When using linear regression, does the researcher want the standard error of estimate to be small or large? Why?

(a) If the standard error of estimate is large, what does this imply about the variability of scores on Y for any given value of X? Why does this make predicting scores on Y from scores on X less accurate?

(b) If the standard error of estimate is small, what does this imply about the variability of scores on Y for any given value of X? Why does this make predicting scores on Y from scores on X more accurate?