Sid Meier famously said that “a game is a series of interesting choices”
This is a very “dense” statement, that is loaded with meaning and implications. Much game design theory has been written just to unpack it, with every word requiring it’s own articles to explore the full extent of it’s implications. Entire schools of thought have sprung from this statement, as I will attempt to spring from it as well.
But putting all this aside, this is a statement that I keep coming back to. It captures perfectly the kind of value I seek in games, specifically strategy games. And it raises the fundamental question of strategy game design, how does one create an environment which leads to the player making a series of interesting choices?
Traditionally, contemporary wargames have been relying on randomness to create interesting choices – cards, dicerolls, RNG. After all, if you can’t rely on the outcome of an action, it does make one consider the entire possibility tree – what if a shot whifs, and now my little army dude is all out in the open? How much can I afford to let that happen, and how much can I afford to not take the shot, or take a safer shot? This does lead to design which is chaotic and heavily random. But the most fascinating thing is how even among the noise and randomness, good plays still exist. Consider X-Com, both the new and the older one. In the old one especially, one could land on a mission map, lead one’s squad out of the ship, and then some aliens randomly show up from the fog and get a bunch of lucky headshots, and drop half your troops. But at the same time, a good X-Com player can beat the campaign on the hardest difficulty every time by making the right calls.
On the other end of the spectrum, ancient wargames such as Go, Shogi and Chess have no randomness whatsoever. It seems like it would be possible to just calculate all the moves in advance and there would be no interesting choices left – just the correct ones. But this is too complex a task even for the most advanced computers! Instead, players and world champion beating AI must rely on guidelines, heuristics, intuition and pattern recognition – all qualities which make decisions interesting, their effects ambiguous.
Even the most random games, to avoid being just gambling machines, need some form of consistency, and even the most deterministic games, to avoid being just math problems, require ambiguity. From this, we can derive a principle that an interesting decision lies somewhere between a guess and a solution.
A very simple way to make for a hard choice is to make the player choose between things which are not comparable. “Comparing apples and oranges” comes to mind as an idiom that reflects the futility of such an attempt – an impossible comparison, not just hard one. But does this saying really mean what people think it means?
Imagine for a second a farmer that wants to plant an orchard. What does he plant? Apple trees? Orange trees? A mix of both perhaps? What proportion of a mix? There are a lot of answers to this question. Some of them are better than others. If one is to plant orange trees in a snowy, northern climate, one is going to have a bad time. And if some options are worse than others, then other options are better.
Suppose this farmer wants to maximize his profits. Now we’re talking! After all, with an objective measurement, we can compare just about anything. But it’s still not an easy comparison. There are still a lot of factors, and the market is so complex, it’s always ambiguous which option is better. If we were to model this question in an economic game, it could result in some very interesting choices – as economic simulators tend to do.
But suppose we make our Farming Simulator, and someone plays it once. They have made some very difficult choices! Perhaps actually our player has made her choices randomly, since this is her first time playing, so she’s just figuring out how it all works. The second play, the choices are a little less random, a bit more concrete. After ten games, our player may even figure out a strategy she likes. Even if we make the set up random to avoid an exact solution, general tendencies may form. “If I spawn in California, I’ll plant oranges” – a flowchart with lots of “IF” statements might be forming.
Now lets have a thousand players play a thousand games. Let’s add multiplayer, where farmers compete to turn the most profit or establish a monopoly. Over time, the game community will move the choices in the game from just random guessing into the territory of more solid solutions. But we want to avoid either extreme, we want the choices in the game to be somewhere in the middle! How do we accomplish that?
Case study: Age of Empires 2
Age of Empires 2 is a really marvelous example of a sturdy strategy game design. The last official expansion by Ensemble Studious came out in the year 2000. The next unofficial official expansion came out in 2013. In between that time, no patches or changes were released. But many high level players remained dedicated to the Age of Conquerors expansion.
Now, one would think that in that time frame, the player community would be able to figure out what the best civs are, what the best build orders and strongest units are, which choices in the game are purely situational or simply are not good enough and should be avoided. That the games would be decided by whoever can execute the build faster or better, and have better unit micro.
But that was not the case. Every year when tournaments were held, the meta was always fresh, always shifting. There were always different “good” and “bad” civilizations, and the players would be opting for vastly different build orders every year. This is with 0 actual balance changes to the game, for 13+ years. How did Ensemble Studious manage it? Let’s take a closer look!
On the surface it seems like it should be easy to figure out what the best unit is. On paper, that’s the Paladin. So whatever civ gets the strongest bonus to them or to getting them, that should be the strongest civ – right? Well, not quite. The weakest military unit in the game, the humble spearline (spearman-pikeman-halberdier) actually has a hidden bonus to attacking cavalry units – so a group of halberdiers can make short work of paladins. Even a spearman or a pikeman would make very good trades with paladins, being very cheap, and paladins being very expensive. And spearmen of course get demolished by archers, militia-line units, and onagers – which in turn get beaten by the knight and scout line units (paladins are the final upgrade to knights).
So a solid loop of counters is formed, much like a game of Rock Paper Scissors. But is it just it, making games about RPS? There’s undeniably skill involved with playing RPS, but Rock Paper Scissors isn’t exactly tantalizing to strategy game enthusiasts. And of course, there’s more to it.
Consider for example the role of the units in Age of Empires. It has three, cheap hard counter units, the pike line, the skirmisher line and the scout line. Collectively, they are called “trash” units since they don’t cost gold. But there’s reasons to build some of these units independently of just countering your opponent. The scouts for example are often used for very early aggression due to their speed. Whereas pikemen, since they can’t catch up to cavalry, they’re more a defensive unit, more of a deterrent. Sometimes players build siege and pikes just to anticipate the enemy counter! And the archers and skirmishers, players use them for their low food cost when they want to tech up and boom their economy more.
So not only do the units counter one another, they align with one of the axis of a Counter Triangle like the one above.
What is important about this relationship is that it disrupts the RPS pattern by giving the players incentives to commit to specific sides of the triangle, but also by giving each side an identity and a niche, preventing the breakdown of comparisons.
This relationship doesn’t just exist in RTSes. Shooters don’t always have an economy, but they have a counter triangle. Shotguns excel at short range and a shotgun rusher can easily clear a nest of snipers. Assault rifles excel at middle range and can take care of shotgunners before they get too close. And sniper rifles excel at long range engagements, outranging assault rifles. Shotguns are the aggressive rushdown option here, snipers are defensive, and assault rifles have the potential to be flexible, either way, without excelling in either situation.
This relationship also exists in Magic The Gathering to some extent – there is a Beatdown deck, and a Control deck and it’s imperative to know your role in a match up. This relationship exists in Fighting Games, where blocks stop attacks, but throws overcome blocks and allow the throwing player to set up an advantageous situation for themselves. In fact, I would argue that this relationship exists in any stable, aged strategy game, since the triangle creates real, stable incomparables which avoids solution. But how do we describe this relationship?
A hard, interesting decision always means choosing from options which represent two different extremes. A good strategy game is about balancing those extremes. The curious thing though is that even though there are many different styles of decision-making, they all happen along a single axis that has a broad abstract form. On one end,of the scale, there are actions that are akin to advancing, being aggressive, rushing and taking risks – doing everything to win. On the other end, one must also retreat, be passive, stall and avoid risks – do everything in order to avoid losing. The exact nature of those extremes is not so straightforward, and can get fairly abstract – so I prefer to use the term Jing, a loan word from eastern martial arts which does not have much baggage. Positive (+Jing) for advancing, and Negative (-Jing) for retreating.
Sun Tzu said: in all fighting, the direct method may be used for joining battle, but indirect methods will be needed in order to secure victory. Indirect tactics, efficiently applied, are inexhaustible as Heaven and Earth, unending as the flow of rivers and streams; like the sun and moon, they end but to begin anew; like the four seasons, they pass away to return once more.
There are not more than five musical notes, yet the combinations of these five give rise to more melodies than can ever be heard.
There are not more than five primary colors (blue, yellow, red, white, and black), yet in combination they produce more hues than can ever been seen.
There are not more than five cardinal tastes (sour, acrid, salt, sweet, bitter), yet combinations of them yield more flavors than can ever be tasted.
In battle, there are not more than two methods of attack–the direct and the indirect; yet these two in combination give rise to an endless series of maneuvers.
The direct and the indirect lead on to each other in turn.
It is like moving in a circle–you never come to an end.
Who can exhaust the possibilities of their combination?
Jing is also tempo control. +Jing shortens the game, -Jing stalls it, and Neutral (=Jing) Jing allows the game to go at it’s natural or average pace, developing the player’s potential for a much strong +Jing or -Jing strategy. Using Age of Empires example from above we see that
Rushing is +Jing. If your rush succeeds, the enemy dies quickly and you win. If it fails, your economy will be crippled and the enemy will be able to counter attack to quickly end the game. Either way, when someone commits to a pure rush build, the match would be over very quickly.
Economy and booming is =Jing. As you expand, you will naturally drain the resources and have to fight for contested expansion sites. Techning up and gathering resources creates the potential for a much stronger attack, or to defend quickly if an attack is coming.
Turtling is -Jing. If both players just sit in their base and turtle, the game can just go on indefinitely!
However, most competitive players alternate between them and rarely go for an all-out rush, preferring to soften the economic self-blow with early aggression and timed push builds, and few build heavy defense unless they scout the early rush coming.
The Jing Line
While the Jing exists as two extremes, the actual options available when making a decision are rarely dedicated 100% to either side. The options can be said to exist along a Jing Line based on how positively or negatively charged it is. Displaying the options along the Jing Line allows us to evaluate how much each option occupies a niche. Options that occupy a unique niche have more of an identity, and the more niches there are, the harder it is to break down the options into a simple numeric comparison. On the other hand, when several options compete for the same niche, one of the options ends up just being straight up better.
Height and Breadth
Height is a measure of how much an option is aligned to a particular Jing charge. An all in rush is a Tall +Jing move. Breadth is a measure of how much area between the positive and negative charges does a decision cover. For example, harassing in RTSes is a Broad+-Jing move, since it is both a form of aggression, but also the purpose is to develop your economy while the enemy deals with the pressure. The taller Jing strategies must always beat out the shorter ones, but the taller Jing options should require more of a commitment, in order to create room for the other player to maneuver. At the same time, shorter Jing options should be broader, and cover more charges at the same time. No tall action should be allowed to have more charges than the shorter options.
So in Starcraft, a group of Mutalisks for Zerg players is usually a Broad +- Jing move. One of the responses is building a force of dedicated anti-air units – say, Valkyries, which are a Tall -Jing move. They, being a hard AA counter, can’t be used to attack the enemy base, and can only be used to deny air – a unit a player has to commit resources to, and wants to avoid doing that.
Width is how far apart the options are on the Jing Line, how unlike one another, and thus harder to compare they are. There are many ways to make options unlike one another, but the easiest rule of thumb is to avoid numbers. Finding numerical values for different options is already how we solve games (ex. the piece value in chess used to determine material advantage), so the harder it is to compare the options numerically, the more varied situations one has to consider, and this generally makes for harder and interesting decisions.
A way to compare how far apart options are is to compare their numbers as factors of one another. Let’s compare the Pump Shotgun and the AK in Counter-Strike. The shotgun carries 8 bullets in the chamber before reloading, the AK carries 30. The factor of difference is 3.75 (x3.75 width). However, comparing the Galil (35) and the AK’s (30) clip capacities the difference is just x1.16 width. This can be done for all the possible statistics for the guns, and if you do, you can see that the Shotgun and the AK are much wider apart than the AK and the Galil.
What if the characteristics of different options cannot be compared numerically? For example, in Chess, the Knight possesses the unique ability to hop over units, the Rook can castle with the King, and the Pawn can be promoted to other pieces, the ability can be represented by a 1 for having it, and 0 for not having it. Since dividing by 0 results in an indeterminable factor, such options are the furthest apart on the Jing scale. The more such incomparable (also called Boolean or Binary) characteristics exist, the wider the overall Jing scale is.
The most important metric is the spacing. So long as the options are evenly spaced along the Jing Line, they can keep their distinct feeling. However if the options are gathered in clusters, they are seen as variants of the same option instead of being unique options. Such clusters are “weak links” in terms of mastering a game, and it’s much easier to determine which of the variants is best for which situation.
Going back to the example of Counter-Strike, the AK and the Shotgun being more distinct are much harder to compare, they represent a more interesting decision as to which one to buy, even though the Shotgun costs less than the AK and by that logic the AK should be an obvious superior choice. AK is regarded as the more superior version of Galil makes Galil only an option to consider when one is low on money. Even if Galil and AK were more distinct, their close proximity would still lead most players to consider buying an Assault Rifle vs a Shotgun, rather than considering buying specific assault rifles.
In chess there is a similar situation with promoting a pawn. One can promote the pawn to any of the backline “noble” pieces – the queen, the rook, the bishop, the knight – but most people think that you can only promote to the queen because the queen, combining the powers of the rook and the bishop, makes the other two choices obsolete except in rare edge cases. The knight however is still picked occasionally, due to it’s unique tile-hopping ability. So in essence, the pawn promotion comes down to two real choices most of the time – Queen or Knight.
In both situations, clumping creates false choices. A good way to work out clumps is to increase the width between the clumped options, by introducing harder to compare statistics to one, the other or both. In Counter-Strike 1.6, the two assault rifle options are the AK and SG553 – a scoped assault rifle. Since having a scope is a binary characteristic, comparing the AK and SG553 is not so easy, and both see a fair amount of play.
However, increasing the width is not the only way to reduce clumping – even if an option is further apart from the other options, widening the other options may not be such a good idea for the style of the game. Consider Starcraft, Warcraft III and DotA. In SC, the food cap is 200, in WC3 it is 100 with incentives to go below that and in DotA you control only one hero. The units in SC are also more easily comparable using numbers, in WC3 most units have some sort of a special ability they can use, plus there are heroes with really unique abilities. If you had 200 food cap army in Warcraft3, managing all those more complex units as well as the hero would prove to be too much. Most serious decisions in WC3 are in regard to hero selection, leveling and control, and everything else is secondary. This is even more apparent in DotA, where making decisions using a single hero is the entire game. If the players had an army, it would be mostly neglected for the sake of the hero, creating many false choices. On the other hand, SC’s simpler units create width because there is a lot of them. That’s not to say that having more options in a game necessarily creates wide, deep decisions – after all, it took Blizzard many years of patching and balancing and reworking to make sure that every unit has a niche to fill and a reason to be used. What is important is the overall style of the decisions, and having options that are consistent with that style.
The impact of every choice is another important metric. Impact here means “how long will this decision remain a factor”, and can be expressed as a percentage from 0 to 100. 0% means “this decision will not be a factor the future at all”. Options with 0 impact are false choices though such things are rare. It can be said that even non-game options, such as screen resolutions, the colour of one’s units or music volume have some minuscule impact. However, for the sake of simplicity we will round that to 0. Some other forms of 0% impact choices are QuickTime Events and other pure types of dexterity/reflex tests. If in Guitar Hero you miss a beat, while it does decrease your performance score, it does not affect your ability to hit beats in the future.
100% impact means “this decision will remain a factor for the rest of the game”. A good example of a 100% impact option is race selection in Starcraft, since you can’t change your race after you pick it. Researching weapon and armor upgrades in SC is another high impact option, since tech persists from the moment you research it until the end of the game, though due to tech ceiling, after a certain amount of time, the tech advantage will become non-existent and a non-factor.
How much impact is ideal depends a lot on the width of the decision. If a decision is wide enough, having high-impact actions results in analysis-paralysis, where the player feels lost and confused and unable to make a decision. On the other hand, low impact, especially impact of 0%, results in a false choice.
For example, many novice Go players feel intimidated by the board in the beginning of the game, since most stones placed are not going to be removed. But for real-time video games, even though they present a huge amount of options, they’re not so intimidating due to low impact. It generally doesn’t matter if you move an inch to the left, because you can always move an inch to the right as well.
Hardness and Softness
Hardness and softness refer to how hard of a counter an option is. A soft option only has a small advantage against the charge it counters, and can be reversed. A hard counter has little opportunity for reversibility – if at all.
In Age of Empires 2, the trash units, which are Medium Height and Broad -Jing, are very hard counters. This is what allows them to generally fulfill their Defensive, -Jing function.
But Hard counters are not particularly interesting. They can exist as part of a general failsafe system, but generally, very hard counters threaten to take the game back into RPS territory. While the Counter Triangle is commonly referred to as an RPS relationship, when making a strategy game we want to avoid recreating Rock Paper Scissors. Soft counters are much more interesting, since the potential to reverse the counter relationships creates interesting edge cases, which give the game depth, and which are exciting to identify and discover.
To put it in other words, hard counters make the game easy to learn, by providing the player with an easy to pick up strategic guideline. But soft counters make the game hard to master, by creating exceptions that a player needs to really understand the way the system works to take advantage of.
One of the imperatives of strategy game design is to make sure that the softer counters are not so soft as to be easily reversible. This is a common issue in Warcraft3. The game has a rough triangle of air>melee>ranged set up, but the ranged units often end up being able to reverse the triangle and defeat the melee units as well. For example, many human and undead players end up building riflemen/crypt fiend armies regardless of what their opponent is playing, since with good micro, they can kite and focus fire any opponent’s group.
As mentioned in the beginning of the article, Jing is meant to be primarily a tool for strategy game developers. Charting Jing Lines for key game moments allows one to clearly see clusters and other unevenness and give an idea of how to work them out. The Jing Line is best when charted for moments typical for a game, though if a special situation is giving issues or if you want to focus on one specific aspect of the game, it should be charted instead.
Since writing down all possible options is impractical, it’s fine to groups options together to make the line clearer, and analyze the strategy space at a more abstract level first before getting into nitty-gritty details. For example, one can attack move a group of marines at co-ordinates X,Y, and each pair of possible co-ordinates (that number is in the billions) is it’s own option. But we can write “move marines home” at the negative jing end, and “move marines to the enemy base” at the positive jing end, and some other key points (chokes, expansions) in between.
In writing or even just thinking out the triangle, the relationship of all these options, I hope to give game designers a good tool to cut back on iteration costs and be able to come up with good gameplay dynamics and squash game breaking imbalance before it even has a chance to be discovered. What’s truly great about it is that the designer can set up a very basic “triangle” system, and then work to give it depth, multiple dimensions, and build a really complex and deep experience on top of that.
One thing that makes it very exciting for me personally, is that writing about and thinking about this idea, looking at games and breaking them down, I have a very solid idea of how I would improve, or make my own version of those games.
For example, Ensemble Studios made for a very rock solid, robust game. They didn’t have the hindsight of 20+ years, so it’s incredible what they managed to make. But also, their design is not very asymmetric. Factions in AoE2 are mostly the same, and it’s understandable that making wildly different factions is both resource intensive, but also introduces all manner of potential balance issues. So it makes sense that they went with a very safe, very doable roster they did. The later installments in the Age series are much more asymmetric, and we can see just how well the formula holds up. Arguably, there are major issues which arise from the fact that they undermined one of the “railguard” systems – the lack of stone, a very strongly -Jing charged resource. Having the hindsight and the luxury to peer further from atop the shoulders of giants, we can see that the basic Age of Empires 2 formula can be further expanded with highly asymmetric factions, so long as the Jing identities and niches are maintained.
On the other hand, we can look at games such as WarCraft3. A game that I love deeply and want to continue to play and love, but which is also plagued by persistent balance issues. First hero choice is basically solved, and the army compositions remain fairly static and don’t get impacted by matchup and enemy maneuvering all that much. But from the perspective of Jing identities, we can start identifying, honing in and finding solutions for the various issues.
All in all, it seems an important and useful tool for designers. Yes, this is nothing revolutionary – the idea of counter triangles has been floating around for a long while. But while many great designers have been relying on intuition and fragmented theory pieces to make their masterwork, I’m hoping to articulate a very useful and deep way to move forward in game design.