Posted on Leave a comment

Blog: Designing countermoves in PvP games

A skill-oriented player in PvP games, when looking at a new game, would ask the following questions:

Players will start to learn the controls, and, in the meanwhile, try to figure out a strategy. That’s obvious. But what happens underneath the player’s consciousness? As game designers, how do we encourage players to play better and longer in our games? One effective way is to design countermoves— emergent solution to specific problems. In this post, I will talk about what they are, how they work and use card games as an example of how to design them.[1]

Background Knowledge

Before we define countermoves, the most important thing to look at is the control. When we say somebody is good at the control of a game, we mean that his or her input is precise and fast. Obviously, there are two ways you can challenge the players: make them unable to see correctly, or unable to act correctly.

Players “can’t see” because the game throws redundant and complex information in a short amount of time. They can solve this problem by:

  1. Analysis
  2. Relying on experience and memorizing levels
  3. Muscle memory that triggers reflectively

Players “can’t act” when they know what to do, but just can’t do it correctly or quickly, because the game requires the player to input a series of commands consecutively, or be strict on the timing of their input. They can solve this problem by:

  1. Familiarizing themselves with the hardware and control scheme
  2. Focusing on fewer things

Definition

In trying to focus on fewer things, the players summarize and abstract their experience of trial and error, and compile multiple steps into one simple solution in their mind. Then they discover a countermove.

What Do Countermoves Do

Countermoves provide players with the following advantages:

  1. Faster reactions, because they shorten the overall time needed for multi-step control.
  2. Greater chance of success, because they increase the chance that the players performed a right counter to a particular move.
  3. Greater ability to multitask, because they save the players the energy to do more.

Now that we know where countermoves came from, and how players can use them, we can design them for the players to discover.

Designing Countermoves

Countermove vs. Strategy

What’s the difference between a strategy and a countermove?

Players form a strategy when they put game elements together according to the rules of the game to gain a larger advantage or to find an optimal solution to the whole game.

On the other hand, countermoves are optimal solutions under particular, specific situations. They are the adjustment or backup plan when trying to play according to a strategy.

In short, we think about strategies before the game, and about countermoves during the game.

What Creates Countermoves?

Since countermoves are reflections under the dome of a strategy, usually, the more strategies a game has, the more countermoves there will be.

If strategies pair countermoves one-to-one, it doesn’t encourage players to think about countermoves. It only forces them to memorize solutions. However, we would love our players to think about countermoves, instead of bragging about a solution they found.

Therefore, what we need is a design that allows multiple solutions to the same problem. This way, players will need to make decisions as they go, and can’t have a dominant solution in every case.

(Taku: a choice) — The Available Countermoves

There is one genre of game that:

  1. Has no dominant strategy.

  2. In which the power of all choices is dynamically changing.

  3. Is not transparent — some information is hidden.

  4. Has multiple choices most of the time.

That particular genre is fighting games: Street Fighters, King of Fighters, etc.

Fighting games have enclosed strategy sets — the characters. The sets won’t intersect with each other since a player only controls one character at a time. And the depth of countermoves usually won’t go more than three layers because of the complexity of the design. It is a genre that simplifies strategy and empowers countermoves to the extreme. 

Yomi — an abstraction of fighting games.
Yomi — an abstraction of fighting games.

Because of such design, players invented the term “taku” — choices, to describe the meaningful countermoves that they can choose.

Above Countermoves

Because fighting games are so about countermoves, there is also a term that describes the movement of players when they have a lot of countermoves.

立ち回り (Tachimawari: moving back and forth)

Tachimawari is the action that happens before an attack, or before the defense against an assault. The goal of these actions is to move into an advantageous position. In fighting games, players move back and forth before attacking, hence the name.

When you have enough countermoves, you will do everything you can to bring yourself an advantage in the next taku moment — that is Tachimawari.

A player may start from a tachimawari, find a way to score, then return to tachimawari again. Let us call it a “Tachimawari Cycle.” Every PvP game will have one or more systems where players will compete for scores — victory points.

The Tachimawari Cycle
The Tachimawari Cycle

Well-designed countermoves — ones that can create tachimawari — gives clear indications of the difference in skill level between two opponents. And a match between the same level would be a duel of mental strength.

So, how do we achieve this phenomenon achieved in other genres? Let us take a look at examples in trading/collective card games.

Why Countermoves Matter in Card Games

The card games that we often think of — Yu-gi-oh, Magic the Gathering, or Hearthstone — are unlike games of Rock–paper–scissors. They are not zero-sum games based on one element beating another. More often, they are “converting” games that provide resources, and judge players by who is the most efficient in converting the resources to victory points.

What Brings Strategy to Card Games

  1. A large card pool and a big number of possible strategies.
  2. Expanding card pool around existing core cards, creating extended strategies through iteration.

What Brings Countermoves to Card Games

  1. Many possible strategies.
  2. Limited choices (taku) for efficient “conversions”. 
    (Number of cards in hand, cost, conditions to use the cards, etc. all creates limitations.)

You can see how card games to fighting games are different. Fighting games accomplish tachimawari through simplification of strategy. On the other hand, card games emphasize the competition in efficiency between strategies.

Winning Elements in Card Games

The winning elements in card games are:

  1. Top decks
  2. Choosing the deck according to current meta
  3. Calculation and countermoves in a match

It is clear that countermoves are only a part of card games. Usually, card games utilize extra rules to make it seem that there are a lot of countermoves:

  1. Side Decks allows the player to have supplementary cards that counter particular meta.
  2. Limited Formats, such as Sealed Deck, forces players to change strategies between matches, thus bringing a dynamic to countermoves.

Optimizing the countermoves in a card game will make repetitive plays using the same deck more enjoyable, and thus boost the lifecycle of the card pool. You don’t want players to try out all the possible decks in three days and quit because the designer doesn’t have that many resources to keep throwing out new cards.

So, designing for countermoves in card games is adding complexity, in a cost-efficient way.

Designing Countermoves in A Card Game

Let’s look at how to design countermoves following the framework that we established before: Allowing taku, allowing tachimawari, and forming the tachimawari cycle.

i) Allowing Taku

Three essential elements will enable taku:

Accumulating possible actions: As the game goes on, you accumulate cost, cards in hand, and permanents on the field. These are your possible actions.

Ways to keep these actions alive: Not allowing direct attack, separating attack and defense, dividing battlefields, or eliminating a threat from the opponent… these mechanics exist to protect the player from losing their accumulated actions too quickly. See the example of Infinity Wars below.

Infinite Wars

  1. The battlefield is separated into Offense and Defense zones.
  2. The defender chooses which cards to take damage.
  3. There are costly spells to eliminate an enemy card directly.

The Balance between removal and accumulation: Or, how long do you want the player to keep their accumulated actions? If you have more ways to remove an opponent’s cards, then you are leaning towards stronger removal, and vice versa.

Stronger accumulation, medium, and stronger removal
You can tell which side of the spectrum a game is leaning towards by counting how many cards survives on the battlefield throughout the match.

When you establish these elements, if you have a viable core mechanic to transform resource into victory, your game has taku in it.

ii) Allowing Tachimawari

Players need to be able to switch from a weaker taku to a stronger one. They don’t want to be stuck with little choices.

In an ideal world, we would want the attacker to have more but weaker takus, and the defender to have fewer but stronger ones so that they are willing to play aggressively. We can do this through:

Flamestrike: Deal 4 damage to all enemy minions
Simple: the more the enemies, the stronger it is.

Cards that have dynamic power. They can be more powerful (valuable) according to some other elements on the field. A simple 4-cost removal becomes more powerful when there is a 10/10 enemy creature. A Flamestrike becomes more powerful when the opponent has more creatures on the field. There are so many ways to achieve this.

Three cards
Dynamic power cards from Realm of Duels

Separating offense and defense. This is how Infinity Wars works. By distinguishing the offensive zone from the defensive zone, the game made it much easier to manage the balance. — But that’s a path that few has taken. It goes against a player’s intuition, and sometimes against the rule of simplicity.

In the real world, what we see often is one card that combines both offense and defense function. And that would require some tricks.

Destroy target tapped creature.
Killing tapped creature

  1. If the game can distinguish offensive and defensive state, players will have more choices while defending. Magic the Gathering uses “tap” to do this. Whenever a creature attacks, it is “tapped.” And some cards kill a tapped creature.
  2. If you can find a value that pivots the power of cards, you would be able to design cards that have a particular use for both offense and defense.
  3. If you can give the disadvantaged player a stronger taku at match point, you would allow a dramatic comeback. Molten Giant is one of those cards.

cost 1 less for each damage your hero has taken.
It’s so easy to make a comeback that this got nerfed — use to be a 20-cost card.

If players don’t have enough way to obtain stronger takus, you would quickly end up with aggro decks dominating the game. The metagame then stagnates.

iii) Encouraging Tachimawari in Card Games

Now we have cards that would allow players go from weaker takus to stronger ones. But, why should they? What motivates the players to do tachimawari, rather than go for the optimal strategy and win easily?

There are ways that are unique to card games which will encourage players to think before they make decisions.

  1. Response: in Magic the Gathering, you can cast an instant spell or activate an ability while another spell or ability is already cast or activated. Your response will take effect before the other one, creating risk for the offensive.
  2. Hidden information: a card can be played face down. When there is secret information, one can never be sure if victory is at hand.
  3. Error in prediction: Since card games don’t happen in real-time, players have the room to predict the opponent’s move. And they will make mistakes. A lot of times, the thought of “they might have this card” causes more loss than the opponent actually playing that card.

By involving risks, we can make it harder for players to predict the outcome of every move, thus preventing an optimal strategy and encouraging Tachimawari.

iv) The Tachimawari Cycle in Card Games

Let’s sum up what we talked about:

  1. The basis of tachimawari in card games is the competition of efficiency in converting resource to victory.

  2. In a cycle, a player accumulates actions, therefore have more choices (taku).
  3. A player can go from disadvantage to advantage because of well-designed game rules and card power.
  4. The game encourages a player to do tachimawari by involving hidden risk in offensive actions.

And we have the Tachimawari Cycle that’s specific to card games:

Card Game Tachimawari Cycle

Bottom Line

What do we see in card games? There are essential cards, countermove cards, and dynamic power cards. And there is the resource, the hidden information, and the victory point. What else can there be? Can there be multiple ways to convert resources into victory points — or even various kinds of victory points? Can some tools make any weak taku stronger? Can you establish a tachimawari cycle anytime at your wish?

There is still a lot more to explore in this genre and in other PvP genres, too. I hope this article answers the question of what is a countermove, why countermoves work, and how do they apply to popular PvP genres.


[1] This article is derived from research that I have done in cooperation with a handful of industry veterans. Special thanks to Jing Xie, my former colleague, and designer of Realm of Duels, for tutoring me about the basics of PvP gameplay and teaching me about the concepts and terms in fighting games. He provided the framework upon which this post takes form.

Posted on Leave a comment

Dota 2 Update – April 18th 2018

Eador. Masters of the Broken World – Valve

In celebration of its 5 year anniversary, you can add Eador. Masters of the Broken World to your account for FREE starting now until Sunday at 6pm Pacific! Once you add the game, it will remain in your account permanently.

Eador is a universe made of countless shards of land drifting in the Great Nothing. Each of the shards is a little world unto itself, with geography and denizens of its own. The power over the shards is bitterly contested by Masters, the immortal beings mortals believe to be gods. Take the role of the mighty Master and shape the destiny of Eador! It is in your power to deliver the world from ultimate destruction – or to choke it with an iron fist of tyranny. Eador: Masters of the Broken World is a turn-based fantasy strategy game, where the decisions you make affect the world even deeper than the battles you win.

Posted on Leave a comment

Life is Strange dev Dontnod headed toward an IPO

Dontnod, the French game developer behind Life is Strange and the upcoming Vampyr, has registered with the French regulator Autorité des marchés financiers (AMF) and taken the first steps toward going public on the Euronext Growth stock exchange.

A translation of a story from the French publication Le Figaro notes that listing on the exchange and opening up for an IPO should give Dontnod enough of a solid cash boost to continue partially funding its own games alongside investments from publishers, as it has with publishers like Focus Home Interactive for Vampyr.

“By financing a small part of the development ourselves, we take a controlled financial risk, but this allows us to generate larger royalties if successful,” Dontnod CEO Oskar Guilbert tells Le Figaro. “We are not abandoning the classic model of full funding by the publisher. We want to achieve a smart balance between the two models.”

The Dontnod CEO continues, explaining that the company had been eyeing an IPO for some time but that it “had to first reach a certain maturity,” something that he notes has taken a significant amount of work, likely in the wake of some financial trouble during the studio’s earlier years.

“The global video game market, which is already worth $109 billion in 2017, has a strong dynamic and we want to take full advantage of it,” said Guilbert in a translation of a statement published by the AFJV. “The original works of our studio, developed by experienced and passionate collaborators, have enabled us to appeal to the world’s largest publishers and a very wide audience, and we intend to continue this momentum through this IPO project.”

Posted on Leave a comment

Blog: Breaking down the technologies of Civilization

The following blog post, unless otherwise noted, was written by a member of Gamasutra’s community.
The thoughts and opinions expressed are those of the writer and not Gamasutra or its parent company.


The tech tree quietly does a lot to reinforce the major themes of the game and to provide the heartbeat that keeps players moving through the game.

Constant Progression

I think the most important thing that it does is ensure constant progress for the player. Civ greatly benefits from always giving the player something they want within a couple of turns. This is a large part of what feeds the one-more-turn compulsion of the games.

It serves much of the same purpose of leveling in an RPG, albeit flavored strongly by the rest of Civ. Like RPG levels, it gives you a major in-game reward on a regular basis and helps make the player feel like they’ve grown over time.

It really forces the game to feel like it’s moving forward. It introduces new pieces regularly and so keeps the game from feeling like it’s stuck in a rut. No matter what else is going on in the game, there’s always a new tech in a few turns and most of those techs have something interesting attached to them.

Meaningful Rewards

This progression is strongly reinforced by the way that the techs tend to have something meaningful attached to each of them. There are a few that I don’t care about, especially ones with non-essential military units, but mostly they have something that I want.

The game as a whole is good about making everything meaningful. It encourages you greatly to have a variety of all of the pieces, such as improvements, units and districts. This gives the game a lot of pieces to give you and so it can distribute them across the tech tree and keep the techs valuable.

This isn’t completely perfect though. Things like aircraft upgrades just don’t have enough of an impact in a normal game for me.

The Tree

The other major theme that it does a lot to reinforce is that of progression through history. It’s the feeling of having started as just a single settler and a warrior and growing to be a massive, advanced empire. The abilities to trace technological advancement and the idea of getting to such things as spaceflight from such techs as animal husbandry and archery are very satisfying.

Also, having techs unlock other techs adds value to early techs. Now, not only are they valuable for what they give, but also for what they unlock. The tension between a tech that has more immediate benefit and one that will let you get to a more powerful tech sooner is also fun for players to navigate.

Asymmetry of Tech Options

Despite my above point of the meaningfulness of the techs, there are some that feel much more powerful than others, and those act as guides for decisions that have no clear answer. In this way, they greatly reduce the effective complexity of the decision of what to research for mid-tier players.

This is a little bit of an involved point, so I’m going to expand on it a little. Essentially, deciding which tech to research next is an individual skill that players develop as they play the game. When this skill is not fully developed (and it may be unreasonably difficult to ever fully develop this skill), there are going to be situations in which the player cannot easily decide between a number of options for which tech to research next, when those techs are judged simply by their immediate value. The existence of powerful techs further down the tree makes these decision points more approachable as they give the player a direction in which to research.

Tech And Units

A sub-point here is that techs which result in a major difference in unit strengths are one of those asymmetries and an interesting one to unpack. First of all, the one in SM:AC is one of my favorite impact moments in video games because when you jump from two attack to four attack, but the defense remains the same, that’s a moment that highly incentivizes attacking while you retain the technological edge. Civ6 also does a good job of making units of different tech levels differ greatly in power, so your crossbowmen one-shot the opposing archers.

That’s very satisfying because it’s the conversion of a scientific advantage into a military one and it’s fun to be in unequal battles that you’ve earned. Also, it is the realization of the fantasy of winning a war due to superior technology, which is a familiar narrative, albeit one I believe to be heavily misapplied.

Complexity

The big issue with this tech tree system is the degree of front-loaded complexity that comes with it. It’s very hard for a beginning player to make a decision of what tech to research, even when only considering the choice menu and not the full tree. There’s a lot of content that requires familiarity from the player and the choice of which tech to research requires a lot of comfort with the rest of the game as well. Complicating this with the tree structure exacerbates the issue. To some degree, the depth of the tree makes this unavoidable however.

Essentially, just having the knowledge required to make a decision is a major barrier to entry for new players. Things like the advisors are meant to help with that, but it’s probably insufficient. To some extent now, new Civ games are meant for people who have played earlier Civ games and so it can be hard for players who don’t have that experience.

Staticness

An interesting point to think about is that the tech tree is static between games. The techs are in the same positions on the tree, cost the same amount and unlock the same pieces every time. This is probably in large part because it’s much easier to design and implement that way, especially when you have to take historical pseudo-realism into account, but the results of that decision are worth examining.

You can ensure consistency in the above points through this static design. A good procedural solution could do that as well, but it is again substantially easier to manage with a hand-crafted tree.

This drastically reduces the complexity of the tech tree on later playthroughs. One of the skills of the game is simply getting better at traversing the tech tree. This naturally also leads to players no longer engaging with the decision point of which tech to research next because they feel they have a good, solved path. The rest of the game does influence the decision, but the raw power of some techs is often enough to overwhelm those wrinkles for some players.

History

Tying the tech tree to historical technologies greatly reduces the complexity. Logical connections between the pieces make them much, much easier to understand and historical context helps the player get an idea of what the techs do.

Also, it pushes a compelling fantasy of how technological research works. Having technologies that lead to other technologies and being able to trace that development is intrinsically very satisfying due to what research is in the real world.

Tech Boosts

This is my favorite feature from Civ6 just for all that it does. As you may have seen by now, I love things that tie different parts of the game together and having your actions in the world have such a direct impact on research is wonderful. I love how this plays and how it generates sub-quests which encourage trade-offs. I really enjoy when the game pushes you to go kill something with your slinger so that you can boost archery for instance.

Also, the flavor of the piece is wonderful. It makes sense that slingers would look to improve their weapons. It makes sense that having a number of musketmen would spur replaceable parts. The whole thing is just something that feels like it makes sense.

Civics

Civ6 does something very elegant in making the techs and civics use the same structure. However, these two things need to feel different somehow, and if it isn’t in the mechanical structure, then it needs to be somewhere else. Civ6 handles this with making the reward for civics mostly be policy cards. This certainly functions effectively as a way to differentiate the two trees, but it is both difficult to understand the ramifications of policy changes, and the system feels divorced from the rest of the game.

By and large, it manages all of the above points, but the rewards don’t feel as meaningful as in the tech tree, and so it doesn’t do the same job of feeding the one-more-turn compulsion that the tech tree does. However, it does not exist in isolation, and so does not need to.

Alternate Approaches

The first alternative to examine is that of Beyond Earth, but the visual difference merely hides an equivalent system. There are minor differences in terms of branching, but despite the difference in appearance, the system is functionally identical to that of the other Civ games and so has the same benefits and weaknesses.

On the other hand, this lack of a tech tree is one of the biggest weaknesses of Colonization. I feel like one of the major issues with that game is that there often comes a point where you feel like you’ve gotten into a bit of a rut. Essentially, the next real goal is quite a distance away and so the one-more-turn compulsion ends up lost and the game founders.

The Stellaris design of giving the player a choice between a group of semi-random options from an underlying tech tree might on first glance feel very similar to the Civ one, but the presentation and randomness detract from the points above about long-term planning and so the feature is much weaker. Additionally, the techs do not all feel meaningful (at least, as of before the major change to combat from a few months ago, when I last played). However, the complexity of this decision for new players is thus also greatly reduced for new players as the decision becomes which one of the limited options is most immediately beneficial. For advanced players, it reduces to something much closer to that of Civ, but not quite there.

Polytopia has the major difference of requiring the same currency for research as for everything else. This completely removes the major point of constant progression detailed above. However, each game of Polytopia is both short and dense and so it can stand to lose that piece.

Sorceror King and Master of Magic have a more implicit tech tree in that buildings unlock other buildings. This again loses the constant progression that is a critical benefit of the tech tree and so the games end up with the same problem as Colonization where there comes a point without a reward that is both compelling and reasonably near and the game starts to sputter.

Conclusion

The Civ games have long been known for that one-more-turn compulsion, and I think that the tech tree is the driving force behind that. It’s quite safe to say that without this, Civ would not be anything like the game that it is.

Posted on Leave a comment

Blog: Creating a hard AI for Terra Mystica

The following blog post, unless otherwise noted, was written by a member of Gamasutra’s community.
The thoughts and opinions expressed are those of the writer and not Gamasutra or its parent company.


Spoiler alert: This doesn’t have a happy ending.  Digidiced has been hard at work for more than a year trying to produce a Hard version of its AI for Terra Mystica using machine learning.  Our results have been a lot less impressive than we were hoping for.  This article will describe a little bit about what we’ve tried and why it hasn’t worked for us.

If you’ve paid attention to the latest developments in AI, you’ve probably heard of AlphaGo and AlphaZero, developed by Google’s DeepMind.  In 2017, AlphaGo defeated Ke Jie, the #1 ranked Go player in the world.  AlphaGo was developed by using a massive neural network and feeding it hundreds of thousands of professional games.  From those games, it learned to predict what it thought a professional would play.  AlphaGo then went on to play millions of games against itself, gradually improving its evaluation function little by little until it became a superhuman monster, better than any human player.  The defeat of a human professional was thought to be decades away for a game as complex as Go.  But AlphaGo shocked everyone with its quantum leap in playing strength.  AlphaGo was able to come up with new strategies, some of which were described as “god-like.”

But it didn’t stop there.  In December of 2017, DeepMind introduced AlphaZero – a method that also learned the game of Go, but this time didn’t use any human-played games.  It learned entirely from self-play, being only told the rules of the game.  It was not given any suggestions or strategies on how to play.  AlphaZero not only able to learn from self-play alone, it was able to get stronger than the original AlphaGo.  And on top of that, the same methodologies were used for Chess and Shogi and the DeepMind team showed results that AlphaZero was able to solidly beat the top existing AI players in both of these games (which were already better than humans).  Since these results have come out, there has been some criticism around if the testing conditions were really fair to the existing AI programs, so there is a little debate as to whether AlphaZero is actually stronger, but it is an outstanding achievement nonetheless. 

It also became quite clear that AlphaZero approached chess differently than Stockfish (the existing AI they competed against).  While Stockfish examined 70 million positions per second, AlphaZero only examined 80,000.  But AlphaZero was able to pack a lot more positional and strategic evaluation into each of those positions.  By examining the games that AlphaZero played against Stockfish it became obvious to a lot of people that AlphaZero was much better at positioning its pieces and relied less on having a material advantage.  In many cases AlphaZero would sacrifice material in order to get a better position, which it later used to come back and secure a win.  It suggested the possibility that there might be a resurgence in chess programming ideas, which had been stagnating in recent years.

 

 

 

 

 

 

 

The DeepMind team was able to show that AlphaZero learned many human-discovered opening moves.  They showed several examples of how different openings gained and lost popularity as it continued to learn.

As Digidiced’s AI developer, these were exciting developments for me.  I’ve had experience with machine learning and neural networks before and have been playing around with them for many years.  I once developed a network as a private commission for a professional poker player that could play triple draw low at a professional level.  I began to wonder if I could use some of these same techniques for Digidiced’s Terra Mystica app.  One of the compelling features of AlphaGo was that it was largely based on something called a convolutional neural network (CNN).  A CNN is also used in other deep learning applications like image recognition and is good at identifying positional relationships between objects.  AlphaGo was able to use this structure to identify patterns on the Go board and determine the complex relationships that could be formed from the different permutations of stones.

While Terra Mystica takes place on a hex-based map instead of a square grid, a CNN can still be applied to it so that the proximity of players’ buildings can be incorporated, which is a critical part of TM strategy.  However, there are several things that make TM a much more complicated game than Go.

  • TM can have anywhere from 2 to 5 players, although it is often played with exactly 4.  For programming AI, the leap from 2 players to more than 2 is actually a lot more difficult than most people realize.  You may have noticed that whenever you hear about an AI reaching superhuman performance, it’s almost always in a 2-player game.
  • While a spot on a Go board can only have 3 states (white stone, black stone, or empty), a hex on a TM map can have 55 different states, taking into account the different terrain types and buildings.  Add things in like towns and bridges and the complexity goes up from there.
  • TM has 20 different factions using the Fire & Ice expansion, and each one of these factions has different special abilities and plays differently.
  • TM has numerous elements that occur off the map including the resources and economies of each player, positioning on the cult tracks, and shared power actions.
  • Each game is different by adding scoring elements and bonus scrolls that are different with each game.  Which elements are present in the particular game can have a massive effect on all of the player’s strategies.  Not to diminish the complexity of Go (a game which I’m still in awe of after casually studying it for over a decade), but you’re always playing the same game.

One of the things that makes TM such a great game and causes it to have a very high skill ceiling is the fact that its economies and player interactions are so tightly interwoven.  The correct action to take on the map can be highly dependent on not only your own situation, but the economic states of your opponents or the selection of available power actions.  All of this makes TM orders of magnitude more complex of a game than Go.


Chaos Magicians, Swarmlings, Darklings, and Dwarves fight it out on the digital version of Terra Mystica.  Complexities abound and an AI needs to know how to read the board.  Darklings will want to upgrade one of their dwellings to get the town bonus. They should upgrade next to the Dwarves to keep power away from the stronger CM player. The choice of towns could affect the flow of the rest of the game:

  • Should they take 7VP & 2 workers so they have enough workers to build a temple and grab a critical favor tile?
  • Or 9VP & 1 priest that they can use to terraform a hex or send to the cults?
  • Or 8VP & free cult advancements which will gain them power and cult positioning?
  • 5VP & 6 coins is sometimes good, but probably not in this situation since the Darklings have other income sources.

  The other town choices seem inferior at this point, which the AI needs to recognize.  Notice what is needed to plan a good turn – the recognition that a town needs to be created this turn, the optimal location of the upgraded building, the knowledge that a critical favor tile exists and how to get it, the relative value of terraforming compared to other actions, the value of cult positioning (not shown) & power, as well as the value of coins which depend on how many coin-producing bonus scrolls are in the game.


The main idea behind training the network to become stronger is called bootstrapping.  I’m simplifying things a bit here, but think of the neural network as an enormously complicated evaluation function.  You feed it all the information about the map, the resources of all the players, and other variables that describe the current game state.  It crunches the numbers and spits out an estimate of the best action to take (each action is given as a percent chance that it is the best action) and an estimate of the final scores for each player.  Let’s say you have a partially trained network that has an okay evaluation function, but not that good.  You now use that, and each time you’re going to make a move you think 2 moves ahead, considering all the options and picking what you think is best.  You’ll now have a (moderately) more informed estimate of your current state because you’ve searched 2 moves ahead.  You now try to tweak that model so that your initial estimate is more similar to your 2-moves-ahead estimate.  If you were able to fully incorporate everything from 2 moves ahead into your evaluation function, when you use this function to search 2 moves ahead, it’s equivalent to searching 4 moves ahead with your original function.  It’s not that simple, but you can see how repeating this over and over again will keep improving the model as long as it has enough features to handle the complexity.  You just have to repeat it billions of times…

In order to train its networks, DeepMind was able to utilize a massive amount of hardware.  According to an Inc.com article, the hardware used to develop AlphaZero would cost about $25 million.  There is no way that a small company like ours would be able to compete with that.  Some people have estimated that if you were to try and replicate the training done on a single machine, it would take 1,700 years!  Even after all the training, when AlphaGo is run on a single machine, it still uses very sophisticated hardware, running dozens of processing threads simultaneously.  We needed to create an AI that was capable of running on your phone.  For each single position that AlphaGo analyzes, its neural network needs to do almost 20 billion operations.  We were hoping to have a network with less than 20 million.  And instead of analyzing 80,000 positions per second, we would be lucky if we could do 10.  We also considered an even smaller network that could look at more positions per second, but it would not have enough complexity to incorporate a lot of the nuances needed for a strong player.

So our goal was to create an AI for a game that was even more complicated than Go, using a network about a thousandth the size.  AlphaZero was able to play over 20 million self-play games in order to help its development.  Even renting several virtual machines and playing games 24/7 for a few months, Digidiced was only able to collect about 40,000 self-play games.  Despite these limitations, we were cautiously optimistic.  We didn’t need super-human and god-like play.  We wanted something that could be a challenge to the entire player base while not taking too long to think for each move.

A tiny peek into the complexity of Alpha Go (from David Foster’s AlphaGo Zero Cheat Sheet).

But even that turned out to be too much of a challenge with our limited budget.  The AlphaZero paper claimed that starting from scratch and completely random play yielded better results than mimicking games played by humans.  We decided to try both methods in parallel: one network would start from random play and build up network sophistication over time while another network was trained on games played on the app.  Neither was able to create a very strong player; in fact, we were never able to create a version that could outperform our Easy version that used fairly standard Monte Carlo Tree Search.  We even tried focusing the development on only 4-player games, but this didn’t help much.

What was really heartbreaking was that we could see the improvement that the network was making.  We could see the improvement over time.  But the rate of improvement was just too slow for the amount of money we were spending.  It was a very difficult decision, but we’ve decided that we’re going to halt development work on this for now.  We still see a possibility of spending some time converting the played games from Juho Snellman’s online implementation of TM, but we don’t have the funds for that now.  Juho had very kindly given us permission to do that much earlier, but the conversion proved to be very difficult for a number of reasons, mostly due to how the platforms differed in accepting power.  So while there is still a chance of further development, we don’t want to promise anything that doesn’t seem likely.

Posted on Leave a comment

Game design portfolio-building tips from a Creative Assembly vet

In this helpful article, Creative Assembly senior designer Simon Mann describes ways to think like a designer and help build a strong game design portfolio to get yourself noticed.

Games design is one of the most varied and exciting roles in the games industry today.

Designers work closely with all other disciplines to deliver the project vision. The discipline encompasses everything from psychology to art, sound to architecture, data analytics to programming and beyond.

However, in an industry where fewer companies have positions for junior designers, and more and more applicants, how do you make yourself stand out in an oversubscribed field?

In this article, I’ll describe ways to think like a designer and help build a strong portfolio to get yourself noticed.

Play Games

Play as many games as you can. Don’t stick to one genre and don’t just play the 80+ Metacritic AAA titles (but you should play them as well). Playing games lets you know your target audience, its trends and fashions, as well as giving you great ideas for your own projects. While playing games, pause for a second, look at what the game is doing. Why did they put that feature in? How is it trying to make you feel? How is it achieving (or not) its intended goals? Could it be improved? If you feel you’ve discovered an interesting theme or idea, share with others or even write articles on the web.

Make Games

Make Games- Small games, big games, ideas, prototypes or card games. Every game you make gives you valuable experience about the design process, as well as developing your ancillary skills. Try, where possible, to work with others on group projects or game jams; good teamwork skills are invaluable to a designer. Making as many games as possible also shows that you’re willing to put in the hard work required to succeed in the games industry, as well as implementing the cool ideas you’ve come across.

“All our best ideas generally come from the most unexpected sources. Give yourself a healthy work/life balance – travel, meet people, go to events.”

Design Always

Games design doesn’t occur in front of a computer. All our best ideas generally come from the most unexpected sources. Give yourself a healthy work/life balance – travel, meet people, go to events… Do anything you want, but always be thinking like a designer, on the lookout for the next great inspiration. Carry a notebook or make notes on your phone, so when the creativity hits, you can get it on paper as quickly as possible.

Strong CV

At the heart of any good design portfolio is a strong CV. Show where you’ve come from, the things you’ve done to further your career and give recruiters a strong sense of where you’re going to fit into their team. Keep your CV to one page, with only relevant information to the role you’re applying for. If your CV isn’t concise, recruiters won’t even look at your portfolio.

When I started looking for jobs in the industry, I shotgunned my CV indiscriminately to hundreds of companies, and funnily enough, never heard back. Later in life I realized that I was more desperate to get on the ladder than think about the roles I was applying for, so my CV was ignored out of hand. When applying, always consider: why does this company need you? Ask yourself why you’re applying for that role and how you feel you can improve it with your presence. This will also come in useful for the interview stages.

Have a portfolio website

While carrying a large folder with reams of paper in it may seem enticing to some, we’re in the 21st Century now. Set up a website for yourself, fill it with your work, blog posts, pictures, text and playable demos. Recent studies have found that you only have a few seconds to create a lasting impression. Let your site lead with your personality, have a strong front-page which really sets the tone for the rest of your portfolio.

Always take pride and enjoy your creations, even when making a portfolio. It’s easy to spot the copy-paste versus the lovingly-handcrafted content, even in static media. I’m afraid there are no shortcuts for design, you have to put in many many hours of hard graft, and if you’re not willing to put it into selling yourself, then you won’t be able to do it later down the line.

Content

Be prepared to show off all your skills! Avoid demo reels on your front page where possible and make project subsections with pictures and gameplay videos. An organized site is an organized mine and will make it easier for recruiters to get all of the relevant information.

Varied content on a portfolio is a great thing to show off your skills. However, while you may have a lot to talk about, be concise. No recruiter will read a 40-page Design Document and they will not be impressed by drawings on napkins. Everything in your portfolio should show off your best work in the shortest time.

Show it Off

Show your portfolio off to as many people as possible. Do not just show it to your friends and family. Even with the best will in the world, they will rarely give you unbiased feedback. Show it to your peers, recruiters and anyone you can. When (not if) you do get poor feedback, don’t just react defensively, think about why the feedback was given and how you could fix it. This is a very important skill for any designer as honest feedback is a huge part of the design process and you have to be a little thick-skinned about it. You may think something is amazing, or satisfactory, but perhaps you forgot about your audience? If you are unsuccessful in an interview or application process, never be afraid to ask for feedback, it’s how we learn and grow.

And breathe. This isn’t an exhaustive list of dos and don’ts but certainly the ones I prescribe to. I’ve been in the position of being an unemployed graduate, desperate for anything games related job-wise, but games isn’t a job, it’s a career, and a passion. Patience is a virtue, and while it may take a while, if you truly love making games, and have the skills, you’ll get the break.

Posted on Leave a comment

Sponsored: Join this free live webinar on cross-platform multiplayer game dev!

Presented by AWS and Amazon GameLift

Supporting cross-platform play for multiplayer games gives your players the most competitive matches and allows them to play with friends regardless of console or device they play on. As a game developer, how do you architect your backend and matchmaking to let everyone play together?

Join AWS and Amazon GameLift and learn how to support cross-platform play for your multiplayer game. Bruce Brown and Peter Chapman will explain design patterns and backend architecture that allow you to accept game requests from multiple gaming devices, match players into games, and route player groups to suitable game servers.

Bruce Brown
Software Development Manager
Amazon GameLift

Bruce Brown is the Software Development Manager of the Amazon GameLift Player Experience team responsible for building the FlexMatch matchmaking service. He has been in the software industry for 12 years including time spent at Microsoft on the Xbox Live Cloud Compute and Xbox Multiplayer teams and at Riot Games working on the League of Legends Live Gameplay and Personalization teams. Currently, he is playing Zelda: Breath of the Wild with his 6 year old son, and he enjoys designing games with his wife in his free time.

Peter Chapman
Solutions Architect
Amazon

Peter Chapman is a Solutions Architect in the Amazon GameLift and Lumberyard teams. He has over 13 years of software development and architecture experience. He has designed solutions in many industry sectors including Retail, Healthcare, and Gaming. Currently, he is having a great time playing Axiom Verge on his Switch.

Moderator: Kris Graft
Editor-in-Chief and Publisher
Gamasutra

Kris Graft is editor-in-chief and publisher of Gamasutra, the leading site dedicated to the art and business of making games. He has worked as a full-time game journalist since 2005, specializing in the business and creative aspects of the industry.