Gamedev Player Moderation Charts (2020)
Preface
Throne of Lies is an online social deduction game, where moderation is essential. Luckily, we boldly claim to have the most efficient moderation system among indie games!
While moderating used to severely damage online indie games due to “revenge reviews”, Steam’s new [amazing] “review reminders” fixed most of that.
Alas, here we are today, with 2 years worth of stats:
Stats (2018~Today)
^ We’ve received 97,668 player reports. Of which, 91.3% (89,184) are innocent. This indicates that a vast amount of reports are either fluff, salt, or unwarranted.
However, this does not mean that moderation isn’t worth doing: 8,484 guilty cases is still a worthy number to do them.
”The biggest frustrations are not actually toxicity, but players playing with other players that do not know the game well enough.”
^ Of the 6 categories of report reasons (leaving out the 7th “misc” category for before this was implemented), 69.4% reported gamethrowing/leaving, which indicates that the biggest frustrations are not actually toxicity, but players playing with other players that do not know the game well enough.
While the smaller %’s can actually have a significantly-larger impact on the player experience, it does not seem to happen as often as simple gamethrowing — either intentional or accidental.
Whether or not it was “real” or not doesn’t matter, as the player felt that they had a negative enough experience
Let’s not forget that <9% of these reports are actually innocent! However, even if these cases ended up being innocent, it was still a frustrating experience in the player’s eye. Whether or not it was “real” or not doesn’t matter, as the player felt that they had a negative enough experience to report someone else for it.
That’s not good! This means we must come up with a way to ensure player experiences aren’t ruined by gamethrowing or disconnection. We’re still a bit far away from social AI bots (although someone has done this on a small scale, already); perhaps clarification that players are new on top of the reinforced tutorial: The first 50 games, force-show a <Newcomer> title? Hmm… But I digress!
^ The initial theory was that we would have more warnings than bans (or rather, a suspension between 1-day and 10-years; off-topic: Yes, 1-year was not not enough for a “permaban”, as we found out the hard way).
However, this was not the case — but why? Did we not offer enough warnings? Were the average cases too “extreme” for only a warning?
Nope; turned out to be an issue with repeat-baddies!
For unintentional cases, we will generally issue warnings, first (unless extreme — you know the types of cases we’re talking about). Especially gamethrowing, we want to make 100% sure that the player knows exactly why players are reporting them. However, after 1 warning? We scale with suspensions: 3-day, 7d, 15d, 30d, then 10-years (whew): We used to issue 1-day suspensions until we realized we were a bit too lenient (and the average person may not even know they were suspended if they skipped a day of playing).
With this many repeat-offenses, we may want to consider scaling down even-more. However, we only recently dropped the 1-day suspension tier, so this may have an impact to show slightly less mercy to fight repeat-offenders.
This also is a good thing! It means the community has <9% overall toxicity/gamethrowers since the majority of these folks are repeat-offenders. Yay! Well, sort of… O_o eh….
Final Thoughts
Although these charts can absolutely be interpreted in numerous ways, especially due to the innocent %, moderation is still an important tool for chat interaction.
How would YOU interpret these charts? What would you do to improve a socially-centric game based on these charts? Comment below now!
