Behavior Update: AFKs & Dodges

Addressing players who heavily abuse our AFK and queue dodge systems.
Game Updates
Ashley Tsao

Hey everyone! I’m Leo Del Real, a product manager on VAL’s Competitive Systems team.

We’ve been taking a closer look at our behavior systems, starting with how they handle AFKs and other similar behaviors like Queue Dodging. We started with these behaviors because of their frequency and impact on the quality of matches and the overall match-to-match experience.

Today we’re hoping to bring y’all along for the ride and provide more insight into what we’ve seen and done as a result, and what we’re doing next.

Tl;dr:

  • What we’re doing
    • Our systems weren’t adequately dealing with small cohorts of players with frequent disruptive behavior, so we made some changes.
    • Back in 11.00, we tuned our systems to more reliably issue firmer penalties for players with frequent AFKs and other similar disruptive behaviors.
    • Then, back in 11.05, we adjusted our penalties for behaviors like AFK Remakes and Queue Dodging in Competitive to scale more quickly for players with frequent disruptive behavior.
  • What we’ve seen so far
    • Competitive: Rate of AFKs and similar behaviors has decreased by 20%.
    • Swiftplay: Rate of AFKs and similar behaviors has decreased by 6%.

Let’s get into the details.

Behavior System Overview

It’s been a while since we’ve talked about how our behavior systems work. So before diving into anything else, here’s a quick rundown.

  • Our systems act more firmly when players have repeat instances of disruptive behavior and generally give players warnings on one-off instances.
  • We use ratings to track behavior over time, with different groups of behaviors measured through different ratings.
  • Ratings go down when a disruptive behavior is detected and slowly recover by playing games without having disruptive behavior.
  • Instances of a given behavior can have different impacts on rating, i.e. AFKing for 2 rounds has a different impact than AFKing for 6 rounds.

Why we took a closer look

Behavior ratings are extremely high across the board, and while this seems like a good thing, it didn’t quite match what we were hearing from players or seeing ourselves.

What we noticed

Turns out, our systems weren’t as good at tracking patterns of behavior over time in specific instances. This allowed a small percentage of players to avoid receiving the right level of penalty for their behavior.

This most notably happened for two cohorts of players that often overlapped:

  1. Players who had frequent, but less severe instances of disruptive behavior. Imagine a player who AFKs for 1-3 rounds every other game, but doesn’t have many larger AFKs.
  2. Players who had frequent disruptive behavior, but also played a ton of games. Imagine a player who plays 150 games in a month, but AFKs in 15 of them.

And while these are strong representations of these cohorts, they’re also prime examples of the consistent player behavior our systems should effectively address.

Now when I said small, I meant small. For AFKs, these cohorts were less than 1% of active players, but were responsible for 13% of all AFKs in Competitive, Swiftplay, and Unrated.

On the bright side, this does mean that VAL players generally aren’t disruptive. However, since it only takes one player to ruin a match, this 1% of players has a substantial impact.

How we responded

When we broke down what we saw, there were two questions we decided to focus on.

  1. Are we moving players down in rating appropriately given current behavior patterns and our behavior expectations?
  2. Are we applying the correct penalty to either deter or prevent a given behavior?

Through answering these questions we landed on two big changes:

  • Movement: Back in 11.00, we tuned our systems to reliably issue firmer penalties by more accurately decreasing ratings for players with frequent AFKs and other similar disruptive behaviors.
  • Penalty: Then, back in 11.05, we adjusted our penalties for behaviors like AFK Remakes and Queue Dodging in Competitive to scale more quickly for players with frequent disruptive behavior.

Although we liked the results from adjusting Movement for AFKs, we didn’t quite see the results we wanted for Remakes, especially in higher ranks like Immortal and Radiant, where excessive Remakes are more common. This signaled a need for firmer Penalties to disincentivize behaviors like spamming Remakes or Dodges to avoid tough matches, win trade, or grief other players.

Unless your behavior is similar to the cohorts we described, you shouldn’t see these updated penalties. What you will see is less games with AFKs and less players repeatedly Remaking or Dodging matches.

Impact of our 11.00 changes

Since launching these changes we’ve seen a notable decrease in the frequency of AFKs and similar behaviors.

Compared to before we made the changes, we’ve seen:

  • Competitive: Rate of AFKs and similar behaviors decreased by 20%.
  • Swiftplay: Rate of AFKs and similar behaviors decreased by 6%.

You may notice Unrated and other modes aren’t mentioned here. We’re spending a bit more time looking into these behaviors in Unrated specifically to see if different changes make sense for the mode.

What’s Next

We’re keeping an eye on these changes to make sure they continue to be effective in combatting AFKs and other behaviors. Player behavior is dynamic and we need to ensure our systems are responding appropriately so players can have a consistently great experience when playing VALORANT. 

Next up, we’re looking into similar patterns in Voice and Text comms. By targeting specific patterns of behavior we can reduce the rate of comms toxicity. More info to come!