Game Balancing Guide
This post is all about what game balancing is and how to do it. Just remember that every game has its own unique needs and challenges, so cherry-pick whatever sounds reasonable for your game. This post will become a “living” post of sorts, added to over time. Tell me about your own balancing tools and tricks at [email protected] or in a comment. Sections in this post: Before we get into the balancing itself, we need to know who we are balancing for. Paradoxically, this doesn’t have to be your target audience per se. A player who bought your game but never installed it is still part of your target audience; but they are not part of your balance targeting. Some players simply don’t care for your game at all for one reason or another, or lose interest after a while. These matter to game balancing because they are inevitable. One element of your balancing is who you are not balancing for. There are two very broad types of player attrition: the burn and the churn . Players who touch your game and get burned, like on a hot stove, never to touch it again. Knowing who you are willing to burn means that you can plan for it and exclude any balancing fixes that cater to that group. Pleasing the burned only wastes time. Burned players will never come back. It can be because the genre or gameplay doesn’t appeal to them, because it didn’t match expectations, or for many other reasons. What’s important to understand is that this is fine . You’re not making a game for everyone. Your multiplayer action game is not going to appeal to a fan of turnbased singleplayer games and no amount of balancing will change this. Churn rate is the rate at which players stop playing your game. In the best of worlds, you are able to collect data that tracks when players are churned. In a puzzle game with linear progress, for example, if many players stop playing halfway through the fifth puzzle, that could be an indication that they don’t understand how to proceed. Or if many players stop playing after dying four times against the second boss, it may mean that it’s not clear enough or too hard. Players churn after giving your game a chance. You can win churned players back or delay the churn by understanding your audience. One goal of good balancing is to keep these players playing. In his book, Introduction to Game System Design , Dax Gazaway employs measurements that are very practical and useful for when you want to figure out your design target. The book has more measurements. For this post, I’ve stuck to those that can be quantified. Understanding how a new game works is an acquired taste. Many a gaming console has been purchased to play Grand Theft Auto and EA Sports FC exclusively. A gamer today can be someone whose hobby is to play Fortnite or League of Legends , and who hasn’t spent more than a handful of hours in total even trying other games. Gazaway expresses interest in challenge as a ratio of failures against successes. A game that can be measured as 20 failures against 1 success (20/1) is clearly a lot harder than one that is measured as 1 failure against 100 successes (1/100). A player that starts getting frustrated after failing twice against each success will be burned by the 20/1 game. But so would a player that thrives at 10/1 while playing a 1/100 game. Not to mention that a game that suddenly goes from 5/1 to 20/1 is likely to churn some of its players. Free-to-play advocate Nicholas Lovell talks about the Starbucks Test , which he paraphrases from then Natural Motion CEO Torsten Reil as, “Can you play your game and have a meaningful experience in the time it takes for a barista to make your macchiato?” This is the very bottom of the time investment rung, where you can fit a play session within your other daily phone activities. On the opposite end, you have games that take many hours, even several days, to finish. According to Midia Research , “[c]onsole gamers spend – on average – 10 hours gaming per week, while PC players spend a little less (9.7 hours).” Gazaway talks about session time vs total time . A single-player roleplaying game may take 80 hours to complete, but you can play sessions of variable length. It may even pass the Starbucks test, letting you jump in and defeat a monster before they call out your name. If you are making your game for middle-aged gamers in the west with stable income, you can expect them to pay more than a kid with no personal bank account, or to a young student in a third-world country. Income inequality is sadly universal. Deciding to charge €69.99 for your game will exclude everyone whose ceiling is €5, for example. Simple mathematics. They’re burned before they even start the game. They don’t get to the stove. Even if they’d give the demo a chance, they could still never afford your game. This needs to be mentioned here, because we tend to forget that the money and the balancing are actually strictly related, even if all it affects is the point of entry. The first step to balancing is setting points of reference . This is the foundation for everything you will do, and can live in a bullet point list, a spreadsheet, a design spec, or some other form. In the excellent book Game Balance , by Ian Schreiber and Brenda Romero, game balancing is divided into three “metaphorical dials that are coming together” to create the sense that a game is balanced. Think of difficulty as how much pushback the game has against your success . This can be expressed as a ratio of success. For example, if you want the average player to win 75% (or 1/3 in Gazaway’s failure to success ratio) of the fights they start on the first try, but only 25% of bossfights (3/1), you can see if those numbers check out by collecting data from play tests. The number of things in your game . Enemies, ammo, loot, levels, coins, dialogue lines, health points, and so on. Quantities are often adjusted for practical reasons more than balancing and must then be balanced accordingly, for example how many enemies that can be rendered at any given time, or how many weapons you have budgeted time for delivering. When to press the key. How fast you need to move and when you should stop moving. For how long an interstitial “hurt” animation plays before returning controls to the player. The duration of the progress bar on that new building you are constructing. Timing is about duration as much as the moment. As has been discussed before , the lack of common language around game design means that we often resort to references to other games. This makes it really hard to talk about balancing with any consensus. Therefore, the first thing we need is everyone’s buy-in. Everyone working on the game, that is. We may want our game to be as difficult as Dark Souls , feel as good to play as Super Mario Odyssey , and take about as long per match as PUBG . These are fairly concrete and measurable comparisons where we can apply a more scientific approach. If it takes us about 10 attempts to kill a boss in Dark Souls , that gives us a number to balance against. Just don’t make the mistake of copying. If you have those 10 attempts against a boss with you, take that number and consider it. Test lower, test higher. Look at it from quantity and timing angles as well. These are inspirational and communicative references, not your rulebook. They say something about how the referenced game works, not how your game should work. It’s also common to set up antithetical references. We don’t want a match to take as long as a full Battlefield 3 Conquest match, or we don’t want timing to feel as punishing as in Ninja Gaiden , or the UI to be as confusing as in Crusader Kings II . Whatever it is, it needs to tap into this same common pool of references. References shouldn’t be limited to games. It’s common to refer to movies, comics, TV shows, books, documentaries, historical events — anything that can act as an antithetical to what you have in mind. Same goes for balance references, of course. The key to references is that everyone on the team can understand them. A designed reference is something you come up with in the team that becomes true for you . It can be a pillar, such as “Character-based play,” that lets you tap into your own project to figure things out. Someone changes something, and you can then go back to check if the thing actually is character-based play. If it’s not, it’s failing, and you may have to rebalance or redesign it. Just like the spend ceiling mentioned before, there are limitations you must contend with. When you print a deck, it has a certain number of cards. A Switch console can only push a certain number of textures per frame at 60 frames per second. The worst part with limitations is that they are extremely inflexible. If you break the card count on a print plate you need a whole second plate, even if all you are after is a single card. Similarly, rendering just a few more textures per frame will make your game run slower. Technical, practical, and financial limitations are important to know before you set your other references. Measurable goals for what you want the experience of play to be. This can be session lengths, more or less pushback, or specialised metrics like how many seconds should be the maximum allowed to pass between game launch and the first kill (“seconds to kill”). Set these goals very broadly. This will make it easier to make exceptions to them the same way a bossfight against an enemy that has a lot of hit points immediately changes the “seconds to kill” dynamic. Many action games have an average lifetime for enemies around five seconds. Maybe you want your enemies to be tougher or smarter, so you state that you want average enemy lifetime to be 15 seconds. This is a type of content goal, where you state how much screen or even CPU time something is intended to get. With many games content-driven or even content-locked by design, this should be more of a consideration than it is. It’s helpful to use an object or variable as an anchor for the central mathematics of your game. Health points is an effective anchor. If you set a player character’s health to 10 and you let this serve as your anchor, you set the damage weapons deal and armor prevents, what ratios you want between gold and health, and how effective the conversion of mana to health points should be. This would turn health points into the main dial for almost all balancing in your game. You can also speak of multiples of health points as measures of difficulty. An easy enemy may have half of the player character’s health, while a tough enemy has ten times the same. When you do this for an object, it can be the game’s default pistol and the experience of playing with it. Which is how John Romero writes about testing new Doom levels: all of them needed to be playable from a pistol start. The pistol is then the anchor for Doom ‘s balancing and all the various numbers that affect the pistol, such as ammo pickups placed in the level, damage, fire rate, etc., will affect the play experience of every single level in the game. A range is a floor and a ceiling and all the numbers between. If you have a numeric anchor, like 10 health points, it should probably represent the bottom of the range in a progression-driven game. However, an object anchor like the pistol could represent an average for multiple ranges. You may have ranges representing distance, damage, armor, gold value, durability, ammo, and so forth. Ranges are a good way to incorporate limitations. If you know that your frame rate will limit the rate of fire for your firearms, for example, you can set the ranges accordingly. A handy thing with ranges is that you can use them in random generation. For example, if you determine the extremes (see Set Extremes, later) you can test with random values within that range when you run your tests. One way to set references that I’m personally very fond of is to use words instead of numbers. To fuzzify the numbers. If you have the concept of range in your game, you can talk about it as Close, Far, Long, and Distant, for example. When something needs to be changed, you bump it on this ladder rather than changing the numbers directly. This makes it easier to talk about the balancing in relative terms. It also lets you determine which ranges you want to have without having to get stuck discussing the numbers. If you need more nuance, you can add more rungs on the ladder. Game economy design is basically a science in itself. But a good way to start building a virtual economy is to tally all of your sinks first. These will show you how much virtual currency even has the potential to flow through your system. If this number is capped or particularly low, you should limit how many sources you have based on player time investment, difficulty, and other targeting goals from the previous stage. Make sure that the player can always afford something , but also make sure that they can never afford everything until your reference goals are reached. When you set out to create a system of balancing that has numbers at its foundation, you always need to start from somewhere. One way to start is to simplify everything down to one of four value systems: 1s, 3s, 5s, or 10s . Start from just 1s or 10s, then introduce some 5s, and finally some 3s. This way, you will quickly have the gut-level foundation for your balancing. To refer back to health, one mana could be worth five health points, while one gold is worth three in the form of a health potion that heals three points. Another way to create base numbers is to use a Fibonacci sequence . A Fibonacci sequence is “a sequence in which each element is the sum of the two elements that precede it.” If you start from 1, this gives you 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144, etc. This sequence can often come in handy, and you can play with the principle of the sequence as much as the sequence itself, for example to set item prices with a good initial spread. Perhaps if your health potion costs 25, then you could price the following item tiers 50, 75, 125, 200, etc., using the same rule of summing up the two preceding numbers to get the next price. If points of reference are there to give us something to anchor our balancing to, points of differentiation are where we design the game. A focus on diversity. Ways to mess with your points of reference and give players something to strategise around. Greg Costikyan wrote about this on Linkedin, in the context of combat systems, saying that “The key is multiple points of differentiation, to create interesting strategic choices for players. If it’s all about DPS, it’s boring.” His examples include things like reload times, costs, damage, damage type, combined effects, and more. Ways to add more ranges and more points of differentiation between one gun and the next. Destroy them, disable them, put them on timers, require resources for their use. Take things away. Once the player is used to something, try removing it and seeing what happens. It can be the worst thing you did, or it can give room for you to create a needed change of pace. Like when the enemies in Batman: Arkham Asylum start mining the gargoyles Batman is using to hide. In game theory, a dominant strategy is one that always trumps other strategies, while a dominated strategy is one that is always trumped by everything else. Players will naturally gravitate towards dominant strategies as their understanding of a game evolves (at least if they are accepting of learning new rules), and will bounce away from dominated strategies. This is not necessarily a bad thing. You may want the starting pistol to be a dominated strategy, same as the BFG 9,000 is a dominant one, but other restrictions of usage means that you will never stop using the first and can’t rely entirely on the latter throughout the game. Many games restrict features by requiring transference. You are highly unlikely to find the best mod in Warframe , if it’s even possible, so instead you engage in the act of combining mods to convert them into better ones in multiple tiers. A process that costs resources at every step along the way. Going back to the Game Balance book and its three dials, most of these systems use time as the essential anchor. You need to play to collect more of the resources you need for the crafting, and there are rarely any guarantees, which means you need to put even more time into it. Controlling when a difference is introduced is a key element of time balancing. If you find the best weapon in the game right out the gate, players are unlikely to ever switch, and if you provide access to it too late, many players will never see it at all and many who do see it may be less inclined to change their winning strategy. You can also control how easily a difference is accessed. Players may have to complete a special quest or spend an expensive amount of resources or time. Perhaps most commonly for Massively Multiplayer Online Role-Playing Games (MMORPGs) and similar genres, where you will have the squeeze the best possible mileage out of every piece of content, changing a color and making something deal a different type of damage (red for fire, blue for lightning) modifies things enough to keep the player playing. Patterns you don’t want the player to fall too deeply into can be altered by diminishing returns. Enemies may start wearing more body armor if you kill many of them, forcing you to switch things up. The following list will serve as a living document, with tools added over time. (Sporadically, I must add.) Make sure to only change one thing at a time, and to document what you changed. This makes it easier both to know what you did, what made the ultimate difference, and to write changelists once you publish it to your players. Sometimes, you have a solution or feature that just works . In a first-person shooters, you’re shooting, aiming, and reloading and shooting again, constantly. Same with card games; you’re drawing and playing and drawing. If a segment of your game isn’t working as well as another, repeat a solution that worked before. Keep a record of go-to solutions for future use. If you get stuck on something you decided on, try doing the exact opposite in your design and see if that takes you anywhere. Maybe you added more health to the enemies because the combat was too short, and it made them feel like bullet sponges. Trying to decrease their health instead and then increase the spawn rate could be the solution you seek. Touch, but don’t taste, is when you have a really cool feature that’s so restricted that players never use it, or they save it for a “later” that never happens before the end of the game. If you see this pattern in your game, you can give the players a bite instead. Let them play with the feature without any risk of losing anything, to teach them that the feature is worth using. First one’s free. If you feel that something lacks depth or isn’t interesting enough, you can try to add an additional dimension. Maybe your dialogue system with the Yes and No response options is making for bland conversations. Add a third option that plays against both Yes and No. Rock, paper, and scissors, is a metaphor for the related variation. Sometimes when an action feels like it doesn’t have enough “oomph,” you can add more feedback to it. Tell the player that the thing is in fact happening. Particle effects, cheer sounds, camera shake, stop frames, easing functions leading into or out of the action. Think of the sounds the Halo energy shield makes as it starts and finishes recharging. If your game has many moving parts that are active at the same time, it can be helpful to divide them into separate stages of interaction. Games that are maths-heavy, such as some styles of role-playing games, will often divide their “buffs” into stages. Something that lasts for 30 minutes is a “prebuff” that you can activate when you are expecting something to turn tough, while the one that lasts for just 30 seconds is something you use in the midst of combat. This division serves as a subtle way to structure the player’s buff cycle. If you see players doing something over and over again, even though it’s the wrong thing to do, you can try to make it viable to play the way they are playing. This is the territory of Coyote time, auto-aim, and other mechanics that try to predict game effects from player intent. This balancing tool is almost cliché at this point. Double or halve. You want the thing you are doing to make a tangible difference, so you double it or you halve it depending on whether it needs an increase or a decrease. When you feel that something needs a bigger impact than doubling or halving it yields, you can try multiplying or dividing it by ten instead. This is a much more drastic move and will rarely be used except as a method to find your extremes. If players are unaware of interesting features in your game, you can make them excited for it before they get to play with it by putting the door before the key. Dangle the treasure, the lever, or the new enemy in front of them, so that they will actively seek out how to get to it. Whether by finding an actual key or by understanding that they need to go up to get inside the locked room somehow. Metroidvanias are fantastic at this. An exercise you can use when numbers feel off is to decrease or increase by a lot (doubling/halving or by ten) until you find the lowest low and highest high. Then you construct a range out of those two values, and you start testing with different values within that range to try and find a sweetspot. It’s a good way to get away from pure feel and narrow it down. Probably the most common trap in balancing is to create a cool new thing and immediately feel a need to make it worse. To “nerf” it. This is a reminder to not do that, but to try to capture why the thing feels cool and to extend the same coolness to as much of the game as possible. Targeting : who you are balancing for. Points of Reference : what you are balancing against. Points of Differentiation : the exceptions you are making to your points of reference. Tools : various methods and techniques that you can use when balancing your game. Refusing : players that don’t want to learn anything new at all, regardless of how you package it. They want things to behave a certain way. Gazaway’s example is how Wii Bowling managed to attract people who didn’t really play games, because it was close enough to real bowling for the rules to transfer readily between the different media. They don’t want new rules, but they can transfer what they already know. Resistant : players that are willing to learn some new things to be able to engage with games in a genre they like or a series they follow, but tend to not leave their comfort zone. The aforementioned Grand Theft Auto and EA Sports FC players. Neutral : where you’d categorise most “casual” players. Willing to learn new rules to play a specific game, but more as something that goes with the territory than something to be actively sought. Accepting : players that don’t just reluctantly embrace new rules but dig deeper and even learn more than they may need to. This is the space for many “hardcore” players, that stick to a genre and explore it deeply with character builds and clever solutions. More likely to try something new that they still recognise. Enthusiast : people that read game rulebooks for fun. They will probably move on to the next game once they have understood this one.
Gaming
Tutorial
Design
Business
Programming
C++
Data
C#
JavaScript
AI
Sports
Shell