Thursday, June 14, 2012

Reflections On A Valley Without Wind's Transformation From 1.0 to 1.1

Whew!  It's been a busy month and a half since A Valley Without Wind 1.0 launched.



The release notes are available online in full, and I won't even attempt to summarize all that has happened during this period.  Microsoft Word informs me that the release notes are about 48,500 words long (the equivalent of 194 pages of a novel).  So, yeah.  Lots and lots of words, more than anybody in their right mind wants to read.

The purpose of those release notes, of course, has been ongoing communication with our core community -- every few days (or sometimes multiple times in a day) as we added to those notes, they'd read and respond.  This back-and-forth between us and the ever-growing community has been absolutely vital for the game.

I believe Erik has written up a number of things about a laundry list of features that we included in 1.1, and so has Josh for the release notes themselves.  I want to take a bit of a wider look from a design perspective, however.

Perception vs Reality: Number And Variety of Enemies And Spells
This is an interesting one, because the number one valid complaint about 1.0 had been that there were not enough enemies.

So we went and added 15 new enemies, 8 new minibosses, six new classes of infestations (as well as the concept of infestations in general), and 16 new elite enemies (as well as the concept of elite enemies in general).  All in all that was actually fewer enemies than I had hoped to add during this period, but it's not like we're stopping development now that 1.1 is done; and already the feel of the game between 1.0 and now is palpably different.

One of the more frustrating things about some of the earlier comments about the game was that there seemed to be a perception that there were only four kinds of enemies in the game (whereas there were a couple of dozen).

To some extent that may have been the fault of the commentators, in that they weren't exploring wide enough and instead were just being completionists (which the game warns players not to do).  But on the flip side, I think that was also to a good extent our own fault, because certain enemies (espers, mainly) were far too common throughout the game.

With terrain as crazy and craggy as this game has, a good mix of walking and flying enemies is paramount: the walking enemies have only so much mobility, and so the flying ones are needed in order to mix things up and keep you on your toes.  At 1.0 we were leaning really heavily on bats and espers to fill that niche.  That was a mistake on our part, because it meant that even though there were a lot more enemy types around in general, the world was really saturated with a few enemy types in particular.

And so the perception was that there were very few enemies; something both understandable and frustrating.  That's something that 1.1 rectifies, and it's also something that we'll be continuing to expand upon as we get further and further into post-release development in general.

Accidental Grind
One of the things that our beta players were urging us to put in our marketing materials was "it's like an RPG without the grind."  Which we were, of course, very proud of -- but thankfully that didn't make it into our bullet points.

Because, naturally, one of the biggest complaints about 1.0 that some people had was the amount of grind.  Go figure.

This was very unexpected and hard for us to understand at first -- after all, we had hundreds of beta testers telling us how great it was that it wasn't grindy.  What we came to realize was that these players had become experts as part of the beta process.  They knew all the tricks and optimal ways to play, and were avoiding the grind by using certain playstyles.

Similarly, some reviewers gave the game glowing reviews partly because their own personal playstyles just meshed so well with the game.  They fell into the good habits either by actually reading the tutorial messages (ahem), and/or by simply luckily having a playstyle that was not grind-prone.

There were many, many changes that we made to the game to combat the grind that came to our attention post-1.0.

Perhaps the largest was the removal of Civilization Progress (CP), which was previously the measure of escalation of the conflict on the continent.  The problem was, the best rewards were tied to CP increases, and so this was encouraging all sorts of un-fun behaviors in even experienced players, adding to the grind.

By removing CP, we removed the anxiety that went along with playing missions on the world map.  Now it's a discrete and player-controlled event that escalates things on the continent: the killing of a lieutenant.  We also went out of our way to make the lieutenant towers more varied and interesting (and briefer), and now they are something that happen periodically rather than all at the end of the continent.  It's kind of surprising we didn't think of that one sooner, but oh well.

There were little things like this all over the place, though.  If you were new and dying a lot, you'd be constantly having to hunt down upgrade stones to improve yourself.  If you were experienced and doing well, you'd have upgrade stones to crazy excess.  If you wanted to use guardian power scrolls you had not only to get the right NPCs and buildings AND scrolls, you also had to collect a bunch of consciousness shards of specific colors.  Removing the shard costs from the scrolls really smoothed that out.

Or let's say that you wanted to extend your wind shelter or buoy network.  Well, you were completely at the mercy of the random number generator to happen to give you one.  With the addition of the opal guardian store (mentioned more below), this provides an outlet where you can take more control of your own fate without having to grind.

There's a ton of other examples as well, and they've all been resolved as part of 1.1.

Predictability In Procedurally Generated Systems
As you may know, this isn't our first procedurally-generated game.  AI War: Fleet Command has procedurally generated galaxies, with enemy forces that are all procedural.  The thing that we've found that is key in procedural systems is for there to be enough layers of algorithms, and enough freedom within each algorithm.

With the terrain and object seeding and room seeding in this game, that's something we had gotten very good at.  Although, there was one bug that was causing the room layouts to be frustratingly homogeneous (relatively speaking) at 1.0.  We had hundreds and hundreds of room layouts created by our staff and by players, but only dozens were actually being seen at 1.0 because of that bug.  Facepalm.

Other than that, though, we had a lot of varied sort of caves and so on; it was something I'd spent a huge amount of the development time of the game personally working on.  Even so, it wasn't fully enough.  Since 1.0 I've added a couple of broad new seeding algorithms that give whole new classes of results, and the new infestations that Keith added bring a whole new layer of challenge to specific chunks.

Because that was part of the problem: previously, we were very region-focused, and had a lot of rules that were region-wide.  The infestations are one of the first things that lets us make a specific chunk extra-special and different.  Running through various caves and then suddenly coming upon one filled with dangerous waterfalls is just... cool.  It's really interesting how much this adds to the sense of exploration and adventure, because it makes things less predictable.

That's a trend we're going to keep on with as we develop the game further, of course: how can we make individual chunks of the game feel ever more unique so that players are always a bit on their toes.  I think we mastered that with AI War, and we're getting there with this one.

The elite enemies also really fall under the same aegis, actually: they only show up sometimes, as the conflict on a continent increases in ferocity.  The overlord gets a certain number of points every time the civilization level goes up, and behind the scenes he spends that on unlocking certain enemy elites that you then have to face for the rest of that continent.  On the next continent you might be facing completely different elites.

This again goes back to helping aid the sense of "I haven't seen it all" and making all the continents more unique and unpredictable.  Because, let's face it, with an adventure game of any sort, the fun is in seeing new things -- so as soon as you start feeling like you've seen it all, you're done with the game.  Fortunately this game evolves through continual updates anyhow, so even if you "saw it all" a month ago, there's a metric ton of new stuff since then.

Attachment To Procedurally-Generated Characters
Let's be realistic here.  You're never going to have the same level of emotional attachment to a randomly-generated character that you do to Link, or Aerith, or Celes, or Cyan.  Early in alpha we thought we'd give that a try, to make you care about procedural characters, but none of our prototypes were promising and so we scrapped the idea.

The thing is, and this is also something we learned from our game AI War, we humans do assign emotional meaning to random occurrences.  Certain randomly-generated planets in AI War take on significance to players not because of their nature, but because of what happened at them: historic victories or defeats, mostly.

In Valley, we see the same sort of thing with some players getting quite attached to a character with a cool name who lives an abnormally long period of time (and possibly who has a number of near misses with death).

The problem is, when characters consistent of a portrait, a sprite, a name, and a very few random stats it is hard to care about them.  Thus we really revamped the whole character selection process to make it so that characters were more unique based on their native time period, as well as more unique individually (with special bonuses).  Thus the odds of ever exactly duplicating a specific character has gone way down, because they are now more unique.

We also added the ability to rename the characters right as soon as you select them, so that you can change a name you hate.  But come on, when the game gives you gems like "Judge Glass" we have a lot of players who are perfectly happy to keep the random names. ;)

Perception Versus Reality With The Citybuilding Segments
Another rude awakening for us with 1.0 was that many commentators thought the citybuilding parts of the game were "pointless."  As in, literally serving no point.  At first this was just frustrating, but after thinking about this some more we realized what was going on.

The purpose of the citybuilding structures is to raise the profession skill of your NPCs.  This in turn lets those NPCs use powerful scrolls from afar to aid you in your fight.  The problem is, that's a four-stage process: find NPC, build structure, find scroll, use scroll.

If you just find an NPC or a structure or a scroll out of context, and you're kind of casually playing through the game, and don't put in a lot of time, it's thus easy to conclude that both NPCs and structures have no point.  Oops.  That's the strategy game developer in us, making it so that certain things have only very long-term rewards and you have to be patient.  Patience isn't often asked on that scale with games of Valley's genre.

So what we did was this: we made it so that all the citybuilding structures now grant you a continent-wide small bonus immediately (and permanently thereafter).  Now you don't have to figure out that the buildings lead to higher-skilled NPCs that can then use better guardian power scrolls.  You'll get to that when you get to that, and there's no rush.  In the meantime, all the structures have at least some obvious point to them right from the start.

This is another great example of where perception trumps reality -- and I'm not just saying that about other people, or as a slam against anyone.  Looking at someone else's game cold, I'd fall into the same traps.  The game designer has to ensure that perception and reality line up with one another as much as they possibly can.

Extrinsic Versus Intrinsic Rewards
Later this year I turn 30.  So I'm not really that old, but in terms of video games I guess I'm ancient.  See, when I was growing up, games were short and brutally hard.  And we all basically sucked at them, because nobody had years of experience playing games (since games hadn't existed in a consistent state for many years).

Getting over that first damn hole in Super Mario Bros 1 is still an event I remember -- and how many Game Overs did we get just trying to get past those first few goombas?  I can still remember my mom jumping up, not really paying attention to her trajectory, and hitting the bricks above as small mario, then bouncing down into the side of the goomba and dying.  Again and again.

It was stuff like that -- but it was still interesting.  These games were like a puzzle, and seeing what was at the end of the level, or the level after that, was what was so cool.  There weren't achievements, but there were plenty of secrets if you wanted to take the time to look around.  There was a score that was kept, but I don't think most people paid attention to it.

And when it came to killing goombas in Mario, or most enemies in most other NES-era games, your reward was... nothing.  You got past the enemy.  Your goal was to get to the flag, the enemy was there in your way, and that was that.  Consequently, speed runs became the order of the day once we got better at things, because you could memorize the levels and run as fast as possible while bouncing off a few enemies and ignoring most of them, to see how fast you could beat the level.  Obviously people still do this.

Finally turning this discussion back to A Valley Without Wind, originally I had been adamant that enemies would work in the same fashion: no rewards for killing enemies.  They are like the enemies in Mario, to be avoided our fought at your pleasure.  Personally I quite like speed-running the surface areas with storm dash as if I'm Sonic, and then exploring the undergrounds and interiors more cautiously as if I'm Samus.  A lot of other players seemed to agree with me.

A lot of others vehemently disagreed, and felt like there should be some reward for killing enemies.  I don't know if it's partially a generational thing, or just how modern games have trained people to think, or a matter of the genre expectations that Valley evokes in some people (it has echoes of an RPG, and even in the oldest CRPGs you get direct rewards for every enemy killed).

It was clear that this wasn't going to be an issue that was going to ever drop, so what wound up happening was a simplification of the consciousness shard system of the game into a single currency instead of six currencies, as well as the addition of a store in the settlement where you could spend them.

This was met with universal joy, which is a rare thing with any decision; and it also allowed us to shore up some situations where "when the random number generator hates you, you can get stuck."  To my knowledge, in 1.1 there aren't any more of those sort of situations in a macro-game sense.

Farming Of Extrinsic Rewards
The risk of any such direct-reward system is this, however: "farming" of those rewards.  Whenever we design a given feature for one of our games, we have to try to imagine how every possible player personality will use that feature.  The middle of the road player is easy: they'll do most activities, miss most of the secrets, never discover the ideal balance (but find something favorite that works for them), and still have a fun time.

But then you have the OCD-like players (of which I usually count myself a member, depending on the genre) who figure out every game mechanic in incredible detail, find the ideal balance, find most or all the secrets, and generally do whatever it takes to play "optimally," even if "optimally" is a less fun way to play than the middle of the road path.

My favorite example is from our game AI War, where players would go to great lengths for "knowledge raiding," an then complain that it was easy but grindy to do so.  Our expectation had been that they wouldn't be knowledge raiding in the first place unless they were in dire straights, but many had discovered that the "optimal" path included the practice.  And so that's how they played, even though it drained a huge amount of the fun of the game from their experience.

It took us months of prototyping and experimentation and feedback with players before we ever solved the knowledge raiding problem such that it was still viable and yet not part of the general-purpose optimal path.

Anyway, back to Valley, the worry had always been that if monsters dropped currency then the OCD-like players would take it upon themselves to farm that currency.  Therefore we had to carefully balance the amount of currency you get versus the cost of items at the store; and we also have alternate sources of currency, such as simply finding stocks of them as you free-roam through the world.

Thus the optimal path is... playing the game.  Which is the goal in the first place.  When the "optimal" path takes players into some sort of farming situation in a tiny subset of the game, that undermines the entire experience.  Generally that tiny portion of the game is not fun or varied enough on its own to really be that great.

I haven't played WoW, but I heard a number of complaints about Mining being that way a few years back.  And in my favorite game of all time, Final Fantasy 6, I find myself always having to stop to grind EXP right before going through the events in Thamasa (the serial boss fights on the airship, for me, were a lot more fun if I was overpowered).

Anyhow, the core point is that players come in all shapes and sizes, and depending on their personality they'll love some things or hate them.  So making any one mechanic too central to progression, if it's not something universally loved, is not grand.  If it were not for the story and music of Final Fantasy 6, no way would that be my favorite game, for instance.

My goal is thus always to provide players with the freedom to do the things they enjoy in a game I work on, while minimizing the things they don't enjoy.  In our game Tidalis, you don't have to play the adventure mode -- nor to you have to do the brainteaser puzzles.  Both are really different forms of gameplay, and you can pick and choose.  In Valley, the mission variety works much the same way.

Bottom line: when it comes to farming currency, we made it so that there are enough other ways to get the same end products that you don't have to farm.  If you find yourself farming, that's your choice.  But it's never the optimal choice unless that's simply what provides the most fun for you (killing trash mobs is what Keith and I call "bubble wrap popping fun," so certainly that can be a stress reliever sometimes on its own and it's nice to get the currency as an added bonus).

Fear of Failure Vs Free Play
In this amazing video of John Cleese talking about creativity, he talks about one of the essential facets of "play:" that it is without consequence.  In other words "let's see what happens when I do this!" and then looking at the result.

At core,  my vision of A Valley Without Wind has always been as a fantasy world where players can come and do "whatever they feel like today."  Of course the overlord is there and needs to be defeated.  But there are so many ways to go about it, and so many intermediate goals (all optional), and so many side adventures to be had, that the player can just come and enjoy themselves in whatever manner and in general make at least some progress toward their larger goals.

The problem with the design of the game as it stood at 1.0 was that we took a little bit too much of a strategy game mindset.  Or rather, too much of an ultra-hardcore strategy game mindset.  It's like the awful father who gives the child a choice of what they want to be beaten with: the wrench, the belt, or the switch.  Choosing the switch still doesn't make this a happy event, right?

On the other hand, if you look at a dessert menu and there are a variety of treats of which you can have only one, that's a positive experience no matter what.  You might be torn between a couple of menu items, and the choice might be agonizing in a pleasurably fun way ("oooh, which to choose!"), but either way this is a happy event.  You still get a treat at the end.

Between 1.0 and 1.1 there's really been a philosophy shift across the board, from everything to the design of many of the missions to the overall progression flow of the game.  Rather than the failure state being quite so threatening, it's more of a lack-of-success state.  You can try a mission type that might not be to your taste, fail, and decide if you want to do it again or if you want to give that a wide berth from then on.  Your prized character "Judge Glass" doesn't have to die as a consequence.

Even more centrally, just because you want the reward of some mission on the world map, you don't have to be increasing the overlord's power and making it more likely that you are going to suddenly be frighteningly underpowered compared to the rest of the continent.  There's still a strategy to how you play, because you can only unlock so many skills and there is a lot of choice on how you customize your character and civilization, but now we're doing it via the carrot rather than the stick.

Or look at it another way: at 1.0, if you did five world map missions and got whatever rewards, then the continent tier would go up and you would be facing harder foes.  You may or may not have spells that are ready to deal with them, depending on your choices.  In 1.1, the only way to make the continent tier go up is to defeat a lieutenant, which is always one tier higher than you.

This creates a more natural flow: the very act of defeating that lieutenant is an active, aggressive move on your part.  It can only be accomplished when you are ready to take on the next continent tier in general.  You don't get to play level 1-2 in Super Mario Bros until you've already mastered level 1-1 well enough to at least get to the flag at the end.  The older style of the design was more akin to having it take you to level 1-2 as soon as you killed 20 goombas; that had no real bearing on your preparedness for the next continent tier.

With 1.1, players are free to experiment because failure means less -- it simply means that you don't progress forward at that time.  At 1.0 it meant that not only did you not go forward, you might actively be thrown back, and that's never fun.  Well, except to hardcore strategy nuts, which most of our pre-1.0 beta testers (like us) were.  Go figure!  Fortunately we figured out a way to retain the strategy while reducing both fear and the learning curve for 1.1.

Artificial Paucity of Spell and Enchant Options
In my long series of articles about designing emergent AI, I discuss the trap of the "best path" for play.  In other words, when players in any genre find a killer combination or weapon or class build that makes all other combos/weapons/builds obsolete.  That sort of thing is the absolute bane of both strategy games and fighting games in particular, but it also affects other genres.

The problem with 1.0 of Valley was twofold when it came to spells.  First of all, our mathematical models for balancing spells were not nearly sophisticated enough.  That's been corrected only in the last week or so.  But secondly, even among spells that were ostensibly balanced, the full use of them was not nearly clear enough.

Because, you see, spells don't exist in isolation.  That's always been the case with ships in AI War, as well.  A ship that seems useless to a new player is often a key player in some advanced strategy involving other ships in combination.  That is also true of a few of the spells in Valley when they are combined with certain enchants (which alter spell behavior and power).

The second part of this problem was compounded by the fact that we started players off with no enchants at all, and the introduced them slowly.  Enchants are now introduced a bit faster in general, and players now get a starting selection of about 10 enchants when they reach the settlement for the first time.

This is after the intro mission, so they've already had time to assimilate the basics of combat and movement and navigation, etc.  At this point they're now reaching the settlement and getting a wider array of spells, and deciding which are useful and which are not.  At 1.0 they were doing so out of a position largely ignorant of the hundreds of thousands of possible enchant combinations.

With 1.1 there's now at least a strong suggestion -- without being overwhelming -- of the sort of customizations that the players can gain.  And in fact each player can start to customize a bit right from the start, since each of them will get different randomized starting enchants in each world.

The problem of "best path" was not at all limited to just the spells, however: at 1.0 the Leg Slot enchants were also heavily biased to this.  There were all manner of leg enchants: running faster, jumping higher, powersliding, and then of course double and triple jumping, among a few others like windstorm resistance.

The problem was, once most players had double or triple jump, that was all they wanted to use.  And of course we wanted them to have that: using wooden platforms to painstakingly traverse terrain is fun at first because it makes the world feel dangerous and large.  But then as you gain more powers of mobility, the need to lay wooden platforms sharply decreases and you feel that now you are a person of power in a dangerous and large world.  This is, quite frankly, fun.

And it meant that the other leg enchants might as well not have even existed for all the use they got.  In all other aspects of the enchant slot design we were very careful in what we set in opposition to what else.  Each slot has its own distinct flavor, so that you are never having to choose between the main course of a meal and a side dish -- of course everyone chooses the main course every time.

What we therefore did with the leg enchants was to split out the double and triple jump into a new Feet Slot.   And we added a kind of throwaway Elven Boots for players who prefer not to use the multi-jump enchants (we already had a few in the forums saying they wanted some sot of alternative).  The result of this change was that suddenly all the enchants were actually viable, versus so many of them being overshadowed by the multi-jump enchants.

Overall Conclusions
It's been a very busy month and a half, indeed.  We learned much from our players during our beta period that stretched from last September until this April, and we've learned yet more still from them throughout this May and June.  The most key fact has been that we've had a broader array of players giving us feedback since 1.0, in particular players coming to the game "cold," without having followed the game's development from alpha through beta and on to release.

That's been a useful perspective to have alongside our longtime stalwarts who have been with us the whole journey: that helps make it so that no one opinion dominates too sharply, leading the game to veer too far into the hardcore or the casual.  My goal has always been for this game to be very easy to get into and have some fun with, but deep and hard to master.

I believe that 1.1 succeeds at this far better than 1.0 did, to the point that they are almost different games no matter how similar they may look at first glance.  On the other hand I also believe that our work is not nearly done if the world of Environ is to reach its utmost potential.  Slow and steady wins the race, and -- players willing -- we at Arcen intend to be adding onto this game for years to come.