Hi Creators,
Since our beta launch of Custom Matchmaking earlier this year, we’ve been hard at work getting it ready for prime time. We’re happy to announce that Custom Matchmaking is now available to all experiences on Roblox! You can access it through your experience’s dashboard under the new ‘Custom Matchmaking’ page.
If you’re just hearing about this feature, we recommend:
- Watching the Roblox Learn video on Custom Matchmaking,
- Reading through our last post, and
- Reviewing the documentation here.
Any one of these resources will be enough to get started, so please take advantage of whichever format is most helpful to you.
When should I use Custom Matchmaking?
Custom Matchmaking lets you adjust the primary signals used to match players and servers together, and gives you the ability to prioritize what is most important for your game. You’re also able to add your own custom signals, letting you pull player info from the experience (e.g. skill rating, character level) to match players.
Here are some common uses that we saw some from our beta testers:
-
Skilled based matchmaking: For competitive games, you can create a ‘skill rating’ that’s adjusted based on whether a player has won or lost a match, factoring in the opponent’s current skill rating. Once the data is in your data store, you can create a custom signal and assign a weight to it, to prioritize matching players of a similar skill rating.
-
Minimize latency: If your game requires more responsive replication from server to client, you can prioritize that players be matched into servers with a lower latency (i.e. ping) by increasing the weight of the ‘Latency’ signal.
-
Match players based on their selected language: If your game is co-op or requires teamwork, it can be difficult for players to coordinate and chat when they speak different languages. For games that require this high-level of coordination or are more social, you can cluster players who speak the same language by increasing the value of the ‘Language’ signal.
How you set up your matchmaking is going to be different for every experience, and it’s important to think about what’s most important to your experience. When editing your settings we suggest:
-
Make small adjustments: To avoid situations in which you accidentally make your matchmaking signals too restrictive in how players are matched, we recommend tweaking only a few signals and increasing/decreasing the weight of the signals a little bit at a time.
-
Prioritize a few signals: While it may be tempting to increase the weight of many of these signals, you should focus on a few that are most important to your experience. Increasing the weight of too many will cause your most important signals to lose their influence, because what matters are the relative weights across all signals.
-
Run experiments: Running a matchmaking experiment will help you accurately measure the impact of the matchmaking configuration on different metrics. However, an experiment is only useful if there are enough servers to pick from. If there aren’t servers you should monitor the matchmaking changes through analytics
How do I know whether my customization is working as intended?
When you’re setting up your matchmaking configuration, you can preview how your changes would impact mock data while configuring your weights to confirm matchmaking is changing as intended.
However, to predict the impact of your changes on your experience before fully rolling them out we highly encourage you to run an experiment; today we announced configs and experimentation. This also enables you to create matchmaking experiments. By going to the ‘Experiments’ or ‘Custom Matchmaking’ tab in your experience dashboard, you can click ‘Create Experiment’ to test different matchmaking configurations to see which is best for your experience based on how it affects playtime, retention, and revenue per user.
Running an experiment will help you measure the impact of a bigger matchmaking change more accurately, building the confidence on rolling a positive matchmaking change out to the entire player base.
You can read more about setting up matchmaking experiments in our documentation here.
FAQs
Will there be more matchmaking signals?
- Yes, Matchmaking team is continually running internal experiments to evaluate more signals that can produce positive impacts on matchmaking.
When new signals are added, how would they be added to my existing custom matchmaking configuration?
- If a place is using Roblox Default matchmaking, the new signal will be added to Roblox Default with its default weight. For your customized matchmaking configuration, the new signal will be added to it as weight 0. You will need to specifically set a weight for it. We will post updates to let you know when adding new signals.
How many custom signals will I be able to add?
- Currently only 2 custom signals are allowed per matchmaking scoring configuration.
How are signals like ‘Latency’ and ‘Device Type’ compared when they’re measured on different scales?
- Different signals are not directly comparable with one another. The server scoring formula is a weighted sum formula, where each signal contributes to the final score with a weight. Please refer to the documentation on Server Scoring and Attributes and signals for how the score is calculated in detail.
Can I control when players will start a new server instead of joining an existing one?
- No for now. Today, players will only start a server when all the games are full, or potentially when available servers are too far from the player.
During Matchmaking Experiments, would players in “Variant 1” be able to join players in “Variant 2”?
- During matchmaking experiments, players in different variants will be isolated from each other only for matchmaking flows. They can continue to play together when joining through direct server join, party joins, and following friends.
What percent rollout is recommended for Matchmaking Experiments?
- We recommend using 100% of traffic, because players are matchmade to servers in the same variant. Using a lower percentage would limit the number of servers to choose from, reducing the chance of getting statistically significant results.
How do I validate my Custom Matchmaking Configuration works as expected if I don’t have enough players to run an experiment?
- If there is not enough statistical power to run a Matchmaking A/B experiment, you could do a pre-post analysis by observing changes in the matchmaking signals analytics dashboards before and after your change.
