Dev
Competitive Gameplay - Mid Year

Thoughts on the current state of behavioral & competitive systems on gameplay and where we’re going.

DevAuthorRiot Codebear
  • Copied to clipboard

The last time we talked about the competitive direction of League was specifically about Clash. Between then and now we’ve been focused on Behavioral Systems rollouts and sustainable development of services like Matchmaking and the Leagues system. We've also been working with the League Client team to find system instabilities like champ select failures and game start hiccups, and squash them out. We have big plans for the future of competitive systems, and keeping our existing services up to date allows us to drive forward on new improvements. With that upkeep effort trending down, let’s take a look at some of the core issues we’re prioritizing on competitive gameplay.

Where Behavioral Systems is Now

In our last Behavioral Systems update, we called out dodge and AFK penalties as our focus areas and promised followup comms in mid-July, a.k.a. this blog. We recently shared a standalone update on dodge penalties (a bug set us back so changes begin rolling out next patch, 11.16) and we'll share a similar update on AFK penalties in a few weeks. As a result, the content here is on the lighter side.

Organizationally, we’re looking to further expand the size of our Behavioral Systems team on League in ways that will allow us to pick up more juicy projects. These include focusing on a sub-set of following problems within our initiative goals.

Detection and Penalties

“Reduce the frequency of disruptive behavior”

  • Further Confidence Based Reporting: We’ve seen positive results from our intentional feeding confidence methodology, and early testing indicates we'll see similar improvements by rolling this out to other disruptive behaviors like hate speech, verbal abuse, and AFKs. (For the first two, the benefits will stack with the upgrades to our text evaluation systems.) This will involve heavy collaboration with central service teams who work on all Riot's games, so our progress will be slow, but we're still aiming to have this out in 2021.
  • Verbal Abuse Detection and Punishment: With our behind-the-scenes machine models constantly actioning on intentional feeding and AFKs, we have a bit of an opportunity to look at verbal abuse issues and assess what newer text evaluation systems can bring to the table. We’re setting aside a bit of time to work with our sibling team, Central Player Dynamics, to explore our options, and we'll share more information on timelines once we identify a specific solution to pursue.

Harm Mitigation

“Reduce the impact of disruptive behavior on other players”

  • Champion Select Reporting - Penalties: We’ve talked about this across the past 9 months, and had promised that it was “coming”. Unfortunately, we’ve been unable to make that promise a reality. For now, we’re looking at what we can do to hook up champ select reports to post-game report categories (and penalties) where corresponding categories exist. Later on in the year, we hope to get our systems into place to issue penalties and mitigations directly in Champion Select. We’ll focus on immediate action in Champ Select, including deterrent solutions for game impacting picks/bans, potentially including scenarios where we can kill the lobby immediately. We don’t have a confident bead on when we can get the latter in place, but we’re prioritizing the former in penalty integration as soon as possible.

We are grateful for the patience as we scale up the team. We know this is a frustrating part of playing online multiplayer games and we’re committed to making long-term systemic solutions work for all players, not just individual instances.

Where Competitive Gameplay is now

Last preseason, we implemented a suite of improvements to Ranked that brought individual skill identification very close to where we want it to be. Across the board, we’ve seen significant improvement in match quality in ranked play where 99% of ranked matches pair opposing teams that average within half a division of each other, and teammates themselves are within ~1 division of each other. This only improves the closer you get to peak player hours during the day. 

We’re very happy with where these have landed, and now we’re ready to jump into more targeted, smaller percentile problems to polish down our rough edges. This includes making some existing Matchmaking functions more responsive to changing conditions, modifying our decay policy, and even some long overdue review of social experiences. Let’s take a look at the different topics.

Matchmaking

"Improve queue matchmaking quality without compromising queue time or availability"

  • Dynamic Position Popularity & Autofill: Position popularity fluctuates throughout the day, but our current algorithm isn’t performing to the standard we want it to. Early testing indicates that by improving the way we calculate position popularity, we can potentially reduce autofill rates as low as ~0.6% of all games, down from the current ~2-5%. Not only are we expecting to see large autofill gains, we’ve also seen strong signals that this will reduce queue times by up to 10% across all MMRs.
    That would be a huge win across the board, especially for players that specialize in one role. The main thing we need to validate before we can ship these improvements is that secondary role rate doesn’t increase substantially, but we’re committed to getting these out to players as soon as testing completes.
  • Premade Skill Balance: In 2020, we shipped Premade Parity to balance out premades across teams. We're looking at adding additional filters to specifically improve skill balance between each team's premades.
  • Dynamic Map Side Advantage: In a perfectly even skill matchup, map side advantage could mean that blue side has a very slight advantage due to meta shifts, which matchmaking corrects for by fielding a red side team with slightly different matchmaking filters. The impact of this map advantage can shift, or even reverse, from patch to patch. We're looking to make matchmaking's side correction dynamic as well, so we can automatically adjust as advantages change.

Ranked

"Improve progression satisfaction and skill expression in our systems."

  • Ranked Decay: Apex tier (Master+) and high Diamond players that take a few weeks off from League are decaying further than we’d like, down to visible ranks that are extremely off from their skill level. This creates an odd mismatch type purgatory that can be demotivating to play in. Though the intent of decay is to ensure that only the current best players are represented at the top of the ladder, it isn’t the intent to send those that have decided to take a few weeks off down to a place that they don’t belong. We’re looking into better options for decay rate that we can roll out in a low-disruption way before the end of the season.
  • Social Comparison: Ranked leagues with people you don’t know, and almost never see in games, are hard to feel a connection with. We’re experimenting with a more personal social comparison leaderboard right where you need it—directly in the ranked lobby. You’ll see the first iteration of Social Leaderboard show up soon where we can then evaluate any additional improvements to make to it based on player feedback. We’re excited to see the competition between friends and acquaintances heat up during the climb.

Competitive Rewards

"Make rewards more recognizable and relevant for time spent in League."

  • Flex Rewards: Since last year, we've seen more and more players spending a significant amount of time and effort climbing with their group in Flex. This season, we’re going to be offering two versions of each per-tier chroma, one for Solo/Duo and one for Flex. This change better differentiates these two achievements and offers teams that climb in Flex something to shoot for that sets them apart.
  • Clash Rewards Visibility: Getting together with your crew and tackling the gauntlet of Clash is thrilling, and exactly how we think League should be experienced. However, we’ve seen indications that once you lose your first two games in a tournament, it’s hard to find the motivation to continue on. Part of the reason for this behavior is that we haven’t done a great job (or even a good job) of showing you what’s still at stake, both within that specific tournament or across the entire season.
    Be it a snapshot of the trophy soon to be sitting in your award case, banners and loot display per win, or even the long-term VP track that signifies your achievement across the season, we haven’t carved out the space to really highlight what you’re shooting for. We’re in development on some small enhancements for the current season, and we’re looking to target more improvements going into the 2022 season so players know what they’re going after right from the start of the tournament

We’re buckling up for what is promising to be an exciting second half of the year. During which, you’ll see more updates coming regularly so we can keep you up-to-date with our competitive progress, as well as what new things are coming your way. As always, happy climbing, and we’ll see you on the Rift.


-Cody ‘Riot Codebear’ Germain



  • Copied to clipboard

Related Articles
Related Articles