Apex Skill Based Matchmaking (SBMM) To Receive Major Overhaul

In a recent blog post, Resawn, the developers of Apex Legends, revealed upcoming changes to Apex Legend's Skill Based Match Making algorithm.

Photo of author

Operehtor

In a recent blog post, a dev team led by Technical Director Samy Duc (@ricklesauceur) revealed new changes to Apex Legend’s Skill-based Matchmaking system. These new changes are aimed at improving fairness and decreasing queuing times to create an overall more enjoyable playing experience for players.

https://twitter.com/Respawn/status/1615377186508374017

What is Skill Based Matchmaking?

Skill based matchmaking (SBMM) is a concept for online, competitive gaming where lobbies are formed by taking consideration of a player’s capabilities. It’s a fairly common concept that is employed by most shooters, especially Battle Royale-style games where you have large lobbies. In a nutshell, SBMM ensures that noobs play with other noobs and pros play with other pros. This way, everyone can have a positive gaming experience. You can liken this system to the different divisions of the NCAA (Division I, II, III) or driver ratings (Bronze, Silver, Gold, Platinum) in endurance racing.

What is Apex Legend’s current Matchmaking algorithm?

At a high level, Apex Legends currently considers a pre-made squad’s best player’s skill rating as well as the pre-made squad’s size when matching with opponents. While users are able to see their level, there is a hidden ‘Skill Rating’ that is only accessible by the Apex Legend’s backend, which is the key metric in determining matching. What’s interesting to note is that Respawn’s data suggests that for players below level 300, the media skill level is fairly similar. It’s only beyond level 300 that we start to see a positive relationship emerge between Skill Rating and level.

apex skillrating levels
Image via Respawn; average Skill Rating (green line) by account level bands

Currently, the Skill Rating of Apex Legend’s players looks fairly normally distributed when viewed as a continuous distribution. However, Respawn has noted that we don’t see the same ‘longtail’ in the worst players that the best players have, which creates an issue with fairness in the low rank matchmaking as the algorithm will not be able to accurately differentiate between the skill of bottom-ranked players, which will create more inconsistent matches.

For simplicity and practicality, Respawn buckets the players into discrete buckets rather than considering their continuous skill rating. This seemingly should make live matchmaking simpler as the algorithm can draw from a singular bucket queue to form a lobby as soon as 60 players are available. From the 60 players, squads are then formed. It is at this point that there is a key difference in the squad formation between Pubs and Ranked:

  • Pub squads are formed such that the average Skill Rating of a squad should be as close as possible (i.e. best player is grouped with worst player)
  • Ranked squads are formed such the difference in Skill Rating between squad members is minimized (i.e. top players are in the same squad together)

You can definitely tell that there are tradeoffs being made for Ranked — this method creates a more consistent experience for the teams while ‘stacking’ things slightly in favor of top players in the bucket.

Furthermore, for pre-made teams, the Apex Legends algorithm considers the highest Skill Rating in a squad to prevent the negative impact of adding a smurf or new player to a squad to lower their lobby strength.

What is the future of Apex Legends Matchmaking?

There are several concepts that Respawn noted when discussing the future direction of Apex Legends Matchmaking:

  • A single Skill Rating system across all game modes (with mode-specific tuning)
  • More granular Skill Rating buckets (where the current algorithm only has 4 buckets to draw from)
  • The new algorithm will be tuned to achieve 2 goals:
    • Must be predictive of actual performance
    • Make the tightest possible matches in a reasonable timeframe

From the data released by Respawn, it seems like early tests are positive and the new algorithm is achieving both goals set by the development team. To solve for fairness and timeframe, the algorithm will now use a demand forecasting model rather than queuing model (i.e. assign players to many lobbies as they arrive rather than taking the first 60 players and forming 1 lobby). Hopefully this all means that the future for Apex Legends SBMM is bright and all players will be able to experience faster and fairer matches.

If you’re looking for way to enhance your Apex Legends experience then be sure to checkout Z League’s Apex Legends tournaments, where you can participate for free and compete for some great prizes. You can also browse Z League’s Team Up feature to find likeminded players to queue together. You’ll be able to get a look at their historical performance, preferences, and playstyle to help you find the right fit.