I am currently writing a pet-project for a browser based game which consists of two parts, first part is a data miner which observes the transactions in the market. And second part is as you might guessed the analyzer.
But between these two steps I want to filter out some transactions since people sometimes use the market to transfer gold between each other and that corrupts the data when someone pays 10x average price of an item. I haven't taken probability and statistics course yet so I am not sure what to use when filtering out those.
First thing I thought of is to filter out 2x average prices but that doesn't feel right, what would you propose? On second thought I came up with filtering out average + 1 std dev. but I don't know what that border means in words, I mean I don't have the insight on it.
At the end I am willing to come up with graphs and maybe functions of things like how does the average price and the median of prices change throughout the day(when do they approach eachother and when do they drift apart), or how differently the market behaves in weekdays or weekend. This all sounds exciting to me but first I have to filter my data set.