At its core, matchmaking is a simple concept. It acts as a means to group a cluster of players into an online match based on their skill levels and a mutually desired experience (like a particular game mode, setting, or map). This seemingly straightforward concept was conceived from a motivation to provide players with a continuously challenging (yet rewarding) experience every time they play online. By launching users with similar profiles into a shared, dynamic experience, games become more enjoyable and, thus, more engaging.
That said, matchmaking isn’t just for the player’s benefit; it can also strongly impact a game's success. By increasing a person's enjoyment during play, engagement and spending are also likely to increase. Better yet, with increased exposure to other players (human or not), we can expect to see a further spike in conversion with players looking to showcase their visual cosmetics and abilities to others.
Numerous approaches can be taken when crafting a matchmaking model, and the path differs from product to product. Ultimately, this is a strategic business decision that seeks to attain growth in particular areas; for example, long-term engagement, average engagement times, or conversion regularity. Teams should build matchmaking systems around a product's and its players' immediate priorities.
This essay explores what it means to build an effective matchmaking system that retains, entertains, and engages an audience. Although a thorough breakdown of the math behind effective matchmaking models is beyond the scope of this essay, we will provide you with a summary of the systems and resources you can use to gain a strong comprehension of matchmaking and key takeaways to help spark your decision-making.
Why Does Effective Matchmaking Matter?
Everyone can agree that seeing someone win against all odds and by the thinnest of margins is one of the most exciting aspects of watching (or participating in) a team sport, whether physical or digital. Everybody loves a comeback, and on the flip side, everyone hates getting annihilated and never having stood a chance. It’s these nail-biting moments and feelings of competitive progress that keep players coming back.
That said, the notion of a great match can drastically vary between individuals. Some like a speedy, social matchmaking experience that prioritizes quick and enjoyable bursts of competition; casual experiences between players (or bots) with a wide range of skill levels but a similar love of the game. In contrast, others favor a challenging session with opponents as skilled as themselves.
This preference is just one factor highlighting the importance of superb matchmaking. Regardless of the model — many of which are based on a skill-based system known as Elo — matchmaking systems can be highly intricate and complex, and getting them right is essential.
Below are a few more ways matchmaking can create a positive impact on a game:
UX Improvements & ‘The Competitive Climb’
By continuously offering players a positive experience and a fair challenge, they will continue to be engaged, thus retaining longer and uplifting LTV potential. A trustworthy rating system (whether it be visual or numeric) can be significantly impactful on a game's longevity and success.
Many competitive gamers love rankings/tiers and the journey (referred to as grind) to climb as high as possible. This competitive climb creates incentives and opens up the gameplay the higher one advances. Often we see games double down on this competitive progression by periodically (usually seasonally or quarterly) resetting online progress to ensure that there is always a new ladder to climb. This keeps players in the matchmaking pool and KPIs high.
FTUE Enhancements
Churn can be high without a matchmaking system that accommodates the certainty of lower skill levels and new adopters. New players are unlikely to enjoy being matched with more skilled ones and experiencing sequential losses shortly after joining the game, and this inaccurate matchmaking makes them feel less inclined to spend more time or money.
As such, building a rating system that makes newcomers' first matches satisfying and empowering is a sure way to improve early results. Players can learn through the gradual exposure to other, more veteran players (or skilled bots).
Reduced Toxicity Online
Matchmaking is primarily considered a means to aid a game's competitive and social elements. However, mixing players always runs the risk of toxicity and arguments, which can lead to a negative experience that ultimately affects core KPIs.
Interestingly, some high-profile games (such as League of Legends) have tackled online toxicity by developing a matchmaking pool specifically for players with a history of toxicity (i.e., players who have negatively influenced the gameplay experience either through abuse, absence, or negativity). This keeps them separated from the larger community, preventing further damage and building trust between the player and the development team.
LTV Scalability
By regularly exposing players to one another, socially and competitively driven player types of all skill levels and playstyle preferences are commonly driven towards one of two things:
- Advantageous Consumables: Ranking boosters, loss protection consumables, etc.
- Bragging Rights: Limited edition cosmetics, skill-based collectibles, etc.
This presents an excellent opportunity to monetize. Monetizing online play can see players retaining for extended periods and becoming more engrossed in adjacent gameplay systems.
Exclusive cosmetics linked to higher ranks are a strong way of rewarding players while also exposing them to the value of premium content — incentivizing them to spend. Enhance this further by using regular and opportunistic special offers that occur at new (and existing) moments of gratification, i.e., increasing rank or playing a certain amount of games when players are most likely to convert.
Matchmaking can scale player engagement, retention, and spending when refined and maintained well.
Fairness & Responsibility
Matchmaking comes with the responsibility to provide something fair and fun to players. But ‘fairness’ has different meanings depending on the context.
In the Elo system, a fair game would be one where two players with an equal chance of winning are matched. However, this isn’t necessarily the most fun. Matchmaking is about fairness and providing players with a rewarding experience — everyone likes to win and feel like they deserve that win.
Beyond the quality of UX, players who win more games than they lose are also more commonly retained. As such, an ideal online matchmaking system might balance true fairness with slight modifications that improve the experience. Strategies such as pairing new players with bots or lowering the negative impact of a loss compared to the gain of winning offer fairly immediate solutions.
Look at Call of Duty Mobile, which according to data.ai, has earned $1.49B in global revenue since its release in June 2019, as an example of brand responsibility. Over the years, the franchise has been offering two online experiences — casual and ranked. Both feature the same game modes and rules but offer different player experiences, matchmaking conditions, and rewards.
Call of Duty Mobile followed the same suit as its PC and console counterparts — a strategy led by recognizing that changing this philosophy for the mobile port would have confused the product vision and probably caused the game to fail. Long-standing players of the numerous titles in the portfolio may have rejected the game due to unmet expectations, which would be a harmful result not just for the game but for the entire portfolio.
How to Deliver Better Matchmaking
Matchmake Newbies with Bots
When new players enter the game, they have no reliable experience determining a rating value. The Silver rank may sound good, but how skilled do you have to be to reach this prestige? Equally, players need to gain more knowledge of how many matches must be played for someone to achieve such a rating.
An accommodating FTUE can help maximize retention and engagement. Players joining the matchmaking pool without familiarity with a game and being matched with more experienced competitors are more likely to face multiple losses early in their journey, which could churn them for good.
Popular mobile titles such as Clash Royale, PUBG, and Fortnite have successfully used bots in matchmaking to facilitate configurable players. This is especially useful for funneling players to the most enjoyable experience early, allowing them to practice without penalty by pairing them against bots for their first online games. This initial period will give them the wins (and thus the confidence) that lift the chances of retention.
Additionally, developers can use bots to solve low liquidity (i.e., number of players in a matchmaking pool) in games with low CCU, which is especially likely in new or soft-launched games.
An alternative to using bots is using ghost data. Ghost data is performance data captured from real players and used to create realistic and often challenging opponents. This is a well-established technique in single-player (and now multiplayer) games that widely apply to many titles.
The most budget-efficient way to test realistic bots is to create a set of variables that help define bot personas based on existing (and ideal) player behaviors. With this, teams can create a collection of bots that encompass those behaviors, adding randomness to the matchmaking pool. Over time, utilizing player data gathered from the game can help inform AI design, thus creating increasingly more lifelike bots.
When data isn’t an option, bots can be developed using variables encompassing a player and their behavior during a match. Below is a list of example bot variables for a weapon-based combat game:
- Name: Determines the pool of names the bot can choose from. A name is selected at random during the matchmaking process.
- Area: Determines which zone the bot will navigate and predominantly stick to (unless strayed) during gameplay.
- Loadout: Sets a preferred arsenal (or weapon group, i.e., close range, automatic, etc.) for the bot, defining a role.
- Visuals: Defines the bot's visual loadout, specifically what cosmetics it has equipped.
- Rank: Determines a bot's ranking, which could (depending on the game) affect the bot's combat stats.
- Aggression: The chance that a bot will initiate combat when it enters a player's radius. This percentage is calculated each time the bot comes within a radius of the player or vice versa.
- Cowardice: The chance that a bot will run away from the player each time it enters a player’s radius.
- Loyalty: The chance that a bot will stay in proximity to the player when they have been found. This is recalculated on a timely basis.
- Stealth: The chance that a bot will utilize stealth environmental props (i.e., bushes), if at all.
- Communication: Determines whether the bot engages with emotes, if at all.
Please get in touch if you’d like a deeper understanding of designing AI for your portfolio!
Track Data to Guide Matchmaking Decisions
Similarly, as you would use data to improve core player behaviors and interactions, teams can also use data to guide matchmaking decisions and improve balancing.
Games as early as Zynga’s Farmville have collected player data to determine more efficient ways to retain players and increase their engagement time and spending habits. For example, Zynga can track common churn points within Farmville, such as aggressive pinch points, and balance them to prevent losses. These systems anticipate player frustrations and complaints in a multiplayer setting, curbing them before they churn out of the game for good.
The best way to understand player behavior and accurately tune matchmaking is by setting up a sophisticated analytical infrastructure within the game (more on this later). Google Analytics and Firebase are brilliant ways to manage your analytical findings, offering solutions that scale with the team's needs. Furthermore, data visualization and analysis software like Tableau create visuals (tables and graphs) to make them much more palatable for the team.
Then, once player behaviors are understood, they can be benchmarked against the industry average using a partner such as data.ai, removing any assumptions from the table and informing product predictions.
Simplicity and Sustainability are Key
Even teams with limited resources, budgets, or relevant experience can have good-quality matchmaking. When out in the wild, a game is compared to others in the market and not by the team that has developed it, so focusing on creating a simple (yet sustainable) system will get it most of the way there.
A team could have the most nuanced and complex skill-based matchmaking system. However, this offers little value if players struggle to find matches quickly or aren't entertained enough to climb the ranks or convert. A straightforward solution to this is to widen the matchmaking criteria over time. This way, players don’t wait for hours, as long wait times can damage engagement and retention.
Matchmaking complexity should rise according to two key factors: the larger a game becomes and the wider the gap in skill becomes. As player scale increases and skills diversify, more sophisticated solutions are required to accommodate as many different player types and abilities as the game makes possible. Teams must scale their matchmaking alongside the success of a product to ensure that the quality standard of matchmaking is maintained.
Secure the UX fundamentals to validate that the core experience is fun and enjoyable before adding more complexity to the system.
Maintain & Balance Continuously
Much like a product’s economy, the value of an online rank evolves and changes alongside a shifting playerbase. In the case of skill-based systems such as Elo, new players who are matched against other humans can provide points to stronger players, shifting the relative value of a player's skill rating.
As more players join the game, the average value of a rating may shift. If players are absent for an extended period, they may return to a rating that no longer reflects their ability — a phenomenon known as ‘level decay.’ This is useful as an engagement tool since it encourages players to regularly log in to the game so as not to lose their progress.
Another form of involuntary rank regression is seasonal refreshes, which are becoming increasingly popular, with games like Call of Duty Mobile, Overwatch 2, League of Legends, and Rocket League adopting some adaptation of the model. With seasonal refreshes, every player's rank is reset (or reduced) regardless of their level of engagement. However, games often reward highly engaged players with spoils exclusive to that season as a token to showcase.
Season resets ensure that ranks grant balance by preventing veteran players from holding the high ranks while giving new and returning players a fair chance to climb. Teams can expect to see a spike in engagement, returning players, and increased spending, with players looking for advantageous purchases during this period.
Seasonal refreshes are a great way to ensure that players always have obtainable goals. This is especially powerful for games with a shallow ranking system; for example, material (bronze, silver, gold) progression. Gains aside, adding seasonal refreshes later in production will likely anger a portion of the playerbase that has spent a long time maintaining its rank. Consider what it would mean to lose these players (and their engagement and revenue) and be prepared to lose this.
It’s vital to ensure that teams track data (as mentioned earlier) as a tool to monitor and maintain a live matchmaking system. Here is a handful of example scenarios to look out for:
- Player Win/Loss Ratio: How does player retention change based on their win/loss ratio? If you want to go deeper, check how this changes based on ranking.
- Matchmaking Times: Are matchmaking times significant? Are a substantial portion of players exiting the game during matchmaking? One way to resolve this could be to add bots to the matchmaking pool or in games that require a selection of roles to be filled, rewarding players for choosing less popular roles (see below).
- Promotion Requirements: Are higher ranks enticing enough to drive players to increase engagement so as to reach them? Lowered requirements or increased rewards may uplift motivation.
- Character Used/Position Played: Are the chances of winning significantly higher when certain characters are being played? Are the majority of players choosing a particular character because of that? If either answer is yes, matches are likely becoming repetitive, harming retention gains.
If bandwidth is a concern, consider pre-established off-the-shelf skill rating systems, or utilize them as a starting base. There are some great options out there, including:
- Glicko-2 is a rating system based on Elo and has seen use in popular titles such as Team Fortress 2, Counter-Strike: Global Offensive, Pokemon GO, and the Splatoon franchise.
- Microsoft’s TrueSkill 2 uses a Bayesian skill rating system broadly comparable to Elo and has been used in the Gears of War and Halo franchises.
Matchmaking Methods Define The Experience
There are several pre-established matchmaking methods to choose from, and what’s most suitable depends on a product’s genre and audience. Selecting a matchmaking model is an important decision that guides the online experience for players. Is this a game focused on progression and skill, or a casual game that focuses on the fun?
Below is a list of matchmaking models split by the market they’re most conventional to and why. Although this guideline suggests matchmaking models by market, teams can interchange them on a product-by-product basis:
Hardcore
Traditionally, hardcore games rely on player rankings as the basis for matchmaking. This method leans on the philosophy of players having strong overall skills rather than strong skills in a specific area.
- Player-based: Player-based matchmaking focuses on a numeric or visual player ranking bound to the player account.
Midcore
Similar to hardcore games, midcore games center matchmaking around player skill. However, rather than utilizing overall player skill with a measurement such as player rank or level, midcore games often allow a more diverse approach by matching based on characters or classes. This enables players to practice new characters without the punishment of a potential demotion.
- Character-based: Character-based matchmaking concentrates on character rankings rather than player-based. Even if a player is high on one character, if they begin playing a new one, their rank on that character will be low.
- Class/weapon-based: Class-based matchmaking is similar to the character-based model. However, it focuses on a particular class or weapon rather than a playable avatar.
Casual
Casual matchmaking often applies a contextual approach, pairing players by desired experience instead of player skill. This is the quickest and most straightforward approach to matchmaking and is perfect for teams looking to test an experience or provide a space that purely exists for fun.
- Mode/map-based: This approach pairs players based on a preferred map or game mode. Players of any skill or ability level can match with one another, making the focus more on casual fun than progression.
Hidden vs. Public Rankings
Displaying the fine details of an Elo rating and a game matchmaking system is rarely a good idea. Revealing the intricacies of matchmaking systems opens them up for manipulation. Additionally, providing players with all the data can confuse, mislead, or even frustrate them, reducing trust in the system with possible claims of a lack of precise or fair matchmaking. Mistrusting, confused players can damage retention, engagement, spending, and reputation for this game and other products in the portfolio.
Instead of exposing players to substantial data, consider representing player ratings through a simplified system, like ranks or symbols that use familiar hierarchies such as bronze, silver, and gold medals.
Simplifying a ranking system to something quickly comprehensible (preferably shareable) helps players understand their rating and compare themselves with others. This creates a space where revenue and engagement can be increased by giving players a means to brag about their experience and status within a game.
Casual games are more likely to benefit from a more conventional, visual form of ranking, preferably with a progression system that is quickly and easily understood globally (i.e., numbers, stars, trophies, or materials). This audience traditionally doesn’t retain long-term and is more likely to play, share, and spend deeply for a shorter period, especially if the content is easily consumable.
Hardcore games should seek a more profound experience, giving players a long journey and several goals to strive towards. If this audience can perceive value, it is usually willing to take the time to understand more complex systems (likely to be a sizable and often changing numeric number), as it is looking to invest its time (and money) long-term.
Midcore should strive for a middle ground that offers value and deep thinking but is still easily consumed and, in some cases, comparable to other titles on the market.
Player Feedback Is Paramount (and Often Unreliable)
In addition to player data, player feedback can provide a meaningful foundation for supporting matchmaking. However, as seen with the Elo hell movement, player understanding of the nuance of matchmaking systems does not reliably reflect reality. Players don't have the complete picture nor a firm grasp on the product’s long-term goals, direction, and previous learnings, so following player feedback as a guide rather than an information source can be dangerous. Therefore, when considering player insight, it is worth focusing on broad trends across that feedback rather than qualms specific to individuals.
If a significant amount of your community is vocal about waiting for long periods to find a match or finding the game too complicated, it’s worth looking into those issues. But remember, the vocal is the minority. Listen to the players but consider that what individuals say may not reflect the bigger picture. Data should help define the difference between what is said and what is meant.
Uncover What Matters Most
Take the time to consider the most appealing elements and the desired player behaviors. Once determined, developers can build matchmaking systems around them:
- Should players specialize in a character or have a diverse range?
- Is matchmaking speed or matchmaking accuracy more important?
- Is social play more important than skill?
Nobody knows a game as well as the playerbase and development team behind it. Player feedback, market research, and analytical data provide an excellent foundation to guide and reinforce product decisions and priorities.
Supercell’s Brawl Stars is an interesting case and has a clear signifier of one of its pillars within its matchmaking philosophy — Brawlers. Unlike many mobile matchmaking systems, Brawl Stars’ matchmaking is character-based rather than player-based, encouraging players looking to climb the ranks to specialize in a character while offering diversity to those looking for variety. A substantial benefit of this is that if a player is highly skilled at one character but interested in trying a new one, they are matchmade with other new players rather than that at the high level, allowing the player to be experimental without risking their prestige. This is great from a product perspective; it increases the likelihood of players raising their spending on new characters, as there is less of a risk involved than, say, League of Legends or Overwatch.
Conclusion
Consider what is most important to the product and its audience and build accordingly. Backtracking from one matchmaking system to another is costly and detrimental to the game's success.
For some games, the enjoyment of close matches keeps players coming back, and matchmaking in those games is closely related to player ability. In others, giving players an easy win periodically is similarly beneficial to designing the best player experience, encouraging them to spend more time (and money) in the game. But what defines the dynamic is that skill-based matchmaking is a business strategy to retain players. How players define fairness is subjective, but their metrics are not.
There are plenty of solutions available to quickly test matchmaking systems in the wild and to a professional standard. As such, fundamental mistakes are often rejected by audiences and may cause irreversible damage to a game's LTV. Off-the-shelf rating systems like Glicko-2 and pre-established and validated models — class-based (Brawl Stars) or player-based (Call of Duty Mobile) matchmaking — offer high value at cheap costs and help teams make informed decisions from the get-go. For the best early results, don't try and reinvent the wheel. Use the tools you have at your disposal and manipulate them to a product's specific needs as you learn what the playerbase wants. Although, in some cases, this may cost a small amount of money upfront, it will cost you much more in the long run if you need to redesign based on misunderstood systems and fundamental mistakes.
And remember, having robust fundamentals is far more critical than nuance for teams of all sizes and prestige. A sophisticated system is worth nothing if it only works on occasion or doesn’t match the quality of the competition — players value enjoyment in consistent measures over enjoyable refinement as a rarity.
A big thanks to Becky Matthew for writing this essay! If Naavik can be of help as you build or fund games, please reach out.
