document, which was used to prototype the game in Adobe Illustrator and proved robust enough to virtually be identical in design to the final implementation. Scripting for quests is a natural approach, because the process of completing a quest appears procedural. First you do this, then you do this, then you do this. And scripting gives you the flexibility to have a wide variety of start and end conditions for a quest: because scripting languages allow you to set up complex logic to verify that these conditions have been met.

But it is these start and end conditions that make scripting problematic because every script requires that these pre and post conditions be correctly met. I've deliberately changed to calling these pre and post conditions, because the terminology is also used in programming to define the inputs and outputs of a function. And as should be clear from Unangband, the risk of taking a programmatic design to systems introduces the potential for bugs. This raises the bar for quest design, because every quest requires programmer-level understanding of game, which limits the total pool of people who can contribute quests to your game (or requires that you check every quest yourself). So the advantage of scripting quests: allowing complex pre and post quest logic, is the same as the disadvantage of quests. Scripting paradoxically limits your ability to use it to good effect.

That is not to say that a data driven design allows you to avoid pre and post conditions, but these are implemented once, in the programming section of the code, where exceptions can be rigorously handled in a single segment of code, as opposed to required for every script. You could equally argue that you can encapsulate the complex logic for pre and post conditions for scripts in functions to the same effect. But at that point, your scripts are not going to look dissimilar from data structures - just without the advantages that a data driven design gives you.

To see those advantages, I'm going to start with the strongest example of data driven quest design that I can find. Or, more correctly, I already have. That example is Shadow Complex, inspired by the Metroid series of the games, and the quest data structure used is the map.

In the Metroid series of games, the limits to your playing space are not scripted. They are instead limited by the current powers that you have - doors require rockets or ice beams to open, lava requires upgrades to your armour, small spaces require you can transform into the Morph ball in order to fit through them. This is a straight forward lock and key design, where the keys can be re-used. What is most important about this design: it is much more robust than a scripted approach. As can be seen in Metroid speed runs, it is possible to break the individual locks, without necessarily breaking your ability to continue playing the game. In a traditional scripted approach, if you somehow bypass a section of the map and do quests out of order, the designer must have explicitly considered this in the pre and post conditions of individual scripts. With the map-as-quest, you can do sections out of order without anyone having considered the ramifications of doing so. It is still possible to end up stuck if you skip multiple power ups, and end up falling into a section of the map which requires these power ups to traverse out of. But falling is the most general case of one-way movement which otherwise only rarely occurs in map design.

And the map as quest allows us to use a mature game design tool set where we are able to prove important attributes about the map, such as connectivity, in a way that we cannot necessarily prove with scripts. We can also take advantage of procedural generation to create sophisticated maps which are more complex in one sense than any script could be.

You may feel at this point I've performed a sophisticated three-card monte to prove data driven design superior to scripting. Every game has a map, and the important thing is the scripted triggers embedded in these maps, and not the maps themselves. In a sense you are correct because it is entirely dependent on whether you have made the map interesting to traverse. In other words, is it about the quest or about the questus interruptus?

Walking to Mordor

If your map degenerates to a series of nodes, interconnected by a graph of edges, then there is no point trying to represent the map as anything more sophisticated. Puzzle Quest is refreshingly honest in this approach, and it's how I've represented the Unangband wilderness. I don't necessarily think it's the most successful map design - it is another great example of data driven quest design though - but it is much more successful than games where you get a quest, then solely use the minimap in the top right corner of the screen to get there. In simple to traverse maps, the push component of the narrative that I talked about in part two is the most important, because there is nothing sophisticated required by the pull component of the narrative. Once you know where you have to go, there is no compelling puzzle stopping you getting there.

If on the other hand, the map is non-trivial to traverse, then pull narrative can become more important. In the idealized pull narrative, you already know the destination, and there is a compelling puzzle blocking you from reaching that destination. For Far Cry 2, this is the checkpoints between you and the destination, both those marked on the map and hidden, and the area around the destination, which you have been given the opportunity to explore previously. Metroid goes one step further, and just shows you compelling puzzles, with the implicit promise of a destination, or reveal, behind the puzzle. This has the effect of both a push and pull, but without the explicit quest narrative pushing you or a destination being shown pulling you.

We should really revisit our theory of how quests and narrative interact at this point. We've found at least two examples which fall outside the bounds of the push and pull based narrative structure velocity I described in part two. The first is a single, compelling push and pull at the start of the game, which is constantly interrupted as you move towards the destination (the Lord of the Rings model). The second is no push or pull at all, but puzzles in multiple directions without destinations (the Metroid model). Both appear to be about interruptions of some kind, one built from the top down, and the other from the bottom up, which allows me to segue to my own interruption until part four.

Wednesday 21 October 2009


An attempt to recreate Angband in Blender. Video footage below:

Tuesday 20 October 2009

Review: Canabalt

Canabalt is the most adrenaline pumping, heart-stopping four dollars* you'll ever spend on the iPhone - a homage to the chase scenes in Point Break, the Matrix and Se7en with but the purity of a John Woo movie and the timing of Buster Keaton.

*Australian iPhone pricing. As per usual, the most expensive world price.

The Quest for Quests - Part Two (The Puzzle)

(You may want to start this series with part one.)

I apologise again for the authorial ellipsis of the previous part of this series. It was one of those fumble in the dark, looking for a light and likely to be eaten by a grue moments where I had to write out my thoughts and objections before I could get to the gist of what I want to say. I should be far more direct from here on in.

My chief objection to quests is how they impart narrative velocity. Narrative velocity is the reason you move forward through the game - you want to continue to see additional content because you've been excited about the rules of the game and of the world it is embedded in. Your disbelief has been suspended, you're experiencing flow, you have invested in the characters and so on. Think of it as all the good reasons to continue playing.

There are two ways to impart this velocity: you can be pushed, and you can be pulled. A push is direct: you are told you have to move forward, but you don't know the unknown you are moving towards. A push is required during the early phases of the first time you play a game, until you've figured out the rules. But it would be wrong to think of the push as the tutorial phase of the game. Many games use push the whole way through, in the form of quests and quest givers. A push can be mechanically very simple: the push in Super Mario is 'move to the right'.

A pull, however, is indirect. You know what the destination is already - 'finding all the coins' in the Super Mario example. The pull comes from the fact the game has made this destination compelling to you. This destination could be as simple as 'reaching the end of the level' - this is a pull, not a push, because the end of a level is known quantity (a 'you've reached the end of the level' screen) - it's the beginning of the next which is unknown. The pull is the payoff from getting there (loud music, bright coloured lights, operant conditioning).

Both pushes and pulls involve a promise from the designer which keeps you moving forward. The promise of a push is 'I will reveal to you something more impressive than you have already seen'. The promise of a pull is 'This thing I have shown you will be as good as you think it is'. With these promises you can see how a combination of pushes and pulls can be used to impart continuous velocity through a game.

But if the challenge of the designer is to deliver on two quite closely related promises, the challenge for the player is quite different. With a push, the player is given a to do list by the designer - as simple as 'keep moving'. With a pull, the player compiles their own to do list - again, as simple as 'get there'. The payoffs of 'Wow' (amazement at narrative resonance) for pushes and 'Yeah' (gratification of operant impulses) for pulls are psychologically quite similar, but the process of player agency is completely different: either externally applied, or internally generated.

It is this internal generation of pull based narrative velocity which is unique to games and why traditional narrative techniques are only weakly effective in motivating a player to continue playing. The two techniques are diametrically opposed in form: the novelty of push countering the repetition required of pull (See The Function of Narrative in Games: A Theory for further argument on the power of repetition in game narrative). Traditional game design requires a balance between introducing new things, and providing the payoff for each new thing introduced (Giving toys, and letting the player play with them).

And this is why, as I intimated in the previous part of this series, the puzzle is such an effective paradigm for quests in many ways. Puzzles operate mechanically as 'given this set of tools, how can I reach the payoff at the end of this known problem?' - they are almost exclusively pull based. You solve a puzzle because it's there and you have been conditioned to solve it. There is almost no narrative backstory required to push you towards it.

But the strength of a puzzle is not just about the logical deduction necessary to solve the problem - for games it is equally about the psychological rewards for using each of the tools, in reality toys, in solving the problem. That is why puzzle games like World of Goo are so successful, while adventure games have done so badly - the difference in the operant payoff between negotiating with a text parser and stretching a viscerally represented ball of malleable goo.

The quests of Far Cry 2 to me feel so successful because they are essentially the same problem repeated with new combinations of shiny, operantly rewarding toys: the weaponry you unlock as you progress. Far Cry 2 is biased far more to pull than pushed based narrative velocity, whereas most traditional quests are push based.

So for the rest of this series, I'll be looking at the quest mechanics I touched on in part one: quest content, branching dialog, main quests vs side quests, procedurally narrative structure, open world vs hub and spoke, and trying to rewrite them as pull, puzzle based mechanics instead of push, narrative mechanics, and see if the results are more compelling. But first I'll touch on an issue dear to me, script vs data driven design, and give an overview of how a data driven design can solve some of the problems of the traditional scripted quest structure. For that, you'll have to travel far from here, to part three of this series.

Sunday 18 October 2009

The Quest for Quests - Part One (The Cast)

I've played a lot of quests recently: Puzzle Quest, Vampire: the Masquerade, The Chronicles of Riddick: Escape from Butcher Bay, Far Cry 2, Cthulhu: Dark Corners of the Earth, S.T.A.L.KE.R.: Clear Sky and the demo for Armed Assault 2, and stumbling out of these games, it is my intuition that the concept of quests in games is deeply troubled one.

But first, an apology. What follows is a rambling forage through the inventory of games I've picked up recently, attempting to look for the right item to use. You'll see plenty of misapplied (ad)verbs from a list which appears at the outset only partly appropriate (only two RPGs?), and attempts to force a somewhat inappropriate syntax to parse. It should get a lot easier to read in part two and beyond when I check the spoilers for the right answer, and come back and solve some of the puzzles I skip over here. This is not a deliberate technique at obfusticating the answers - it's simply bad design and underwhelming and tangled prose.

It doesn't help that in many of these games, quests themselves are broken and bizarre. I've discussed elsewhere the fascist fantasy politics of Puzzle Quest, and a stillborn review I'm writing of Cthulhu staggers around in a semblance of life stalled by a character not recognising a quest item I hold. Equally, I blame ready access to, a refuge from broken user interfaces and game bugs, for spoiling the meat of some of the story delivered in quest form. But to my refined roguelike palate, it's not the stories which are problematic, but the way quests are used to shape and propel the game play.

Strangely, it is the most repetitive of these games, Far Cry 2, which has felt the most quest friendly, with the similarly story-lite free form quests of S.T.A.L.K.E.R. and Armed Assault 2 providing an incentive to play through the bugs (to a point: in S.T.A.L.K.E.R.: Clear Sky, I gave up at the buggy warehouse in Garbage, reproduced with both the old bugs and additional bizarreness from Shadow of Chernobyl). All three games feature an attempt at procedural narrative, clumsily told. Armed Assault 2 builds clipped dialog from context specific information ('Machine. gunner. at. 300 metres. ahead'.) - whereas the rapid fire dialog of Far Cry 2 is delivered by different characters playing similar roles, in an ostensibly dynamic, but in reality quite symmetrical structured narrative. S.T.A.L.K.E.R (either version) plays best when the Russian jokes remain untranslated.

Contrast these failures in story telling with the pitch perfect notes played by the certain characters in Vampire: the Masquerade - Troika making the Source engine sing to be the only game to make me feel like I was cheating on my wife by playing it; and the wonderful opening scenes and smart decision in Cthulhu to have characters turn the uncanny valley into the Atlantic trench (A line I'll reuse in the review should I finish it). In Chronicles, Riddick's bad guy persona voiced with journeyman effectiveness by vin Diesel conveys a compelling series of events, and the broken inmates of Butcher Bay are no less believable than the damaged stalkers of Chernobyl.

As an aside, Source remains the only engine with verisimilitude to deliver something which successfully approximates a human being in game. Every other 3d game engine seems to me to leave their 'people' with a Cthulhoid plasticity - although Far Cry 2 is not far from the mark. Only military shooter Armed Assault 2 and Puzzle Quest from the above list have characters with close to what could be called 'normal' mental states, with every other game I've listed featuring a range of inhumanity and mental pathology - let's just use that as an excuse for their facial tics, stiff movements and unnatural sheen.

So the story component of the traditional narrative driven games is not the source of my quest frustration. But then what is?

It's not the quest content. All the games feature plenty of fetch and kill this quests, as well as light puzzle action and lever pulling. I will admit to fixed puzzles being my least favourite part of quests, but nothing which a quick perusal of can't fix. I'm not especially interested in solving word games, or shuffling through inventory screens looking for the right item. Luckily, the world agrees with me that the death of the genre which featured these the most - adventure games - was more than timely.

It's not branching dialog. The games which feature it wholesale, Puzzle Quest and Vampire, avoid the typical pitfalls of branching dialog by allowing real and interesting choices respectively to make most of the conversation decision points both natural and not obvious. In Puzzle Quest, you are presented at points choices which are more freedom vs. obedience than good vs. evil, and the seemingly consequence free outcomes allow you to shape the internal morality of your character to suit the tale at hand. In Vampire, negotiating the moral morass of social interaction is as important a choice in character development as your combat skills, and where maximising your seduction skill seems a natural decision. The only oddity is presenting choices where your character sets out to deliberately antagonise without clear motivation why you would do so: I'll have to sample those on another play through.

It's not the tension between the main quest and side quests. The RPG trope of rushing off to save the world, but then stopping to rescue a cat stuck up a tree, then solve the marital problems of two strangers, then choose the right colour curtains for a neighbour, etc. didn't feature in any of the above games - although the primary driver to find Lord Bane mysteriously disappears towards the end of the first chapter of Puzzle Quest. Vampire: the Masquerade has side quests which are more compelling than the main quest simply because they feel more important (stopping a plague, finding a serial killer, solving family disputes) than the somewhat dusty and archly political orders of a vampire prince.

It's not the procedural narrative structure. The narrative in Far Cry 2 is only procedural in form, not function. You experience a narrative where your buddies are selected dynamically, telling their individual story vignettes in different orders as they send you on the same stock missions, while the larger hunt for the Jackal has only a fixed number of possible outcomes: all controlled by decisions you make. I can't tell you if this variation in form changes the emotional arc of the narrative because I haven't finished the game. But it's the mechanical moment to moment interactions which give Far Cry 2 it's emotional punch - watching a wounded enemy lie writhing on the ground in pain, waiting for his comrades to try to collect him, the brief pause before the clinical assassination of a target. Far Cry 2 is procedural because these moments are procedural: fire, weather, the AI, guns jamming and malaria flaring.

It might be the way the worlds are structured. The 'good' quest games I highlighted are all open worlds, where you are free, essentially, to explore the entire terrain of the game with minimal gating of content; while the 'bad' quest games have a hub and spoke design, with most of the content being unlocked by completing previous quests. This is a continuum - S.T.A.L.K.E.R.: Clear Sky has one significant gate which requires the main quest arc be partly completed, Far Cry 2 a couple, but the locked doors of Vampire: the Masquerade and Riddick are a frequent barrier, and Puzzle Quest refuses to open it's fantasy network up until you progress. Cthulhu is a cleverly disguised corridor shooter for the most part: with a few limited hubs that have only a little more scope than Half-Life.

Paradoxically, I believe the fixed puzzles I don't like are the solution to the dilemma I find myself in. Having said that, I've still not answered why I feel uneasy about quests, and to get that answer, I'll have to ask you to wait here for part two.

The Rules of Game Design:The Golden rule

In Towards a Moral Code for Game Designers, I sketched out some brief rules of game design, and made the promise I'll expand on them - which I'm doing in this series. I'll keep the rules punchy, short and easy to remember - the ideas are usually not mine, and I'll attempt to credit them appropriately. Feel free to respond in the comments, buttress my arguments or undermine them. Remember, I'm trying to be a force for good to Ernest Adams' evil.

'The Golden rule:' The most valuable resource in your game is the player's time and attention.

Named after: Gold, particularly gold in World of Warcraft.

Why is it important: It should be self explanatory that the player's attention is the most important resource in the game - far more important that anything else that the developers can deliver - but many people design and play games as if this as not the case. Consider the following:

1. The majority of people never reach the end of a video game. Their attention gets exhausted before the game content does.
2. The best players in Starcraft are often measured by their actions per minute - the total number of actions that they can execute effectively in parallel. In the best players, this is over 200 per minute or more than 3 meaningful clicks per second. This resource is usually more strategically important than perfecting a build queue, or the population cap, or the decision to rush, boom or turtle.
3. The number one killer of characters in Angband is tiredness, followed by boredom. The best strategy to winning the game is to attempt this in as short a time as possible - a counter intuitive approach in a game which is effectively infinitely grindable.
4. No matter how you design a game, you cannot prevent a player from choosing not to play it. In fact, trading of characters and in-game currency in MMORPGs make it possible to put economic value on skipping playing parts of the game. Some people are clearly willing to spend much more than they did on the purchase of the game, to avoid playing it.

Consequences: Design your game so that playing better is more important than playing longer. Allow good players to short cut the process of playing by opening up non-linear progression options, and skipping over any unnecessary content. Leave it up to the player to decide what content is unnecessary, rather than lock them into cut scenes and out of in-game content.

Further reading:
Half-Life 2 completion stats: Episode 1, Episode 2.
Starcraft explanation of actions per minute. UC Berkley Starcraft class, week 1.

Further viewing:

Speed runs for Metroid, Half-Life, Morrowind, Angband.

Any other recommendations?

Saturday 17 October 2009

Poll results for 'What is your Brain Hex class'; new poll

Thanks for those of you who voted in the Brain Hex poll. The 71 of you who did so help contribute to over the 7,500 responses they had received as of the 30th of September. The results for readers of this blog were:

7 (9%)
17 (23%)
6 (8%)
37 (52%)
22 (30%)
11 (15%)
9 (12%)

Contrasting this with the responses in the overall survey population:

Conqueror: 2073
Mastermind: 1597
Seeker: 1475
Achiever: 809
Socialiser: 744
Survivor: 592
Daredevil: 464

suggests that the self selecting sample at this site is paradoxically less competitive than the self-selecting population that the survey has received - with a much lower proportion of Conqueror types. [Edit: Xopods points out that I should have done some maths. We have a normal distribution of Conquerors. It's Masterminds which appears to be over represented. And that may simply be an artefact of the fact I let you choose multiple types in the poll]. Other than that, the proportions are similar to the general population. I don't know how much more you can read into samples of this size which have not been randomly selected, but that single statistic is food for thought.

The next poll is some administrative catch up. I've sporadically asked which of the articles I write on the blog you find most useful. Unfortunately, I've not done this since September last year so there's a few articles here for you to reminisce over.

You may notice as well that the overall blog output year to date is half of what I've achieved in previous years. That's simply been an impact of the time constraints of both trying to get a major release of Unangband out the door, and now having a longer commute than previously (The two are unrelated).

I would have mentioned this earlier, but bloggers writing blog posts apologizing for the lack of blog updates is a little.. hmm... I guess if Warren Ellis does it, it's okay. (BTW: Friend him on Facebook if you haven't done already).

The Canon of Procedural Games

While updating the PCG wiki today, I've come across the difficult notion of canon in procedural content generated games. I've flirted with this concept before, by defining games which are prototypically procedural, but in general I've tried to be inclusive rather than exclusive when it comes to including games in the PCG wiki.

I've hit a stumbling point writing up an article on adaptive difficulty - always a controversial point in games. I'll quote the whole article to saving you having to go to the original link:

Adaptive difficulty is the process of adjusting the game in reaction to the player. By spawning new enemies or powering up existing enemies if the player is progressing quickly through the game, or by decreasing the frequency and/or difficulty of existing enemies if the player appears to be having problems progressing, adaptive difficulty techniques attempt to create the 'optimal' game experience.

Classically, adaptive difficulty has been seen as a hard problem, requiring a level of artificial intelligence in the game to attempt to model the player to attempt to determine if they are finding the game easy or difficult.

However simpler RPG style mechanisms can also be seen as adaptive difficulty techniques. Allowing the player to level up by playing through additional easier content can ensure the player is able to grind their way through parts of the game in order to decrease the difficulty of sections of the game where the difficulty level increases. Paradoxically, adaptive difficulty techniques which increase the difficulty of the game by scaling up enemy strength have been fiercely resisted by RPG players, as can be seen by the negative reactions to the difficulty scaling in Oblivion.

Adaptive difficulty is not usually seen as a procedural content generation technique, but it has most of the features of such techniques. It could be seen as decreasing a game's randomness instead of increasing it which would make games which feature it without other PCG features to fall outside the 'canon' of PCG games.

Should I include games which have adaptive difficulty in the PCG wiki? There are plenty of examples of games which have adaptive difficulty and are procedural (Oblivion I've already mentioned, Left4Dead) but there are plenty of games which are not (SiN: Episodes). And I don't want to include every RPG, based on the argument I've made above.

The real question is not whether I should include these games, and the answer to that is probably not, but why? What good reason can I give to not include SiN: Episodes, for instance, as a procedural game?

Again, the randomness argument is the most plausible, but it is not completely convincing.

Convince me.

Tuesday 6 October 2009

Duct Tape Programmers

I didn't link to this at the time, but it popped up again in an article on Slashdot, and I think it's worth reading: Joel Spolsky discusses how to get programming done.

Saturday 3 October 2009

Iso-Angband is back

Hajo, how we missed you.

I hope the Angband community is more welcoming this time around.

Read more here and here.

Friday 2 October 2009

Review: Puzzle Quest: Challenge of the Warlords: Chapters 1 & 2 for the iPhone

It was the word cull that finally made me balk. The sallow humans who needed me to thin the numbers of goblins in the nearby forest: sunken eyed and dark-lipped as if inbred with their ghoulish neighbours to the south. They wanted the goblins dead, but phrased the process in terms of clinical population management. The same goblin race that I was trading with across the mountains were here no more than cattle or game overrunning it's permitted habitat.

Sure, I killed a few. But the word echoed in my head as I butchered the first three: burnt flesh and blood on my hands (and boots, and hair and everywhere else that the clubbing blows I inflicted sprayed their arterial viscera). And I stopped. Not because of any moral code but mere distaste for the work. (and the decision fell nowhere in the deliberate morality I had chosen: freeing the daughter from an arranged marriage, who still travelled with me, never stealing or cheating - but brutal, brutal bloody murder...)

The storm giants were next. Their chieftain I had somehow chosen to spare, turned and asked for the head of one of the thieves who had guided me to his mountain habitat. I agreed to his request, but no thief will feel my stormblade. I did hunt pegasus eggs in the mountains for them, but the nobility of the mother in her throes of death turned me from hunting any more - even though taming one of these beautiful creatures as a steed would have greatly benefitted me.

The collection of heads as bounties is no trivial matter. The tendons in the neck are tough, and need to be cut, or more often hacked through with a sharpened blade post-mortem, with particular attention paid to popping the vertebrae apart between segments. You should move the body between three and four a.m. to minimise the risk of being noticed, cut into four or five approximately equally sized pieces in the bath or shower and double-bagged in plastic garbage bags. Store the parts in the freezer to avoid the stench of decomposing flesh until you're ready to proceed, and change the sumps that could trap human matter to minimize the forensic footprint.

Every fantasy world feels like a Pangaea through which you wade waist deep in a bloody river of your own design. Humanity has always acted this way: from some primeval fear when we stared deep into the eyes of Neanderthal man and saw something truly alien staring back at us. Our so-called civilisation still gives lip-service to the idea of universal intelligence, saving dolphins to sack the rest of the sea. But as soon as we can stuff it in our mouths and swallow, we'll 'harvest' it until the wellsprings run dry. Witness the fiasco of governments stumbling to avoid declaring the bluefin tuna an endangered species. But governments have always stumbled: perpetual extensions of our partitioned, hypocritical by design brains.

Humanity as a xenophobic monoculture - eusocial and spreading through the stars in a genocidal crusade. That's why science fiction to me rings truer than those tired recreations of the hero myth.

Distract me with the shiny gems of achievement, and let me not think.