Update on the reliability of detection and reporting.
Riot has re-announced how work is underway on systems to lower toxicity and improve the overall League of Legends experience. This time there are actually two interesting pieces of information.
One concerns “spam” news reports. Now the system will ignore reports from people who use them in every game. In addition, toxic people will be able to return to LoL if their behavior improves.
In addition, many features are still being worked on to make gamers’ lives easier. For example, less points loss etc. Below, full information directly from the developers.
Goals and update their status
A few months ago we started our journey with the goal of going away from the computer and intentionally losing:
- Mitigating the effects of negative behavior, if it happens.
- Early surrender (in progress).
- Compensation for PL: Providing compensation for PL in ranked games where there are confirmed instances of behavior such as stepping away from the computer and deliberately dying (pending).
- Discourage attempts to engage in these negative behaviors.
- More severe penalties and queue block in cases of confirmed departure from the computer (introduced in 10.19).
- 10% increase in penalties for leaving the computer on a global scale.
- The increase in penalties is related to the overall decline in computer abandonment by region.
- Improved on intentional losing (in progress – probably in patch 10.21).
- All behavior detection reliability update (pending – details below!).
In patch 10.21 you will see an improvement in the detection of intentional dying and malicious play. We analyzed our current detection system that identifies players who die multiple times and have multiple deaths. The update we’ll be rolling out in patch 10.21 looks at the details of each death to pick the ones that look intentional. Thanks to this, even in the case of players with only a few deaths, we are able to distinguish deliberate dying from playing under unfavorable circumstances.
The goal of any detection relevance update is to meet one or more of the following goals without compromising the other’s metrics:
- Better than before detecting policy violators.
- Detect policy violators faster.
- Severity reduction for wrongly punished innocent players.
We remain focused on punishing significant violators after one game, but certain nuances of behavior require several games for us to be able to impose the penalty with full conviction. In the next update, we will try to move behaviors that require observation over several games to be reasonably sanctioned to the category of behavior that we confidently punish after one game. In the longer term, we will be working on ways to catch people who are deliberately dying during the game.
Hot Topic: Reporting Reliability and Accuracy Updates and You
In our August update two months ago, we mentioned manual browsing through hundreds of games – we learned a lot during this review. Getting straight to the point, the most important conclusions are:
- Player reporting is more effective at identifying signs of more subtle / nuanced behavior than our attempts to detect these signs through theoretical considerations.
- Not all reports are equally accurate. Some players are great at consistently spotting game corruption, while others are spamming us using the report button in every match.
- We have full confidence that we are making progress in improving our detection systems based on one game, but we have found that analyzing individual games will not give us a complete picture of players’ gameplay behavior – even if each individual game is correctly assessed.
That’s why we can’t wait to introduce you to our latest detection system, which we’ve been fine-tuning for months. Using the black magic of large numbers, we completed our first draft of the new post-game behavior awareness system.
The system works by measuring and controlling variations in the player’s behavior over a long period of time. This gives us the ability to tell the difference between games you’re failing because you’re stuck in lane and games where your teammate has been hovering around your Nexus for 20 minutes. By taking advantage of a longer period of game history, we have the opportunity to more consciously decide whether to impose a penalty or refrain.
Players can expect the final version of the system to be:
- Identify players who persistently engage in game disruptions in order to be penalized more severely than the one-game detection system. The more often someone engages in spoiling the game, the more severe the penalties. This system already works for quitting and exiting penalties, but we’ll extend it to other behaviors.
- Differentiate between players who, with great sensitivity, constantly pick up behavior that spoils the game, from those who spam us by hitting the report button in each game.
- Enable players who engage in spoilage behavior to improve. One of the advantages of our new system is the ability to track changes in behavior over time. We want to bring back players who have revised their behavior, while still preventing repeat offenders from breaking the game.
We have tested this system on players who die on purpose, play maliciously and quit the game, and we have very optimistic results. Our early testing shows that we can both increase the number of behaviors detected and improve the accuracy of the detection.
In the coming months, we will be regulating this system to increase the accuracy of detection, add more behavioral tracking to the system, and implement it.