Lyte on a new feature – Restricted Chat Mode! / Poppy’s rework discussed and Akali fixes!
Welcome to another detailed discussion on the issues of Champion Select and toxicity ingame, featuring non other than lead social designer Lyte, he whose walls of wisdom you actually read.
For easier following, I’ve split all posts into categories:
- Introduction (defining toxic players and the Restricted Chat Mode)
- The Tribunal
- Matchmaking & Champion Select
- Restricted Chat Mode
- Assisting Players
- Additional Info
What does a “toxic player” mean?
In our observations and research on player behavior, we find that a single source of negative behavior can ripple through hundreds or thousands of games. For example, let’s say we have a game with 10 players–9 are positive, and 1 is negative. The 1 negative player is racist, rages all game, and intentionally leaves the game.
This experience can negatively influence some of the other 9 positive players in the game. Some of these 9 players will play another game of League of Legends and instead of being positive, they might start the game neutral or negative. Their actions can then influence 9 other players in the game, and toxicity spreads.
A single negative action could result in thousands of games with negative incidences–this is why many of our latest experiments focus on shielding players from toxicity (Restricted Chat Mode — forcing players to make a decision between using limited chat resources for cooperative communication, or still using it for negative purposes and being message capped), or stemming the spread of toxicity (our next experiment).
Is being neutral bad?
The Future of the Tribunal
Lyte: At the beginning of the Tribunal launch, Player Support staff reviewed every single case and assigned appropriate punishments. After hundreds of thousands of data points, we found that the level of agreement between players and Player Support was extremely high for the vast majority of cases; in fact, the level of agreement was over 80%.
Some of you might think 80% is a low level of agreement, but it turns out that the discrepancies are mainly because Player Support was even more strict than the average random sample of the player base. Given this data, we opted to slowly allow obvious, clear-cut cases be automatically punished while still maintaining Player Support review on the more ambiguous cases, or cases with severe punishments.
In saying this, we’re moving more and more towards experiments where Tribunal can hand out account restrictions like Restricted Chat Mode instead of time-bans. We feel account restrictions are stronger for reform because it forces players to play games on their main account to unlock the restrictions instead of simply taking a break on a smurf account. While they are playing on their main accounts, features like Restricted Chat Mode can shield other players from their toxicity while they work on improving their behaviors and becoming more sportsmanlike.
Are you trying to get rid of trash talking?
Lyte: One of the inspiring observations from player behavior research in League of Legends is how the online community has evolved over time. Think about online communities as their own infant culture, evolving over time and maturing with their own set of rules. For the first time, we can document and record every action and see how they impact the emergence of social norms in a culture.
We don’t believe that online behavior is worse than behavior in real life because players are anonymous. We believe it has to do with accountability and consequences. For decades, online communities matured without any semblance of accountability; in fact, in many cases when someone behaved negatively they were reinforced as others joined them because they saw that there were no consequences.
Very early on, a lot of players thought the player behavior team was out to get rid of offensive language or trash talking–that isn’t true. Profanity is OK, the problem is harassment and abuse. The problem is when a player directs profanity and targets someone else. We agree with the vast majority of players in League of Legends–it’s not OK to throw around homophobic, sexist or racist terms. In fact, Tribunal cases with these examples are among the most highly punished cases in the entire system. Over time, you can see that League of Legends players began to align on the idea that profanity is OK, but harassment and abuse is not–that’s cool to see.
You can say that racism or the usage of the word “f*g” has been in gaming since the beginning and is just a part of the culture. I have to ask, so what? Who says that is the way things have to be? Time and time again, the players in League have shown that they are a mature, sophisticated group that don’t care for that kind of language.
Has anyone been permabanned due to false reports?
Lyte: No one has been falsely reported enough to be permabanned, and I can’t remember the last time a permaban was overturned. I’m not saying the system is perfect, but it is highly effective at what it is supposed to do.
What type of players use the Tribunal?
Lyte: We did a research study just on players who visit the Tribunal. As many know, early in the Tribunal’s life cycle we used to reward IP (currency) for completing Tribunal cases correctly; however, as many players noted, currency incentives might not be the ideal motivator for this type of task. We removed IP rewards September, 2012, and saw a 10% drop in the number of players that used the Tribunal; however, interestingly, the accuracy went up.
When we introduced Justice Reviews, which were profile pages that showed your personal contribution to the community, we saw a 99% increase in the number of Tribunal judges; surprisingly, the accuracy of the system jumped a second time. This experiment confirmed our original hypothesis that players visit the Tribunal to make a difference in the community–not to simply grab some currency.
We also did some demographic analysis and found that Tribunal judges tended to have 3% fewer reports per game than an average random sample of the active playerbase–the vast majority of them also had 0 ban histories. This data suggests that Tribunal judges aren’t necessarily the super paragons of the community; however, they are a decent (and slightly more sportsmanlike) representation of the active playerbase. This data really inspired us because it suggested that the results and social norms emerging from the Tribunal were representative of the true community opinion.
Can pre-/postgame chat be included in the Tribunal cases?
Lyte: When we investigated pre- and post-game chat, the question we asked was: “What is the incremental value we would get by adding this feature? Would we identify more toxic players, and if so, how many more?” It turns out that many players who are toxic in pre- and post-game chat are also toxic in-game, and tend to be caught by the Tribunal. However, Riot does have access to the logs–they just aren’t hooked up to the Tribunal system. Given the choice, should the player behavior team work on Champ Select issues, or work on pre- and post-game chat? The answer is pretty clear.
In saying this, as you might have heard from other player behavior members, we have been shifting away from feature development on punishment systems. Like I mentioned in the GDC talk linked earlier in this thread, we’re focusing much of our development on making the game more enjoyable for neutral and sportsmanlike players instead of focusing on the toxic players.
Do you punish those who respond to toxic players?
This is part of the reason why we need to create systems other than the Tribunal to also tackle the player behavior problem. The Tribunal is designed to identify and reform or remove the worst of the worst. We need other systems to provide feedback and nudges to the players who are sportsmanlike 90% of the time, but may have a bad day or be influenced by a toxic player on rare occasions.
Are players warned before being chat-banned?
GG is a backlash from a bad game experience
Matchmaking & Champion Select
The issues with Champion Select
Lyte: We agree, Champ Select isn’t the ideal environment for setting teams up for success right now. I talked to a few scientists recently about Champion Select, and mentioned that it’s like putting 5 strangers together and asking them to negotiate a plan they all agree on within 90 seconds. The scientists laughed and said, “That’s a problem we’ve been studying for decades.”
There’s no easy solution to Champion Select; in fact, it might be one of the most difficult problem spaces we’ve ever had to tackle. However, it’s currently a major focus of the player behavior team, and we hope to fix the core issues with Champ Select and find a way to really build trust among strangers before the game even begins.
Why don’t you display Win/Loss ratios?
Lyte: We have considered things like showing teammates your Win/Loss records on certain champions, but there’s a red flag with this suggestion. On average, humans are notoriously bad at probability and statistics–this even includes smart people like doctors, lawyers and scientists.
For example, if you are in a Champ Select and someone chooses Jax and a 0-2 Win Loss record shows up… a lot of players would be instantly frustrated and politely (or not so politely!) ask the player to choose a different champion; however, realistically, 0-2 isn’t a significant data point. It’s meaningless.
You could argue that we wouldn’t show Win/Loss until a certain number of games have played, but this creates a similar effect where if you don’t have a Win/Loss record, then players lose trust because you haven’t even played enough games to have your record show–it’s a lose-lose situation.
What issues have you identified with matchmaking?
Lyte: Matchmaking is an interesting problem space. Let’s say that we had a perfect matchmaker, and every game you have a 50% chance to win. However, let’s say Team A and Team B play 1000 games. Team A wins 500 games, Team B wins 500 games. Technically, this means the matchmaker was quite good at matching these teams together.
However, what if every game had scores like 30-5? 25-3? If every game is lopsided, what does this mean? Players might claim that the matchmaker is completely broken. We have to separate the fairness of the potential match outcomes, and the in-game closeness of the match. In MOBAs, there’s an inherent snowball factor to the games–some games have a higher snowball factor, and some have a low snowball factor; however, in all MOBAs an advantage you build early game can lead to advantages later game that can secure victory. Snowball factor can be tuned, but it’s an extremely complicated parameter.
For example, if there’s no snowball factor, then there’s really no gravity or weight to the early game–advantages you earn are neutralized, and you don’t get the satisfaction of making plays and pulling ahead. Then there’s the fact that there is a delicate balance between building powerful champions and enjoying being powerful versus being on the losing side and suffering through the rest of the game.
Right now, the vast majority of games do have a 50% win chance. It’s highly accurate in that aspect; however, we could debate about whether the game is too snowballey and thus feels unfair or unenjoyable to play.
Restricted Chat Mode
Do you find Restricted Chat mode to be useful?
1) It is correctly forcing a subset of players to consciously think about their chat resources. At the end of the day, players want to win, and they are learning about the difference between positive chat providing a distinct advantage versus the destructiveness of negative chat. We’re mapping out the usage of Restricted Chat Mode, and are interested in seeing things like what ratio of messages are used for cooperation versus destruction. We’re also interested in seeing how many games of Restricted Chat Mode is required before a player’s personality or character fundamentally improves for the better.
2) A lot of players are self-aware of their own outbursts and rage in games and have asked to opt-in to Restricted Chat Mode voluntarily. This is interesting because for a long time, many developers and game studios have assumed that a lot (if not all) of toxic players simply lack self-awareness–they don’t realize that their behaviors are toxic, or that racism isn’t OK. However, through Restricted Chat Mode, we’re finding a demographic of players that are very self-aware, but need help–they need a nudge in the right direction, and they can’t do it alone. I often talk to the player behavior team about whether it’s our responsibility to collaborate and work with players in these cases and time and time again, we find that the answer is “Yes.”
Any plans of allowing non-toxic players to use Restricted Chat Mode?
Lyte: We’re debating it, but right now we are focused on analyzing the current effectiveness of Restricted Chat Mode and whether it is increasing sportsmanlike behavior. Until we know the impact of the feature, it would be irresponsible to create an opt-in version.
Can we help players who have been banned improve?
What about teaching new players sportsmanship?
Do you think teams other than Player Behavior will be working on this aspect?
Is your team getting any outside help?
Lyte: Up to this point, everything is done in house. We’ve created a pretty awesome team over the past year, and I’m constantly inspired by the talented people we have who were formerly top of their fields.
However, when we visited MIT/Harvard to give talks on this work, potential opportunities for collaboration were a common topic. We aren’t sure how to traverse this space yet, but we’re open to the possibilities and are talking to labs.
If you’d like to hear (and see!) Lyte discuss the issues with player toxicity in League of Legends in a more scientific way, check out his GDC presentation on the matter.
Also, if you have any helpful suggestions, you can find Lyte at: lyte [at] riotgames [dot] com
Future Changes to Poppy
What has Riot in store for our beloved Yordle?
What are your plans for Poppy?
* Rework Poppy and make her a real pick in League, balance appropriately.
* Don’t rework Poppy, and never ever buff her, and nerf her if she ever sees play.
Can you explain the reasoning behind your second choice?
Morello: Because if Poppy’s good, she supports terrible counterplay and unreadable skills with a slew of mechanical overload. Current Poppy being strong damages the game more than Poppy players get to derive joy from playing Poppy in competitive settings.
Will you keep Heroic Charge?
Akali’s glow effects return!
Recently Akali lost the glow on her hands she got from her passive. Originally she was required to get a certain amounts of AD and AP before getting access to Twin Disciplines. With the removal of this requirement, the art team saw the effect was unnecessary, but, as QA Analyst ricklessabandon explains, that part of Akali is going back on:
ricklessabandon: I followed up on this, and the removal was intentional (for reasons similar to what you mentioned). After speaking with some of the artists and the developer that made the change, we’ve decided to add them back—I’ll have the change up on the PBE with one of the next couple updates—basically our sentiments are that while the style is likely not something we’d make today, and we’d like them to have more gameplay significance.
Those are opportunities that would best be tackled in a visual update for Akali. In the meantime, we’ve decided that since Akali has had those particles for nearly 3 years, it’s fine for her to hang onto them for a bit longer until the relaunch team turns their eye onto her (which could be a long time out).
tl;dr – they were removed on purpose, but after talking about it we’re going to keep them until we can update akali’s visuals as a whole.
Don’t be a hater and likewise you won’t be hated!