I just saw a pre-game tip about "players who curse lose 13% more games" or something. I was just wondering what statistical test you ran for this, and what your alpha, etc. values were or if you checked for confounders and such. It would help me with a school thingy.
**You can check out the experimental design of that system here: **http://gdcvault.com/play/1017940/The-Science-Behind-Shaping-Player
There's no point running significance testing on some of these studies because each condition has over 10 million data points. When you're working with this type of data, you need to rely on other measures to determine true differences. When you're also measuring "true" population values (so literally you have every data point on a population), the actual statistics applied are very different than in academics.
Finally, 10% of all games always had no tips at all, which acted as a control condition. Because the analysis was all games that had a specific tip versus games that had no tips, and they were randomly assigned... most confounds were taken care of.
We've also presented this work at academic conferences, which you can look up too.
I swear in all my games... not at my team mates, but when I screw up or when someone makes a great play and deserves a "HELL YES" - I think that stat is likely more something akin to 100% of people that drink soda will die, because everyone dies.
It's probably just a simple comparison test,
Just pull out the match history profile of the entire server, filter out the matches that don't involve *'s in the chat history.
Then compare the win-loss ration between the 2 segments, the non-swearing and the swearing games.
Done.
Of course there are more complex ways to validity the stats, making sure that it only uses the matches where the cursings are typed in harmful instead of joking manner.
But I doubt they did it, considering these stats were not actually there to prove a point, but to promote certain positive mindset to the player so I wouldn't be surprised this stat is completely off by a large margin in reality, hell there might not even be a correlation between swearing and winning games who knows.
Oh, the actual analysis behind the tips themselves are correlational studies.
For example, take Player A. You look at all the games where Player A was toxic (i.e = says a homophobic slur) and then you compare it to all the games where Player A did not say a homophobic slur. You find a difference between those two sets of games which is your delta. You then measure that delta over the whole population.
