Ok, so if you are doing such detective work comparing past results with what they publish later, then just count up if house edge goes close to 2.7%..
if there is enough games, you should be able to count total winings devided by total losses and you should get close to 0.097 if roulette is fair.. if not, you get lower number.. probably much lower.. so what did you find out?
This is proved cheating in concrete game, that statistics is not needed at all. And it is not single event, I have couple other examples.