Time completion of games

[RedCloud]

After purchasing your game, one of the last things you want to happen is to finish it on the first day you bought it. It would seem in cases to me, that the only really lengthy games out there are RPGs. They could take anything from 30 - 70 hours to complete these days. And after completion in games, is there nothing else you can do except start again?

Are developers focusing on lengthy game play, or graphics to the best of their potential in a short time frame? What really keeps a player interested in the game? What added features should their be included to keep players captured?

[zoombapup]

Procedurally/randomly generated content

User definable content

There is NOTHING wrong with not requiring 70 hours to "complete". In fact, a lot of casual games offer gameplay that NEVER completes. There is no end. You simply play the game, like a hand of cards.

Randomness (like shuffling a deck) can add a huge order of complexity to the variations a game can have. I would highly recommend it as a starting point for any game which wants to involve players for a long time.

That doesnt mean huge completion times or long play sessions. It just means that if you want to make a game that is replayable, then make good rules and procedurally generate the game-state initially.

[Anthony Flack]

I like games that tell me how many hours I've been playing them. It's a recent trend that I'm definitely in favour of. And it tells me that most retail games that I buy are doing well to be played for more than 8-10 hours in total. Even games I thought I was playing quite a lot often haven't clocked up that much time yet.

I probably wouldn't even attempt (to play, let alone to write) a game that required 70 hours to complete. Somewhere around 15 hours is just fine, providing there are a few replay incentives and extras for people that really, really like it.

More than that, and I just find it unwelcome stretching. The game is probably getting you to do lots of repetitive chores to last that long. I hate threadbare games and I'd much prefer a short, action-packed game than the same amount of content spread four times as thin.

It all varies from game to game of course, but just filling up a couple of hours of "straight-through" gameplay in a way that is polished, varied and entertaining is usually more than enough to keep a developer occupied for a year or two.

[Ste Pickford]

Our game keeps track of how long each player has played, and we're offering a guarantee of at least 20 hours worth of gameplay for a $20 subscription (you buy credits to play Naked War, so you gradually 'consume' your $20 worth by starting new games). We're pretty confident that 20 hours is a safe bet, but what's surprised us is how much longer than that some players play for.

We only have a small number of beta subscribers at the moment, but a pretty high percentage of these players have their total playtime measured in days rather than hours, and that's in just about 40 or 50 days since the beta programme launched.

Its a lot of fun keeping track of how long people play, and I suspect it will also prove to be very valuable information in time.

[LilGames]

Depends on the kind of game. Alot of people were miffed about the length of Half Life 2: Episode 1. FPS games are typically played for long sittings, since the story engages the player and makes it difficult to stop and say to yourself "ok Ive had enough for now". So what happened is some people finished it the same day they bought it, and then of course felt jipped.

[Christian]

I find it unbelivable that retail games that where on the making for 5 years can be completed within 2 weeks, its ridiculous. This games are driven by content, that means that is the desire to see the content that keeps you playing (more or less). I prefer making games like soccer, or tennis, or a card game, they are "play" driven, the desire to play it is what makes you play it, not finish it.

Procedurally generated stuff = the best thing you can do for your players.

[Anthony Flack]

And a movie that was 5+ years in the making can be completed in an hour and a half. It's not ridiculous at all, it's just an extremely labour-intensive form of entertainment.

There's nothing wrong with content-driven games. Play-driven games are good too, and Ste's game is a good example of how much time can be spent playing this sort of game.

And of course, content-driven games should be play-driven as well, otherwise they're just content-grinders, and rubbish games.

But procedurally-generated content? I'm yet to see an example of it that wasn't dull. Procedural content generally isn't strong enough to drive a content-driven game.

[lapskaus]

I like my games short, really. It's rare that i like a game so much that i'll play it for more than 10 hours, if i'll play it that much. If anyone started making games that you could complete in an hour or two that cost $5 or $10, i'd buy lots of them. Maybe i'd even play them twice! They would be much cheaper to make too. Halflife and Sin are going in the right direction, but they're still a bit long IMO. I hope to start doing something like this when i complete my current project.

[Davaris]

>If anyone started making games that you could complete in an hour or two that cost $5 or $10, i'd buy lots of them.

Hmmm... I have an idea for a short but very hard RPG with patented "no save" technology.

[zoombapup]

Anthony: the entire game doesnt have to be procedurally generated, but parts of it being so adds replay value. Worms had a procedural landscape and could randomly generate weapons settings and team names and pretty much everything else.

The major part of it, was to show the settings so that people could click through a whole bunch of procedural maps and see which ones might look interesting. We had a central "current map" and I think it was 6 "picker" maps which were deviations on the current map. Kind of like Kai's PowerTool Texture generator if youve ever seen that.

Anyway, I think for indie games, its HUGELY important to be able to reuse the little content you have in meaningful ways. Procedural maps are a great way of doing that.

[Anthony Flack]

What I liked about worms' system was that it would seed the maps from a word you gave it; so you could always recall a map if you wanted it.

However, that's an example of a play-driven multiplayer game where the level design really isn't that important. A single-player Mario-style game with procedurally-generated levels would be terrible! Even if it worked, playing it would have a depressing futility. I don't think it's fun to beat a procedurally-generated level. But it may be fun to beat another player in a procedurally-generated level.

[Hamumu]

No, roguelikes with their randomness are endlessly fun, and they don't have people to beat! It just depends on the emphasis of the game. A random Super Mario level is probably not that great, because your emphasis is on interacting with the obstacles (although Icy Tower is an example of this, and it is quite a good game, so maybe it's not such a bad thing!). A random dungeon in a dungeony game isn't really about the terrain, it's about leveling up, finding random gear, and getting to those stairs (which would be a lame task on repeat plays if it weren't random). In fact, a lack of procedural content could be terrible in such a game. Imagine an RPG that didn't randomly generate equipment. Boring!! You get things like the very first console RPGs where the only upgrades were buying slightly better gear in each new town. The gamble for gear is a large part of the fun.

In my current game, I'm hand-making the levels, but for most of them, I'm just kind of throwing down terrain to give you a place to fight monsters in. It's not the terrain that matters (beyond a general "you're in a forest now, and a river is north" sense), it's fighting the randomly appearing enemies. The game would be almost as fun if it just took place in a series of empty rooms with monsters teleporting in (Crimsonland/Smash TV). The entertainment lies in the conflict, and in the choices you make with gear and leveling up.

I don't know what point that makes, but it relates to the topic, so I will post it now.

[Christian]

Anyway, I think for indie games, its HUGELY important to be able to reuse the little content you have in meaningful ways. Procedural maps are a great way of doing that.

I totally agree with that, thats what i tried to say. Since we are indies we cant compete with such great amount of content that retail titles have, because they are more like movies than games and i personally dont like that, its simply boring, games should me made to play not to watch, but i recognize that when i was younger and i had all the time of my life, i ejoyed playing games like Monkey Island, bioforge, full throttle, story/content driven games, but that has changed, i want to play and have instant satisfaction.

[Anthony Flack]

That... well, er, that's fine, but it doesn't really make a lot of sense in the context of the discussion. The majority of games are content-driven. Even breakout-style games are generally content-driven, and a breakout-style game with procedurally-generated levels would be pretty dull.

Although playing Psychonauts recently, I'm reminded that most stories in games fail because almost nobody does them well, and when they are done well they can make an otherwise-average game extremely compelling. I think that on a smaller scale, DROD demonstrates this as well (DROD would also be pretty dull if procedurally generated).

But a bit of randomised content is all good fun that helps with replayability. Random items in RPGs work really well; it's like a lucky dip. But each item is a self-contained, designed object; you don't really need much internal structure there. But I found the randomised dungeons in Diablo pretty characterless.

Rogue may be the great exception here; I guess that the simple tech behind it means that there can be plenty of variety, and your mind fills in the gaps in the story. Your own imagination in place of the designers', as it were.

[JPickford]

I love the idea of generated, procedural content - It's long been my ambition to create a game around this idea. But, it's very difficult to do well and is more likely than not to lead to dull content.

I long ago banned myself (and anyone working for me when I ran a dev team) from using random numbers to make gameplay decisions. Random numbers can be incredibly useful but they are also a terrible crutch for programmers to use. It's far better to think of a logical reason for something to happen than to stick random numbers in all over the place. Naked War only uses random numbers in a couple of places ; cosmetic stuff (randomised trees\grass etc.) and randomly deciding which 'end' of a map each team starts on. Other than that there isn't a single random number in the game. Even the contents of the pick-ups are entirely predictable and a smart player could theoretically work them out before opening the crates; although I'm not sure anyone has noticed yet.

It's like fireworks at the end of a level in Super Mario Bros. When you first play it seems random, at that seems fine. But when you discover its related to the last digit of the clock, you've a new, fun little wrinkle in the gameplay.

[Anthony Flack]

Hmm, that's very interesting. This is exactly what I've been pondering in the last month or so. My own experience has been quite different though.

Originally, in order to encourage score-attack replaying of levels, I thought it was important to make everything non-random and reproduceable. I always had randomly-occurring score objects in the game (I think it's fun, and I was sure to include plenty - more than a hundred - so that new surprises will be popping up even after you've been playing for many hours), but most of them are worth similar points values, and I had seperated high-scoring objects from low ones in the randomiser. And it shows the extent of my non-random bias that I was intending to implement a score-count for these objects that ensured the exact same amount of points was always available in any given level. This is the sort of thing I was worrying about. I also included lots of secrets you could learn to trigger, to earn more large bonuses.

But then I found, when playing the game myself, that most of these score-increasing tactics I wouldn't actually bother with. A lot of these repeatable exploits, while perhaps being fun to discover, became a chore to actually perform repeatedly. I'd rather not do them, and I think it sends the wrong message if a player is rewarded for playing the game in a less-fun way.

So I have decided to tone them right down; make them worth less points. I also found that by including a large random element, it actually made score-attack more fun. I would play the levels casually, and if something really lucky happened, it was a chance to try every trick and get every bonus multiplier to get the highest score possible - knowing that it was a rare shot at increasing my all-time high score.

[JPickford]

I think it's better to think of an obscure trigger for these event rather than making them random.

Shoot 3 aliens in a row from the left and...

The 10th consecutive hit without missing is a ONESHOTKILL!

If the player lets three health objects expire (doesn't collect them) then put a mega health object there.

Go for 5 seconds without jumping and...


etc.


It all appears random to the casual player but it rewards the dedicated player.


But who am I to tell you what's fun? This is just my take on it.

I think my main point is that I love random stuff but it's really, really easy to use it badly. I find a no-random rule very useful when developing because it forces me to think of a reason for everything. And of course any rule can be broken.

[Anthony Flack]

Yeah, that kind of thinking was exactly what I was going for, originally. I'd been playing a lot of Radiant Silvergun and the amount of exploits in there for a skilled player to use is unreal.

But then I found, as a person who naturally knows all the exploits in my own game, that actually using them was less fun than just playing slap-dash, more often than not. If actually pulling off the exploit is an extra challenge, and more importantly an enjoyable extra challenge, then it works. But I'd gotten myself into a situation where pulling off all the special tricks was more of a grind. And I thought, well, if other players start to use all these exploits, they might find the game less fun, too. But they'll be the ones who win.

And then I found that random stuff made it all much more exciting for me, who of course has played the game far too much already.

[Maupin]

I have to say that I love randomly generated content in games - especially levels/maps, etc. For example, as in Angband or Diablo 2. I don't think games go far enough in this regard.

It obviously depends upon the game type, but for me a perfect game would have the normal hard-coded version, and then an infinite number of random variations. I used to have a lot of fun playing through computer-generated adventures in Adventure Construction Set.

And even when content is (or must be) hard-coded, it would be nice if npc movements, etc. differed between playings.

[JPickford]

And even when content is (or must be) hard-coded, it would be nice if npc movements, etc. differed between playings.

There are plenty of ways to do that without random numbers. If the NPCs intelligently respond to what you do (and in such a way that your behaviour is affected) then you'll get pretty much infinite variety without ever witnessing pointless or seemingly irrational behaviour.

We did a lot of work in this area on a game called Future Tactics. The AI worked in a very natual manner - only acting on information it could realistically have (the NPC's didn't magically know where your units were for example). You could play any level of that game a thousand times without seeing the same NPC behaviour. Sure, if you did *exactly* the same thing then so would they but in a game where you have direct control of the characters that's almost impossible.

[Christian]

Procedurally generated content is not the same that random generated content... and it doesnt need to be dull either, it all depends in how you design that procedure, you can make it so it makes dull levels for example, or you can design it so it creates good and fun levels, it all depends on your investigation and analisis on how a level would work well.

Mario and any other game that doesnt use procedurally generated content would not work well with procedurally generated content since they were not designed to work that way, i agree, but we can design games that work perfectly well with procedurally generated content, it all depends on how we design those games.
These are games that mainly work with the challenge that the level provides (as said by other person before), so, having that in mind you could design different levels that provide many kinds of challenge, or, you could not focus on the levels but instead focus on the interactivity, the AI, in those kinds of games the level is not of much importance, its like designing soccer, the "level" never changes, but the players in it allways do, thats why its so interesting i think.

[JPickford]

Yeah. I'm convinced it's possible to make a great game with entirely generated content. It's just not very easy and there are a lot of pitfalls along the way which can lead to very dull gameplay. Making this game remains an ambition of mine - one I've held since playing Elite back in 85(ish).

[Anthony Flack]

These are games that mainly work with the challenge that the level provides (as said by other person before), so, having that in mind you could design different levels that provide many kinds of challenge, or, you could not focus on the levels but instead focus on the interactivity, the AI...

Yeah, that's what I was getting at earlier. "Content", in some games, is just a freeform bed for the action to play out on. In othe games, it very strictly defines what goes on. The former is obviously more amenable to pregenerated content.

But beyond that, any game where the primary mechanic involves defeating this content is likely to be diminished when you lose the human touch. Which is why you wouldn't want to pre-generate breakout levels, even though you could. Human-made levels will always be more interesting; the advantage to procedurally-generated levels is that you get an infinite number of them (or, alternatively, a much larger number of them than you could realistically hope to make). So I guess it depends which is more important. Some games rely on the element of surprise and others depend on a degree of familiarity. But I wouldn't resort to procedural generation unless I thought having infinite variety was really vital to the design.

You could play any level of that game a thousand times without seeing the same NPC behaviour. Sure, if you did *exactly* the same thing then so would they but in a game where you have direct control of the characters that's almost impossible.

I'm wondering what the philosopy behind this was, because in practical terms it doesn't seem that different than using randoms when the NPC's choice of behaviour isn't clear.

Is it to prevent statistical abberations? Or just in order to enforce a general principle that it's better to have a reason for every action? Also, how do you deal with things like NPCs making (deliberate, so as not to be perfect) errors? I guess you could have some kind of algorithm for that, but in this case it doesn't seem that different from using randoms, which are also algorithmically generated of course.

[Ste Pickford]

One trick with procedurally generated content is to hide the fact from the player.

I agree with Anthony that Diablo felt pretty dull and empty, despite a perfectly good game engine. The fact that the dungeons were generated wasn't hidden at all, and once you know this a voice inside your head keeps reminding you what an utter waste of time playing through the random data actually is.

Look at Elite though. That had a procedurally generated universe, but it didn't generate a different universe every time you played, it generated the same one from the same seed, and started you in the same place. Everyone started with the same planets, with the same trading conditions, and worked their way outward. They could talk about particular planets with their friends, and share their experience. You belived that the elite universe was a real place in a way that you wouldn't have done if you could keep hitting space to generate a new random universe when you started the game. Yet it probably only took a 1 byte seed for the universe generation code to make that universe for every copy of the game.

[Anthony Flack]

Yes, as soon as John mentioned Elite it kind of made me think... instead of procedural content as constantly shifting sand; procedural content as a way of creating a fixed, but very large, space.

You'd have to shuffle it from a very large deck of cards to make it interesting, though. And it would probably still be best if you added little dabs of human design throughout it. Hm, I guess some sort of massively multiplayer game would allow players to go out and terraform the landscape into more interesting shapes; but now I'm talking about Spore, and passing the content generation onto the player, rather than onto the computer.

[JPickford]

I'm wondering what the philosopy behind this was, because in practical terms it doesn't seem that different than using randoms when the NPC's choice of behaviour isn't clear.

Is it to prevent statistical abberations? Or just in order to enforce a general principle that it's better to have a reason for every action?

Yeah, I'd say the latter. The game in question was a strategy\tactics game so what the NPC's did couldn't (IMO) be random. Perhaps their behaviour in a single turn wouldn't look too different to random behaviour but over the course of a game it would be quite apparent that they were behaving fairly intelligently and had some sort of plan of attack.

Also, how do you deal with things like NPCs making (deliberate, so as not to be perfect) errors? I guess you could have some kind of algorithm for that, but in this case it doesn't seem that different from using randoms, which are also algorithmically generated of course

Well the original (flawed) idea was to give the enemy perfect knowledge of the environment and location of player team. From this they could create 'perfect' attack strategies (well as perfect as my algorithms). I could then perhaps randomly add some error to that. I never got as far as adding the random error because the 'perfect' approach lead to incredibly clinical and unnatural feeling gameplay.

Eventually we arrived at the idea of limiting the NPC's information in a natural manner. An NPC would know of an enemy's existance only if it had seen it, like wise the enemys location. So the NPC's didn't even know the player controlled characters existed at the start of the level. We also had the NPC's exchange this information when they could see each other - this was the magical element which really brought the AI to life. Pop your head out from behing a rock and there would be a flurry of NPC activity as one spotted you, told another, who told another etc.

With this type of system we found we didn't need to artificially introduce errors. You could 'trick' the NPC's by showing yourself and then running around behind them (under cover) while they were still thinking you were close to where you were last seen. This sort of fallability feels really natural and it's massively 'dynamic' in that you rarely saw repeated behaviour.

[Anthony Flack]

That sounds like great fun, and it makes perfect sense for a strategy game. I was considering coding something that had enemies alert each other and I may still do that in a simple form. My little aliens are either attacking or wandering, and they only switch to attacking when they see you. But they're very stupid, and the system is super-primitive and needs a bit more refining - and they're currently not aware of each other at all. Or noises. But you can at least sneak up behind them and stand there, waiting to for it to notice. It gets a surprise when it does.

But being an action game rather than a strategy one, I'm thinking about things like having bullets that fire with a 5% margin of error on either side of their intended target, for example. Or the way that the aliens in my game sometimes accidentally shoot each other by mistake (instead of the bullet passing them by in the 2d plane). The chances of this happening increase as the bullet gets further from where it was fired from, but it is still a random event.

[JPickford]

That sounds like great fun, and it makes perfect sense for a strategy game. I was considering coding something that had enemies alert each other and I may still do that in a simple form. My little aliens are either attacking or wandering, and they only switch to attacking when they see you. But they're very stupid, and the system is super-primitive and needs a bit more refining - and they're currently not aware of each other at all. Or noises. But you can at least sneak up behind them and stand there, waiting to for it to notice. It gets a surprise when it does.

But being an action game rather than a strategy one, I'm thinking about things like having bullets that fire with a 5% margin of error on either side of their intended target, for example. Or the way that the aliens in my game sometimes accidentally shoot each other by mistake (instead of the bullet passing them by in the 2d plane). The chances of this happening increase as the bullet gets further from where it was fired from, but it is still a random event.

My approach would be to create a simple AI model of how they should behave which is inherently fallible.

One thing I did in FT which wasn't really used in the final game was a 'patrol' mode. Basically each enemy had a map of the environmen in it's 'brain' with each cell containing a timestamp of when he last saw it. The enemy would always move toward the location with the oldest timestamp. If presented with a corridor and a number of rooms he would walk into one room, 'see' all the locations and leave. If there was part of the room not visible from the door he would walk in just far enough to see that spot then retreat (his new oldest location would be elsewhere). When two enemies could see each other they would 'compare notes' basically merge databases. This would result in cooperation where one enemy would emerge from a room and effectively tell his colleague that it was clear - thus they didn't waste time both searching the same room. Emergent cooperation among NPC's.

I can see a simple version of this working for a platformer. Perhaps rather then vision the NPC's could head to the location they least recently stood on. But by refreshing each others data when they can see each other they would patrol a level in a fairly natural but not repetitive manner.

The shooting each other stuff would happen when they attempt to shoot the player character and he avoids the shot. Or do you mean they would actually mistake each other for the player character sometimes?

I'm not really sure why it's desirable to have a margin of error with these things. Isn't it more satisfying to defeat\evade attack by outwitting the enemy?

[soniCron]

We did a lot of work in this area on a game called Future Tactics.

So you're responsible for my losing some 60 hours one weekend? Great piece of work -- I show it off to all my friends!

Back on topic: An excellent example of weighted random rewards can be found in the Zelda series. The items the enemies drop are random but more often than not reflect what you need. As such, killing monsters in Zelda is almost never boring or tedious, where killing a monster in Diablo for yet another 10 experience points can feel somewhat pointless and certainly not as satisfying.

I don't think it's wise to stay away from random events entirely, provided you're responsible and creative with the generation of the numbers. Fulfilling the player's needs through pseudo-semi-random generation is an excellent way to breathe life into an otherwise dull game. (And further enhance an even better one.)

[RedCloud]

I often find that with retail RPGs like the Final Fantasy series, the player is so captured by the story that they want it to stretch out like a journey that they were actually apart of, because, face it, journeys like that don't end in three days. Also, there are a lot of side quests and sub quests that the player doesn't need to take part of to complete the game, but if the player wants to complete it ALL properly and get 100% enjoyment they'll have fun in doing so.

With fighting games, racing games and such, there doesn't need to be an extremely stretched out gameplay, just some pure action and fun that will keep the player entertained for hours on end and get them thrilled about the whole experience. And also, there needs to be some great multiplayer functionality, and finding the right combination for the game being made will take some time.