I defined: X as the number of people in 1000 total people that supports Obama. P=the percentage of people that supports Obama. Therefore: E[X]=1000P, Var(X)=1000P(1-p) Pr(the result is wrong)<=Pr(|x-E[X]>=100|)<=1000p(1-p)/10000
Then, we can find the upper bound by finding the max value of the expression 1000p(1-p)/10000, the method is on the class notes.
