Maximum Likelihood Estimator:
We use DATA to find Probability and Probability to predict(future) DATA.
Data => Prob
Prob => Data
Take the case of tossing a coin. Lets try to predict best probability(of head) for following data
DATA : 1 0 1 0 1
1 refers to Head, 0 refers to Tails
Prob = 3/5 = .66
In cases like with Data as  or [1,1,1], our data is n0t good for P(head). So we add fake data to make it bring close to .5, just as it expected for a coin.
So when we have less data, it is good to have some fake data to put our stats in control to give us better results. Below is the summary video.