Matchmaker In-Depth
Last updated
Was this helpful?
Last updated
Was this helpful?
Learn more about Edgegap’s no-code matchmaker concepts in-depth and customize to your needs.
Edgegap Matchmaker is a managed service using your own rules to find the optimal group of teammates and opponents in seconds. Get started in 5 minutes and test the entire feature set for free. Upgrade when you're ready to a private dedicated cluster with enhanced security. Native integration with Edgegap Deployments delivers best-in-class ping no matter where your players are located.
Our matchmaking philosophy is driven by player perspective, with core principles of matchmaking:
maximizing match fill rate and social integration (friends and lobbies),
fast matches and high match quality controls (low latency, similar preferences),
high service availability, observability, analytics, and user-friendly error handling.
To deliver on our core principles, we prioritize first and foremost:
real world examples, game engine SDKs, and step by step guides,
modular and configurable features, focused on player experience,
no-code configuration and support for zero downtime service updates.
There are three essential concepts to each Matchmaker:
☁️ Hosting Cluster - underlying server infrastructure, fully managed and operated by Edgegap.
📏 Configuration - set of rules and settings that define how the matchmaker will operate.
🌐 Service Instance - the 24/7 live matchmaking service running on the Cluster, using Configuration to match players together and produce deployment (server) assignments.
To take advantage of new features and bug fixes, you’ll need to update your matchmaker version.
To make things easy, we host and manage matchmaking entirely for you 24/7. You can use our
Free Cluster (shared) to test all features and explore synergies with your design,
matchmakers on free cluster shut down after 3 hours automatically, requiring restart,
Private Cluster to ensure stable and dedicated environment for your production needs.
Upgrading to a private cluster only takes one click. Changing Private Cluster Tiers after launch, without any player downtime, is also possible with ⏩ Rolling Updates. Managed clusters provide you high-availability service hosting maintained by Edgegap.
Prepare for success and optimize after launch, so you don’t block your players on release day.
Every matchmaker is based on JSON configuration, validated automatically during restarts.
See Getting Started for our SDKs and detailed example scenarios.
Editing a running matchmaker will trigger a quick reload, deleting all tickets and causing short downtime.
Profiles represent entirely separated matchmaking queues, only sharing the same matchmaker version. You can configure any number of profiles for your matchmakers, though splitting up your player base will result in longer queue times for your players.
Some game modes may require more vCPU / RAM, especially if they support a higher number of players. A single matchmaker may include multiple profiles, each linked to an app version with adjusted resources.
Each matchmaker profile points to Apps and Versions in your Edgegap organization. Separating development and production environments with separate matchmaker may help prevent accidentally releasing a version which is not production ready.
Every ticket joins matchmaking queue based on initial rules defined for their chosen profile.
An openAPI specification will be generated based on your config once you create your matchmaker.
Rules
Each entry in profile at path .rules.initial represents a rule, where:
key is a string value to name the rule however you prefer; e.g. match_size , and
value is an object defining the type and attributes of the rule, adhering to our standard ruleset.
Operators player_count and latencies may only be used once per profile.
Operators (rule type)
To demonstrate various operators, we will refer to our example configuration:
All rules have to be met simultaneously to initiate assignment. You may match using operators:
player_count is a special rule defining how many players need to match to initiate assignment,
example defaults to 1 team (no team split), see Teams for more details,
matchmaker always strives to maximize match fill rate, up to specified max_team_size :
if max team size is reached the match is made immediately,
otherwise, players wait in queue to fill match until expansion (or expiration) is about to elapse,
shortly before expanding (or expiring), if a partial match is possible (>min and <max team size), this match will be made with all players in same expansion stage and any preceding expansion stages with matching rules (other than player count).
Rule player_count is designed to maximize match fill rate (players per match).
Rule player_count is required and may only be defined once in your initial configuration rules.
string_equality matches players with the exact same string value, for example:
example selected_game_mode rule will match players case sensitively:
Alice
Erin
Frank
Bob
Charlie
Dave
number_difference matches players within the absolute numerical difference from each other:
example elo_rating rule above with "max_difference": 50 initially:
Rule latencies is optional and may only be defined once in your initial configuration rules.
intersection matches players with one or more overlapping string values, case sensitively:
example selected_map rule above with "overlap": 1 :
Optionally, expansions can be defined to modify each rule’s attributes at time intervals after ticket creation, effectively expanding the pool of players which can be matched, speeding up queue times.
Initially, we require 1 team composed of exactly 4 players (possibly split in groups) with:
maximum of 125 ms latency against the same (any one) beacon,
latency difference of 125 ms or less between lowest and highest player value for that beacon,
skill rating difference of 50 points or less between lowest and highest ranking player,
the exact same (case sensitive) selected game mode,
at least one matching map selection (case sensitive) amongst players,
at least one matching backfill group size value amongst players - see Backfill.
In the example above, we expand the search by modifying attributes after:
To support multiple teams, increase team_count value. Get total number of players by multiplying team size and team count:
for "team_count": 2 and "max_team_size": 5, a total of 10 players is allowed for a match,
for cooperative and free-for-all game modes, set "team_count": 1 .
Team IDs will be assigned to 🎫 Player Tickets (accessible by game clients), and injected in deployments as Injected Environment Variables (accessible by dedicated servers).
Each player creates tickets to join the matchmaking queue. Tickets are compared and if criteria defined in the 📏 Configuration is met, a deployment will be assigned, and players may establish connection with game server.
The matchmaking process can be summarized in a few steps:
Player creates a ticket status="SEARCHING" ,
Player repeatedly reads their ticket every second; looking for matches:
Team has been found status="TEAM_FOUND" (see Teams and Join as Group),
Optionally team disbanded status="SEARCHING" , shuffling groups.
Optionally team disbanded status="SEARCHING" , teammate abandoned.
Optionally group disbanded status="CANCELLED" , group member abandoned.
Match has been found status="MATCH_FOUND" ,
Deployment started, waiting while Assignment is null,
Host has been assigned status="HOST_ASSIGNED" ,
Deployment is Ready, server process starts initializing,
Players read the assignment (URL & external port) from their tickets,
Players retry connection to game server until ready.
Matchmaking may stop before "MATCH_FOUND" due to either:
Player abandoning (deleting) their ticket,
Ticket is immediately deleted and reading it will return 404 Not Found.
Ticket expiration period reached status="CANCELLED" ,
Ticket can be read until removal period is reached, then it’s deleted (returns 404).
Maximum deployment retries exhausted status="CANCELLED" ,
Ticket can be read until removal period is reached, then it’s deleted (returns 404).
If players experience any issues, or a long queue time, they may attempt deleting their current ticket and restarting the process by creating a new ticket.
If the player has been matched and assigned a game server, their ticket will be deleted automatically. Players who abandon queue after HOST_ASSIGNED may be replaced with server Backfill.
To provide best possible latency, 🎫 Player Tickets may include game clients’ latency for one or more Ping Beacons. Once clients perform a “best effort” measurement, they construct an object with all beacon-specific latency values:
{
"Los Angeles": 224.4,
"Manama": 23.2,
"Tokyo": 167.4
}Beacon latency values are used to prevent matching with players in a vastly different region.
Joining matchmaking with your friends relies on a third party identity system. Most games will use Epic Online Services, Nakama, Steamworks, brainCloud, Lootlocker, Playfab, Apple Identity, Google Play, or others… These systems usually expose a platform-dependent friend system and a lobby/party system which can be used to communicate across the group.
To join matchmaking with a group, once the group is assembled:
Group leader requests ticket attributes from other members,
Group members provide requested information to leader,
Group leader creates a Group Ticket using our SDKs or the API,
upon receiving assignments, Group leader sends each member one assignment,
Group members follow regular Matchmaking Process as if they joined alone.
In case a group member decides to abandon their ticket prior to HOST_ASSIGNED, all tickets of group members immediately update to status="CANCELLED" pending removal after the configured period. Afterwards, this matchmaking process may be repeated from scratch with a new group ticket, disregarding the previous tickets.
If a group member decides to abandon match after HOST_ASSIGNED , game server is responsible for waiting for a pre-defined period of time and finally issuing new Backfill tickets to replace the player.
If a group member’s game client crashes, they may reconnect as long as they have saved their assignment ID persistently (on drive) between game restarts.
See Getting Started and our SDKs with automated reconnect and more.
Optionally, some games may have special matchmaking needs, such as:
allow new players to join games in progress (friends or strangers),
replace players who leave (leavers) after game starts to avoid restarting match,
allow spectators to join and observe tournament or friends’ matches (e-sports),
centralize players in larger servers to provide more social interactions (MMOs),
let players choose a server to join through a server browser (custom).
Any profile can be used for Backfill as long as attributes of backfill are valid.
This feature is in development, scheduled for late 2025, stay tuned for updates! 🚀
Some games may want to let players join servers with open capacity. We're working on additional matchmaking features allowing this in addition or as a replacement of classic matchmaking.
Reach out to let us know about your server browser requirements and needs!
To pass information to game servers and provide operational insights, we provide additional features.
Map players to tickets - players should send their ticket ID to game server when they connect.
Instances use CPU and memory to perform matchmaking and assignments, associating a hosting cost with each private matchmaker. See resources and prices associated with each tier on our pricing page.
The resource requirements for your instance will depend on factors:
number of players - more players make more tickets,
number of requests - polling for status more often creates more load on matchmaker,
configuration complexity - number of profiles, rules, and expansions,
average match duration - shorter duration means more matches, tickets, and higher load,
ticket expiration and removal periods - keeping stale tickets for longer consumes more memory,
client retry fallback logic - retrying with short jittered backoff helps spread traffic burst peaks.
If a matchmaker is experiencing high load:
if CPU is throttling, matchmaking could be slowed down,
if matchmaker runs out of memory, it will restart without losing ticket information, hoping that clients will implement exponential backoff and the burst is retried over longer period of time.
Display ticket IDs and deployment IDs in client match history UI to trace players when troubleshooting.
Gain insights into your matchmaker load and performance, no code or configuration required.
🌟 Upgrade Matchmaker to Enterprise Tier to unlock matchmaking metrics and insights:
Keeping track of compatibility between server and client versions can get confusing. Follow our tips for reliable releases, updates, and preventing downtime or compatibility issues.
Your Matchmaker URL will always remain the same after restarting.
Create separate matchmakers for dev & production environments to safeguard against human error.
Start matchmakers at least 3-7 days before release to propagate DNS records to all ISPs globally. Quick restart doesn't require DNS update propagation and applies changes with 5-10 min downtime.
To prepare for zero-downtime updates we recommend creating at least 3x copies of your matchmaker ahead of time, for example green, blue and orange. Each game client version will create tickets for one of these, rotating after each update.
Multiple app versions can use the same server image tag, removing the need to restart your matchmaker when rolling out a new server version. Simply update the associated tag in your app version.
We strongly recommend implementing blue/green strategy for update releases.
Prerequisites:
This section assumes you’ve completed 🟠 Before Going Live.
Live game clients c1 are currently connecting to matchmaker green, pointing to app version v-green linked with image tag t1.1.0.
You’ve previously created matchmaker blue pointing to app version v-blue linked with an older image tag which is no longer used t1.0.0.
In order to release game client + server updates, you may:
Prepare new server app version v1.2.0 on Edgegap:
push a new image tag to your container registry t1.2.0,
create new app version v1.2.0,
Perform any dev tests by deploying your new app version v1.2.0:
connect your game engine’s Editor to the provided URL + external port,
Update unused matchmaker blue to link to your new image tag t1.2.0,
enable caching for new app version v1.2.0 , enabling cache for this version will ensure the image is also cached for version v-blue since they reference the same tag,
Update your new game client c2 to use the new version v-blue when creating tickets:
update your base URL and Authorization token in game client,
Perform QA tests and final verifications of your new game client c2:
if you find and resolve any issues, repeat process from the beginning,
wait 3-7 days to propagate matchmaker DNS changes to ISPs globally, after the matchmaker has been stopped (quick restart doesn't require DNS updates or waiting period),
Release your new game client update c2 on game distribution platforms,
Allow time for new game client c2 to distribute to player devices (typically up to 3-7 days):
Clean up unused resources in your Edgegap account:
delete image tag t1.0.0 to free up Container Registry capacity,
delete image tag t1.1.0 to free up Container Registry capacity,
turn off your green matchmaker to pause billing until your next update.
For your next update, increase version numbers and swap green and blue keywords in the guide.
Prerequisites:
This section assumes you’ve completed 🟠 Before Going Live.
Live game clients c1 are currently connecting to matchmaker green, pointing to app version v-green linked with image tag t1.1.0.
To release a server patch without requiring a game client update, you may:
Prepare new server app version v1.2.0 on Edgegap:
push a new image tag to your container registry t1.2.0,
create new app version v1.2.0,
Perform tests and verifications by deploying your new app version v1.2.0:
connect your game engine’s Editor to the provided URL + external port,
if you find and resolve any issues, repeat process from the beginning,
enable caching for new app version v1.2.0 , enabling cache for this version will ensure the image is also cached for version v-green later since they will reference the same tag,
Update version v-green to link to your new image tag t1.2.0,
new matches will automatically initiate assignment with the updated tag t1.2.0,
Clean up of unused resources in your Edgegap account:
delete image tag t1.1.0 to free up Container Registry capacity.
Clients and servers may call API directly or with game engine SDKs, see also Getting Started.
Game clients and servers authenticate with Matchmaker using Authorization HTTP header containing your secret token (Auth Token):
Authorization: xxxxxxxx-e458-4592-b607-c2c28afd8b62This token may be included in your game client safely, as it doesn't grant access to Edgegap API.
Keep your tokens secret and safe! Edgegap staff will never ask you for your tokens.
Matchmaking API token provides unrestricted access to all matchmaking API endpoints.
Individual players can be identified using the ticket ID shared between clients and server:
Enhance security and integrate custom authorization with Server to Server (below).
to authenticate players with your own third party service acting as a proxy,
These endpoints can be used directly by game clients or with Server to Server integration.
Monitoring endpoint provides a convenient way to verify uptime of your Matchmaker Service from game clients or backend services, so you can reach out for assistance in case of urgent issues. You may integrate redundant (multiple) monitoring systems checking this endpoint for your own visibility.
Ping Beacons provide a simple yet effective way to approximate connection quality to major networking nodes and prevent high latency matches. See Matchmaker In-Depth for more.
Create Ticket (POST) - pick a matchmaking profile & submit ticket attributes:
Read a Ticket (GET) - check on matchmaking progress and host assignment.
Delete a Ticket (DELETE) - abandon matchmaking immediately.
Unity/Android - consider using raw string interpolation to prevent code stripping of hardcoded JSONs.
Join as Group endpoints allows matching deployments as a group/party/lobby:
Create Group Ticket (POST) - submit yours and your group members' tickets.
Your designated group host/leader is responsible to collect ticket attributes from players through a third party of your choice, and to distribute assignment IDs to individual group members to allow managing individual Matchmaking Process.
See Join as Group for details and Getting Started for genre-specific scenarios.
For webGL games hosted on third party distribution platforms (e.g. itch.io), sending any requests to Matchmaker from game client may result in Cross-Origin Resource Sharing policy violations. Most modern web browsers send a pre-flight request to verify that a backend service (the Matchmaker) understands and accepts communication from your game client.
Failing pre-flight check (default for security reasons) can result in one of several possible CORS-related errors, most commonly CORS header 'Access-Control-Allow-Origin' missing .
To resolve this error, add allowed_cors_origin parameter to your configuration to either:
whitelist your exact client hosting domains:
or whitelist a wildcard domain (including all subdomains):
These endpoints are used by the dedicated game server or your game backend.
Backfill allows your game server or your backend service to manage player capacity of servers:
Create a Backfill (POST) - submit current player tickets along with matchmaking profile.
Read a Backfill (GET) - retrieve ticket information about the filled player.
Delete a Backfill (DELETE) - stop backfill depending on game server logic and design.
See Backfill for details and Getting Started for genre-specific scenarios.
Inspect API provides real time matchmaking insights. The changes you make to your configuration and the impact on your players will be more predictable if you run tests on your matchmaker.
Inspect Tickets (GET) - list current player tickets, their status, and assignment (if available).
Inspect Matches (GET) - list current matches and resolved Rules values.
Coming Soon: pause and resume matchmaking at specific ticket stages.
Your success is our priority. If you'd like to send custom requests, ask for missing critical features , or express any thoughts, please reach out in our Community Discord.
Our matchmaker uses official Semantic Versioning guidelines. Each configuration JSON is tied to a specific version of the matchmaker. Matchmaker version dictates which rules and settings are available in your configuration and API. Once a version is released, it will never change.
Your configuration file will be validated depending on matchmaker version used, make sure your rules are matching the matchmaker version’s capabilities.
Stop, Edit, and Start your matchmaker to upgrade your version. Quick Restart won't do it!
This is the latest matchmaker service version, recommended for production use.
🩹 Bugfixes:
Matchmakers now correctly validate tickets with multiple profiles including different rules.
✨ Improvements and new features:
More optimizations to maximize match fill rate with player_count rule. Tickets will now wait until end of expansion (or expiration) if only partial match is possible (>min and <max team size).
Full matches (max team size reached) are made immediately (no change).
Upgrade to Enterprise ☁️ Hosting Cluster to unlock matchmaking Analytics! Gain insights into matchmaker load and performance, no code or configuration required. Metrics at launch include:
total tickets, backfills, assignments, and deployments made over custom period of time,
per minute rates for the above metrics over custom period of time,
totals and time series insights into expired tickets, expanded matches, match fill rate,
API usage metrics, and more.
Improved Rules documentation with better examples and visuals.
⚠️ Breaking changes:
Use min/max team size to fill teams efficiently (replaces player count expansions):
in your configuration player_count rule, replace team_size with min_team_size and max_team_size to achieve "best effort" matching attempting to maximize match fill rate,
to require a specific number of players per team, set both min and max to the same value,
backfills bypass player_count rule and always match with 1 ticket (unchanged).
Tickets, group tickets, and backfills with all latencies above the highest max_latency in a given profile will be immediately rejected with 400 Bad Request response to ticket create request, instead of expiring:
only applies if latency rule is configured,
to bypass this behavior, create an expansion with max_latency: 99999 (any value higher than your client latency measurement timeout).
Injected environment variables containing ticket data now include field id (ticket ID) so they can be reused more easily when creating /backfills.
🩹 Bugfixes:
Backfill now uses configured deletion and expiration period (like tickets and group tickets).
Backfill now correctly matches using configured intersection rules.
Fixed openAPI specification for POST /backfills request (requires public_ip) and GET /tickets response (team_id is optional), including examples.
✨ Improvements and new features:
Up to 3x more potential matches are considered now, producing more optimal groups and maximizing match fill rate.
Up to 200% faster matching speed due to concurrency optimizations.
Up to 40% increased match fill rate due to optimization of expansions algorithm.
Improved service stability and increased speed of quick restarts.
⚠️ Breaking changes:
Separated game profile and expansion stage information in the Injected Environment Variables:
MM_MATCH_PROFILE will now only include the profile name as it appears in the configuration.
Introduced MM_EXPANSION_STAGE which will contain the expansion stage as a string (e.g. "initial", "15", "30").
Ticket assignments now include the group ID when Matchmaker In-Depth. Group ID is also included as an Injected Environment Variables, as a mapping of group ID to a list of the group's player IDs.
Ticket assignments now include the team ID when Matchmaker In-Depth. The team ID is also included in every ticket data Injected Environment Variables.
Matchmaker In-Depth now returns 409 Conflict HTTP code instead of 204 No Content to indicate the ticket can't be deleted since the deployment is starting. To replace leavers, use a Backfill issued by the server after a pre-specified timeout period.
Matchmaker In-Depth request body parameter attributes.deployment_request_id has been moved to attributes.assignment.request_id.
Matchmaker In-Depth request body now requires full assignment details as part of attributes parameter in addition to the request_id.
🩹 Bugfixes:
Resolved intersection rule values are now Injected Environment Variables in the MM_INTERSECTION environment variable.
Quick restart feature now reliably regenerates API endpoints and openAPI specification when configuration is changed.
Fixed several bugs during matchmaker (re)start causing prolonged startup time or getting matchmaker stuck.
✨ Improvements and new features:
Increased rate limits and scalability of all API endpoints, across all matchmaker tiers.
When assigning a player to a Backfill, the new player's ticket ID will be added as a tag to the Backfill's Deployments.
Added swagger UI authentication feature to test API directly in web UI without needing postman.
Improved openAPI examples to reflect realistic requests and responses more closely.
Added new Matchmaker In-Depth meant for development and debugging purposes.
Allows listing all current player tickets in a paginated list.
Allows listing all current matches in a paginated list.
Backfill: Upon (popular) request, we’re adding backfill with automated ticket assignment, which allows you to reuse server seats when players leave the session.
Ideal for filling empty player seats after a match has begun, or for replacing players that have left during a match.
Join as Group: We’re adding the ability to join as a group to the already available ability to fill multiple teams with players.
Ideal for joining up in a matchmaking queue with a group of friends or coming from a common lobby.
To make integration easier, we’re now offering Software Development Kits for the most popular game engines.
Fixed a bug where the Matchmaker In-Depth was not applied correctly.
Tickets will now be automatically canceled after a Matchmaking Process if they haven't been assigned to a deployment.
You can now Matchmaker In-Depth to enhance the flow of your matchmaking process.
Deployments made by the matchmaker are now tagged with ticket IDs.
You can now edit your configuration while the matchmaker is running. This triggers a quick reload of your configuration without requiring a full on/off cycle for your matchmaker. Note: This feature is not recommended for production environments, as it deletes all current tickets and temporarily makes the API unresponsive.
Fixed Injected Environment Variables to use the correct primitive types instead of arrays.
Fixed Injected Environment Variables JSON values, which previously contained escaped characters.
Fixed a bug where certain headers were not accepted by the matchmaker when requests were made from a WebGL application (CORS policies).
Fixed issue with TLS/SSL certificate validation preventing matchmaker from launching.
Fixed a bug causing the beacons endpoint to return a 500 error.
Basic authentication is now mandatory for all endpoints.
Added the ability to configure the number of retries on server assignment failure.
Team-based matchmaking is now the default for all matchmaking configurations.
Both application and version are now required fields in all profiles.
Introduced a new endpoint to monitor the matchmaker's status.
Updated the format of the tickets environment variable in the deployment.
Added a configuration option to allow hosts to communicate with the matchmaker.
The debug API is now only available when explicitly enabled in the configuration (it is currently disabled for rework).
The game_profile key in the GET ticket response has been replaced by profile.
Alice + Bob + Dave may match,
Alice + Erin, or Charlie + Frank will never match.
Alice + Bob may match, or Bob + Charlie may match,
Alice + Bob + Charlie will never match.
Alice + Bob + Charlie may match, or Alice + Bob + Dave may match,
Alice + Bob + Charlie + Dave will never match.
Deployment may be retried several times if occurs,
See for a minimal example of a backfill-enabled configuration.
If you need a server browser feature urgently, consider using our API to .
Your server might need to know details about it’s players. Ticket attributes, resolved rule values, and other values are injected to your deployment, in addition to default .
Use or implement exponential jittered backoff to recover from high load.
If your players experience any issues, tracing their path to server logs can be helpful. Each Matchmaker deployment will be tagged with assigned player ticket IDs so you can easily and find to help you troubleshoot.
See to learn about deployment troubleshooting.
wait for caching indicator in version v1.2.0 to reach green,
monitor outdated game clients c1 using deployment ,
wait for caching indicator in version v1.2.0 to reach green,
monitor outdated game clients c1 using deployment ,
Include parameter player_ip in player ticket with the player’s public IP address to take advantage of and ensure lowest possible player latency:
or when creating tickets from a custom lobby/party service (e.g. ).
🎫 Player Tickets provide a simple interface for players to match together, see Matchmaking Process and .
override by using beacon IP in place of player_ip in your ticket.
and matchmaking SDKs:
30 seconds:
4 players
150 skill rating range
max 250 ms latency
60 seconds:
4 players
200 skill rating range
max 250ms latency
3 minutes (180s):
1-4 players
200 skill rating range
any latency
See Getting Started for your first steps with Matchmaker, game integration, and detailed examples.
See Getting Started for your first steps with Matchmaker, game integration, and detailed examples.
See Getting Started for your first steps with Matchmaker, game integration, and detailed examples.
{
"version": "3.1.0",
"inspect": true,
"max_deployment_retry_count": 3,
"ticket_expiration_period": "5m",
"ticket_removal_period": "1m",
"profiles": {
"casual-example": {
"application": {
"name": "my-game-server",
"version": "2024.01.30-16.23.00-UTC"
},
"rules": {
"initial": {
"match_size": {
"type": "player_count",
"attributes": {
"team_count": 2,
"min_team_size": 5,
"max_team_size": 5
}
},
"beacons": {
"type": "latencies",
"attributes": {
"difference": 125,
"max_latency": 150
}
},
"selected_maps": {
"type": "intersection",
"attributes": {
"overlap": 1
}
},
"backfill_group_size": {
"type": "intersection",
"attributes": {
"overlap": 1
}
}
},
"expansions": {
"30": {
"beacons": {
"difference": 125,
"max_latency": 250
}
},
"180": {
"beacons": {
"difference": 99999,
"max_latency": 99999
}
}
}
}
},
"competitive-example": {
"application": {
"name": "my-game-server",
"version": "2024.01.30-16.23.00-UTC"
},
"rules": {
"initial": {
"match_size": {
"type": "player_count",
"attributes": {
"team_count": 2,
"min_team_size": 5,
"max_team_size": 5
}
},
"beacons": {
"type": "latencies",
"attributes": {
"difference": 125,
"max_latency": 150
}
},
"versus_ranks": {
"type": "intersection",
"attributes": {
"overlap": 1
}
}
},
"expansions": {
"120": {
"beacons": {
"difference": 125,
"max_latency": 250
}
}
}
}
},
"challenger-example": {
"application": {
"name": "my-game-server",
"version": "2024.01.30-16.23.00-UTC"
},
"rules": {
"initial": {
"match_size": {
"type": "player_count",
"attributes": {
"team_count": 2,
"min_team_size": 5,
"max_team_size": 5
}
},
"beacons": {
"type": "latencies",
"attributes": {
"difference": 125,
"max_latency": 150
}
},
"elo_rating": {
"type": "number_difference",
"attributes": {
"max_difference": 50
}
}
},
"expansions": {
"120": {
"beacons": {
"difference": 125,
"max_latency": 250
}
}
}
}
}
}
}See Getting Started for automated ping measurement using our SDKs. Learn how to start low and increase the allowed latency gradually with Rule Expansion.
We currently offer 3 private cluster tiers to cater to everybody’s needs:
Best Suited For
enthusiasts, solo developers
commercial releases
high-traffic launches
Resources
1 vCPU + 2GB RAM
6 vCPU + 12GB RAM
18 vCPU + 48GB RAM
Redundancy
1x virtual node
3x virtual nodes
3x virtual nodes
Price (hourly)
$0.0312
$0.146
$0.548
Price (30 days)
$22.464
$105.12
$394.56
latencies is a special rule matching players within a specific difference range in ping values:
this presents a “soft” solution to splitting your player base, enabling matching with neighboring regions, especially improving match speed for less populated regions,
max_latency should be specified to discard locations above a given threshold,
Our example beacons rule above with "difference": 100, "max_latency": 200 initially:
Alice and Bob will match:
Tokyo is discarded (>200 ms),
latency for Chicago within 100 ms absolute difference.
Chicago
75.0
12.3
87.3
Los Angeles
113.2
145.6
32.4
Tokyo
n/a
n/a
233.2
253.2
Alice and Charlie will never match:
no beacons have < 200 ms latency for both players,
Alice lives in North America - Illinois,
Charlie lives in Asia - Japan.
Chicago
n/a
n/a
12.3
215.6
Los Angeles
n/a
n/a
145.6
238.3
Tokyo
n/a
n/a
233.2
24.2
Beacons are automatically rescaled in real time - adding/removing/replacing existing beacons. Your clients and backend should account for this and reload list of beacons before each matchmaking round.
Swagger Web UI: deploying a matchmaker will generate an openAPI specification and a convenient web UI. Open the URL in your browser to view and test all API endpoints, and to review payload examples.
Use Backfill to replace Seat/Match sessions. Matchmaker only supports Default session type.
To initiate successful backfill:
Server creates /backfills for each team separately, with values:
real assignment data retrieved from Injected Environment Variables (deployment).
adding backfill_group_size to tickets, listing group sizes up to available team capacity.
real tickets data retrieved from Injected Environment Variables (matchmaker),
or mock tickets data fabricated to match specific (real) player tickets,
Clients create /tickets or /group-tickets with backfill_group_size array:
including value "1" if the player is matchmaking alone, or number of group members,
additionally including value "new" if players wish to allow starting new games,
or omitting value "new" if players only wish to join a game in progress (backfill).
Matchmaking Process proceeds to assignment and clients connect to the server.
Backfills ignore player_count rule, and always match a single group. Values in backfill_group_size dictate how many players may be matched. Creating one backfill per team ensures team round-robin strategy, with server creating new backfills to fill remaining seats as needed.
Once game server initialization concludes, your server should continually:
Start idle disconnect countdown for each new player (recommend up to 3 min):
Start a loading scene/level while players connect - a full fledged 3D scene, a lobby-like social UI, or a loading screen with a progress bar, to indicate initialization is progressing.
Trigger loading for new players on server initialization.
Trigger loading for players replacing leavers (joining via Backfills).
Keep track of new player connections or existing players leaving over time:
Listen to custom client RPCs containing new player ticket IDs (initial or backfilled), to pair tickets to netcode entities/actors.
Read and verify assigned_ticket player data retrieved from /backfills.
Create new Backfills for unused player capacity (leavers).
Renew expired Backfills, which are deleted after ticket_expiration_period.
Delete any leftover Backfills once the 5. Deployment Stopped:
Unity - OnApplicationQuit callback,
Unreal Engine - OnWorldDestroyed or PreExit callback.
See Mirror Seat Management and FishNet Seat Management for player connection monitoring.
🎈 Social Game Example