An Interview with Brian Reynolds
The designer of Alpha Centauri answers a few tough questions, including why he decided to restrict gameplay to one planet and how he would respond to critics who think his game might just be another Civilization II
Some are calling this new strategy game from Firaxis Software "Civilization in space." Designer Brian Reynolds would beg to differ. Although the game might tap into his experiences as designer of Civilization II, there are some fundamental differences. And who better than Reynolds himself to take you through the process of his design.
In this final entry, lead designer Brian Reynolds shares his closing thoughts on Alpha Centauri and discusses the critical importance of feedback received from beta testers and fans who downloaded the Alpha Centauri demo.
Alpha Centauri is in the home stretch. Brian Reynolds talks about the final programming stages, diplomacy, and waiting your turn.
In this entry, Brian Reynolds gives producer Tim Trian a turn to talk about play-balancing the seven different factions in the game.
Designing a game that takes place in a "future historical" world is different yet similar to designing a game like Civilization.
Tweaking AI involves discovering why a computer opponent does "dumb" things. Brian gives a glimpse into performing this task.
By Brian Reynolds March 15, 1999
There were a lot of "firsts" for our team in the making of Alpha Centauri. We had never done a public beta test before Alpha Centauri, and this was also the first time we released a demo before the game was out. Since we'd not done one before, we didn't know exactly what to expect when we released it, but it turned out to fit right in with Firaxis' iterative design method.
For those of you who aren't familiar with our design techniques, one of the most important ways we go about making our games fun is by "prototyping" early and often. When it's time to start a new game, we don't start by creating a 50-pound design document - there's no such thing at Firaxis. Instead, we spend a couple weeks creating a prototype game that can actually be played; the first prototypes of a game are very simple and don't have great graphics, but you can actually play them. By playing the game (rather than just trying to imagine it, which is all we could do without a working prototype) we can see what is actually fun and what isn't, what works and what doesn't.
Once the first prototype is ready, the "feedback loop" begins: We figure out what's fun and what isn't, then we strengthen the things that are fun and fix or take out the parts that aren't. Sid calls this process "surrounding the fun." Once we've revised the prototype, it's time to play it again - and then revise it again, play it again, revise, play, ad infinitum! Over the course of the project, our simple little prototype grows one step at a time into a full-fledged game.
This process has been around since Sid's early days at MicroProse. It's how he's always designed games, and it's how he's taught us all to go about it. But in the Firaxis years we've added some significant innovations to this process.
We've discovered that as time goes on it is valuable to expand the group giving feedback to the prototyping process - once we've played and developed the prototype "in house" for a while, it's time to bring in some outside gamers with fresh perspectivesand see how we're doing. We do "blind" tests, where we bring in players who've never seen/heard of the game and see how they perceive it. This gives us valuable insights at a time when we've probably become too close to our own creation to spot all of the "new user" issues.
Next - and, as I've said, this was a first with Alpha Centauri - we expanded our feedback loop to include a group of volunteer public beta testers. This not only supplemented our professional beta-testing efforts, it showed us how experienced game players would react to the game. We got a great feedback loop going with our beta testers and made many changes to our game system based on their excellent recommendations. We'd never done a public beta test before, but we'll definitely do it again.
Then, it was time to release our demo. We had originally thought of the demo entirely as a marketing tool, but as soon as we released it we realized that it made perfect sense as part of our process: Here was a fairly polished, almost-final version of our game in the hands of the public, and as soon as the thing was out we started getting all sorts of feedback from it. Since the release version of Alpha Centauri was still in final testing, and our actual ship date was more than a month away, we had time to incorporate many suggestions from e-mail and Internet newsgroups into the retail version - much to the game's benefit.
Thanks to the demo release, we could make improvements to the game's interface, add a couple of new features, and fix a few glitches found by demo users that had somehow slipped through our testing process. Most significantly, the design of the game itself was still "in play" during the demo process - we could improve some rules, fine-tune the game balance, and improve the AI, all based on feedback we got from the Internet. Even though we hadn't planned it that way, the demo release really proved to be a perfect addition to our design repertoire.
And now, with Alpha Centauri out on the market, our feedback loop still continues - we've released one free enhancement download, and another is expected imminently. All these enhancements were primarily inspired by the excellent suggestions and feedback we received from players.
By Brian Reynolds December 9, 1998
With only a few weeks left until our "master date," Alpha Centauri is finally in the home stretch. In this industry, there's nothing more rewarding than working on a project that is clearly "coming together" to become something big, and that's definitely the feeling here at Firaxis Games right now. The gameplay feels well balanced and addictive (we've had some early "hey is that the sun coming up?" casualties), our new polished interface is just about ready, the movies are the coolest we've ever done, and we've beaten the bug list down to a reasonable level. An even better sign, our 25 public beta testers seem really happy with us.
Given the glaring omission of Net support in many of the previous games I've written, one news item that I'm sure will please our fans to no end is that Alpha Centauri's multiplayer mode has, to our considerable surprise, turned out to be not only stable, but lots of fun. Games like Alpha Centauri are quite difficult to take to the Net - turn-based and complex, with a vast variety of player options and opportunities for input, they don't fit too well with the "sure, we lose a packet here and there but who really cares" world of network play. But it turns out I didn't spend my college years doing parallel programming at defense contractors for nothing! We came up with a good algorithm and darned if the thing isn't working.
As with all parts of Alpha Centauri, multiplayer AC will have lots of user configurability and player preferences. Multiplayer AC features "simultaneous movement," meaning that all players take their turns at once; we use a record-locking system similar to an airline booking system to make sure that two players don't try to do something to the same square at once - whoever hits the key first gets first shot. We have an optional and configurable "turn clock" to keep the slowpokes moving along. You can assign a fixed amount of time per turn and can also add bonus time as more units, bases, and events come along. Players can also earn "bonus time pills," which they can expend to add extra time to a particularly hairy turn.
Diplomacy in multiplayer AC games includes the usual human-to-AI diplomacy, but also direct human-to-human negotiation: Each person adds things (tech, energy, treaties, maps, commlinks, war plans) to the list of what they're offering. If both players click "agree," the trade takes place, and everything on each list is traded simultaneously. You can also convene the Planetary Council, where all players in the game (human and AI) vote together to elect a Planetary Governor and set global policies.
A "full game" of Alpha Centauri, from Planetfall to Transcendence, can take quite a while to play, so for multiplayer purposes we're also working on some nifty scenarios, both cooperative and competitive. Find the artifacts, escort the ambassador, wipe out Morgan Industries, etc. Elegant "mini-centauri's" that you can bite off in an hour or two of Net play. But the true diehards can go all out for marathon games on huge maps (and yes, of course, we do support "save game" for multiplayer).
Any other questions? Good. We'll go spend a few more weeks polishing this thing, and then we'll see you on the Net!
by Tim Trian August 11, 1998
Factional Strife - Balancing a Masterpiece Often, when I have to tell people what I do for a living, I get a similar response: "So you sit around and play games all day, right?" As the producer on Sid Meier's Alpha Centauri, the best days of my job are when I can actually sit at my desk for eight hours straight and just play the game, occasionally tabbing over to MS Word to note some mistake the AI had just made. The completely addictive quality of Alpha Centauri just makes these days all the sweeter. At 6:30pm I look up glassy-eyed and think, "Where did the day go?" We've been doing a lot of that lately at Firaxis, as we work to balance each of the seven factions you can play in the game.
Each faction in Alpha Centauri is founded on an ideology, with strengths and weaknesses related to its beliefs. We have several goals in balancing those attributes: Each faction should have certain strategies suggested by its culture. However, the faction should not be "locked in" to those strategies - if the game situation warrants it, a faction should be able to deviate from its optimal path. Also, the faction strengths should give the player interesting things to do without unbalancing the game. Finally, and most important, the faction must be fun to play, ensuring the player has a new and different series of interesting and challenging decisions to make each time he or she plays.
Each faction excels at one or two important aspects of the game. Players can maximize these strengths by pursuing certain strategies. For example, the University of Planet under its leader, Saratov, acquires new technologies quickly. Saratov can maximize this strength by building Network Nodes (which increase Labs production) and assigning workers as engineers (also increasing Labs). He can also build Secret Projects, which speed up research, such as the Supercollider and the Network Backbone. As a result, his is often the most advanced faction in a game, tooling around the mountains of Alpha Centauri in Graviton Hovercraft when everyone else is still using Missile Crawlers. As their weaknesses, however, University citizens often riot, and their academic networks provide easy access for enemy infiltrators. As a result, those Graviton Hovercraft may have to spend some of the game acting as police units to quell rioters or moving back and forth between patrol markers on watch for sneaky enemy Probe Teams.
Our goal is to balance these strengths and weaknesses. For starters, you don't want a faction to be dependent on its strength. For example, sometimes the University may have to fight an early war or may have to terraform its way out of a difficult starting position. In these cases, another faction may get a jump on tech advances, grabbing the important Secret Projects - but the University should still be able to come back. Even when deprived of its optimal strategy, the faction's weaknesses should not be insurmountable. The University player in such a situation can pursue alternate strategies by building Secret Projects that suppress drone riots or base facilities that protect against Probe Teams.
At the same time, no faction power should be dominant over the rest. For quite a while, the Lord's Believers, under Sister Miriam, could build Colony Pods at an accelerated rate, expanding their empire quickly and building on their inevitable early lead. We experimented with severe weaknesses to compensate for her power, but this left her vulnerable to the problem described above: She was dependent on her strength. If she found herself on an island, where fast expansion was difficult, her weaknesses dragged her down to the point where her faction was no longer significant. Now, the Believers get a "fanatic attack" bonus when on the offensive, countered by somewhat slow tech research. These attributes balance themselves out nicely - the Believers may have technologically obsolete units, but their fanatic bonus gives them an edge if they initiate an attack. These attributes also suggest a strategy: Go on the offensive and capture techs from your enemies to compensate for your weakness.
Of course, none of this balancing means anything unless the faction powers are fun. For us, this means always giving the factions something cool to do with their attributes. For example, an old power for Gaia's Stepdaughters helped reduce ecological disruption resulting from base pollution. The Gaians gained protection against negative game effects resulting from ecological disruption - fewer bad things happened to them - but they still had nothing cool to do with themselves. This was not fun. Now, the Gaians can capture native Mindworms early in the game, giving them a weak but numerous military force of scouts and defenders. We are still experimenting to see if this makes the Gaians too powerful.
Happily, the only way to balance these powers is to play the game, repeatedly and in-depth, which is a rule at Firaxis. In single-player mode, we try all sorts of extreme strategies to find any sure-fire paths to victory We also note how often a particular AI faction ends up at the bottom of the heap - a dead giveaway that its powers are too weak. Of course, multiplayer games provide more insight into balance. A human's intelligence exploits loopholes in the game system more effectively than the AI does, since we train the AI to follow certain pathways. Of course, I've noticed that when I win using a particular strategy, it gets chalked up to "unbalanced powers" (and a new set of attribute rules soon follows); when designer Brian Reynolds wins, it's "superior gameplay."
The end result is a product of unprecedented depth, scope, replayability, and addictiveness, where the player is always challenged by the game to come up with new strategies. There are no all-powerful factions, no unstoppable tactics. If we've done our job, then Sid Meier fans across the world will be spending sleepless nights immersed in creating their own approaches for the future of mankind on a tiny planet warmed by Alpha Centauri.
By Brian Reynolds
One of the best things about game design is getting to pretend to do a whole variety of really interesting jobs. Depending on the game I'm making, I may get to be a historian, an anthropologist, an archaeologist, or a military officer. These days, with Alpha Centauri, I'm pretending to be a scientist, a science-fiction writer, and perhaps a philosopher.
Since the Firaxis team has won considerable praise for its historical games, it seems fitting that our first foray into the world of science fiction should be a "future history." As a colleague recently pointed out, good science fiction thrives on constraint - you need some things you can't do, as well as things you can. So rather than simply concoct a fantasy galactic empire and a host of shallowly conceived alien races, we chose to begin with near-future science - technologies just over the horizon. As we proceed into the future, we try to present a coherent picture of future developments in physics, biology, infotech, economics, society, government, and philosophy. We're not saying everything depicted in the game is known to be possible, but we do promise considerable attention to detail in extrapolating future possibilities.
Of course, a coherent future history is just so much extra back story if it isn't integrated into the gameplay, so one of our challenges is to integrate interesting science fiction into an equally compelling game design. At the "Sid Meier School For Better Game Design," a successful design presents the player with a series of interesting decisions - decisions where he (or she) feels genuinely torn between choice A and choice B - and allows the player to win by any of several different strategies. A decision tree with only one right answer, where the player always wins with choice A and loses with choice B, isn't an interesting decision at all, and the fun of creating your own strategy quickly becomes the chore of trying to read the designer's mind and find the "right" answer.
So what I'd like Alpha Centauri to present is a series of logical, interesting, and coherent choices that allow the player to create the future. You're not learning about human history, you're creating human history. With that in mind, let's proceed to the part you've all been waiting for, examples from gameplay....
The thought of colonizing an alien world immediately brings to mind the idea of altering, or "terraforming," that world to make it more livable for humans. To allow the player to actually make meaningful changes in the climate of a world required us to create a world builder and climatic model far more powerful than anything we'd done before. Temperature, wind, and rainfall patterns have been modeled in ways that allow players to make changes and watch the effects. For instance, with a modest array of future technology, you can raise up a new ridgeline or mountain range. With the prevailing wind coming off the sea to the west, your new mountains will tend to empty moisture from the air on their western slopes, leaving their eastern slopes and the downwind valleys quite dry. So you can raise ridgelines to increase the rainfall (and thus nutrient production) for your own bases while simultaneously drying out your downwind enemies!
As your technology progresses, you can also divert the courses of rivers and dig huge thermal boreholes down to the planet's mantle to release heat (not to mention the mining opportunities). Still later, if enough players work together, you can alter the climate on a worldwide basis - global warming to melt the icecaps and raise sea levels, for instance, where you can watch enemy bases sink beneath the waves. The ultimate terraforming triumph is a breathable atmosphere, but watch out, since some changes may lead to a population explosion among the native life-forms!
In addition to scientific advances, Alpha Centauri speculates on the future development of human societies. As the leader of your faction, you map out your vision of future utopia by deciding where you stand on a whole series of value choices - economics, liberties, religion, health care, military service, education, environment, and information. On each issue, you can choose to take a "ruthless," "moderate," or "idealistic" stance, with a variety of consequences. Choose the ruthless "free market" economy, which promotes efficiency and economic growth at the expense of the environment, or the idealistic "fair market" economy, which encourages citizen loyalty, population expansion, and raw industrial output at the expense of economic growth.
Similarly, you may choose (ruthlessly) to conscript a massive, inexpensive, but ill-trained army, or you can (idealistically) raise an all-volunteer, highly trained, and very expensive force. Ruthlessly, you maintain ironfisted control of your information networks, preventing enemy infiltration, not to mention the corrupting influences of pornography and subversive literature. Or idealistically, you open your networks to the free exchange of ideas and information, reaping the rewards of greater creativity and productivity but running the risks of infiltration, corruption, or even open rebellion. Idealistic public health care keeps your citizens happy and healthy, not to mention loyal; a ruthless health care "for profit" scheme encourages economic efficiency, not to mention advanced medical research.
In each case, we've gone out of our way to avoid promoting a single "right" answer. As you create your vision of future society, each value choice has positive and negative consequences, and the choice between good and evil will rarely be black and white. You can create literally thousands of different societies in Alpha Centauri - an atheistic, polluting police state with a free market economy, universal education, and all-volunteer military. Or perhaps a devoutly religious democracy with a heavily censored information network, conscript army, and cradle-to-grave health care. One way or the other, we'll make you think, and therein lies the secret to an addictive game.
By Brian Reynolds
Even in this new "Age of Multiplayer Gaming," one of our most important goals at Firaxis is to create games that have (pardon my buzzwords) lots of depth and replayability in single-player or solitaire mode. The Internet is cool, multiplayer games are cool, and Firaxis plans to include full multiplayer support in all of its games, including Sid Meier's Alpha Centauri. So many companies are turning away from single-player play or think of it as mainly a "tutorial mode" to be used before graduating to the "real" multiplayer experience, but we think a vast and mostly silent majority of gamers still spends most of its time playing solitaire - and we plan to be there.
So what gives a game the coveted replayability that turns a single-player game into a classic? Well, a lot of things, of course, but one of the most important is good AI - smart computer opponents. One of the most common reasons gamers give for preferring multiplayer games is that human opponents "are smarter and make more intelligent, unexpected moves" or conversely that in single-player games, "the AI is crappy." Knowing this, some developers are ready to throw in the towel on single-player AI and go all multiplayer. As a designer, though, I take it as a challenge to try to create the kind of algorithms that keep players coming back for more.
I've often been quoted as saying that when designing AI "I try to make the computer play the way I do." At first glance, this may seem like a meaningless statement, since by definition artificial intelligence is an attempt to emulate human decision-making. But what I am really describing is the process, not the end product. Now I'm (blush) a pretty good strategy gamer, so I begin by playing the game myself, keeping track of the decisions I am making and even more importantly why I am making them. I also turn on the omniscient view and watch what the computer players are up to, making notes when they do something "dumb." The trick with AI is to find the right "angle of attack" for a particular problem - when a computer is faced with a choice between several attractive alternatives, I want it to be able to think through the problem the same way I did.
So, in Mission Year 2150, as human life gains a tenuous foothold on an alien world, Lady Deirdre offers to trade me CEO Morgan's comm frequency for five energy credits. (Hmmm, I would have given her a whole tech for that, knowing I would likely recoup several from trading with Morgan....) My archenemy Sheng-ji Yang sends a trickle of Alpha Crawlers against my outlying bases, one every couple turns. I defeat them in detail. (Hmmm, I would have waited and massed an overwhelming attack force, then gone in all at once....) Colonel Santiago's forces blunder mindlessly into my territory, trying to find a way through to attack Dr. Saratov's University, not realizing they're violating their agreement with me as they do so (got to give these guys a better concept of territorial integrity).
It can be a frustrating process, since in the early versions of a game the computer players do something rock stupid every couple of milliseconds, and it feels like there's no way to cover every possibility. But if I can put my finger on why a particular decision was stupid and identify the reasoning process I would use to come up with a smarter one, then I have the angle of attack I need to teach the computer how to solve that kind of problem. Deirdre's problem, for instance, is a simple one - the principle involved is merely "information is power." Let's increase the value of information (even information as seemingly innocuous as another player's comm frequency) and give the computer a better algorithm to compare the value of information with that of other tradable commodities.
Yang's problem, in detail, again involves a fairly simple principle (amass your forces before attacking), but the implementation is a bit more difficult. I teach the computer to pick out a base close to enemy territory as a marshalling point for its attack force. Then, as new offensive weapons are produced, a sizeable number of them are tagged as members of the soon-to-be amassed attack force (the remainder are left free to patrol friendly territory for intruders or make small raids against enemy strongholds to "keep him honest" while the attack force is amassed). "Tagged" weapons know to proceed immediately to the marshalling point, avoiding distractions and entanglements along the way. Once they arrive, they are told to wait until the rest of the force arrives. When the attack force reaches a certain critical mass (based on friendly production capacity and the personality of the attacker), its members are "released" all at once and can proceed together into enemy territory using the normal attack algorithm.
Colonel Santiago's territorial blundering is probably the most difficult problem to correct, since irregular geometry is not the computer's natural forte. First, we need a decent concept of what constitutes "my territory." The working definition I chose to start with is "anything closer to one of my bases than to anybody else's on the same land mass." Therefore, it is relatively easy to compute a table showing whose territory each map square lies in. Now, to teach these guys how to tell if another faction is an immediate neighbor or someone far away.... We start at each friendly base on a continent and use our "go to" algorithm to try to find a path to each known enemy base on the land mass with one added restriction: The path cannot pass through land claimed by a faction whose territory we've agreed not to violate. If such a path can be drawn, the enemy faction is a contiguous over-land neighbor; otherwise, we need to make alternate arrangements for our assault.
So, armed with some new algorithms, let's return to Mission Year 2150.... The cagey Lady Deirdre refuses to reveal CEO Morgan's whereabouts unless I produce my files on biogenetics... eight Alpha Crawlers rumble out of the fungal forests and pull up outside my headquarters along with an ultimatum from Chairman Yang... and Colonel Santiago patches up her differences with the University in order to concentrate on expanding into my territory. Mission accomplished. Now if I can just figure out why those transport cruisers keep steaming around in circles....
Got a news tip or want to contact us directly? Email email@example.com