@Slimebeast
Take for example a game that has the following scores:
90,85,90,85,90,90,90,95,90,90,85,90,90,90,100,100,95,90,100,90,85,90,80,85,90,85,90,90,90,100,100,100,95,95,95,90,80,85,85,80,90,90,100
If you average it (and that's 43 different scores) you'll get a 92,67 score. Even if you add, let's say, a 70 score, the score will be 92,15.
As you add scores that are kept to a normal pattern (and in my estimations I did too many variations so that my Gausian Curve was as wide as possible), it will come to a time that even adding a score that differs in a huge way from the normal pattern (the low 5% or the upper 5% of the scores), it's effects won't be felt too much.
This is what happens with reviews. As long as there are tons of 10's - 9's in a game score, you could have an odd out 8 or 7 score and it won't reflect as much on the final result, thus that score being "discredited" in the statistical analysis.