Leaverbuster punished for unstable connection issues - And a computational proposal to improve this
I have a fairly unstable connection because of where I live. This results in eventual game crashes and some difficulties loading into the game. I have now been punished for "leaving games", and the message displays my behaviour being "a punishable offence".
Regardless of whether trying to play with a bad connection is a negative attitude or not, I believe one thing is clear: there is a clear difference between toxic players who ragequit or afk trolls who load into the game but do not actually play, and people who load into games late because their connection fails during loading screen. And the game does not acknowledge this. In fact, you are not technically "leaving a game" if you load late into it.
I think anyone who has been punished in this way immediately think how unfair it is to report really toxic players and not receiving a notification that they have been rightly punished, and yet being themselves punished for trying to play a game in unfavourable circumstances.
To ammend this, and inspired by my background and current studies, I suggest that the developers create some computational tools based on statistical learning and machine learning to automatically analyse and classify the players' behaviour when afk's or even chat conversations in game take place. These techniques' scope is already incredibly wide and it is still a growing field. I think it is quite possible, provided a sufficiently large training data set (and game chats and ping records provide such a set), to let the infrastructure of the servers to learn of different behaviours, thus leading to a more organised (and ultimately fairer) system of punishments and bans.
Some quantitative-qualitative ideas at the top of my mind:
- Correlation between certain words issued on chat and the number of bans this player received after that game.
- Ping distribution over the game (not time series, but only basic statistics) of afk'ed players.
- Some of the players' actions in-game, such as feeding or unuseful pinging and its correlation to teammates' response.
I will finish by re-stating the motivation behind this: to analyse, in a more structured, fairer and justified way, the players' doubtful actions in-game and respond accordingly in an automated way that is capable of dealing with the vast volume of games played every day.