Linear Regression Equation - error term

Regarding the linear regression equation of the form:

y = b0 + b1x + E

Where E = the error term Epsilon.

When using this equation to predict values for y, do I need to leave the error term in the equation? (I mean suppose E = 10, do I need to add 10 to my value of y)? Also if I do need to leave it in, shouldn't it be +/- E? The equation seems to indicate it's just + *or* - E but since it's unaccounted for variation I would have thought + *and* - would be more appropriate?

Thanks

Rob