Here's a little statistical summary for you math nerds.
Alabama has scored, on average, 57.7% more than our opponents have averaged allowing (e.g., if a team's scoring defense is 25.57, then we have, on average this season, scored 25.57 * 1.577 = 40.32 pts).
We have held offenses to, on average, 30.4% of what they average scoring. So, a team whose scoring offense is 23.14, we have, on average this season, held them to 23.14 * .304 = 7.03 pts).
MSU's offense averages scoring 17% more than their opponents allow on average (ignoring the non-FCS schools).
MSU's defense is holding offenses to 64.5% of their average.
Utilizing those stats:
Alabama's offense averages 41 ppg. MSU, on average, holds opponents to 64.5% of that. Based on that, we would expect to score 26.44 points. Their defense averages allowing 16.4 ppg. We have, on average, scored 57.7% more than that. Based on that, we would expect to score 25.9 points. So, either way, we would be expected to score 26 points. Give some leeway for a fluke score or whatnot.
The same analysis yields 9.7 and 10.4 points for MSU - or about 10 points, give or take. So, the predicted score would be 26-10. There's not a true statistical margin of error, but I usually consider 7 points to be the margin of error for each team's score.
A lot of flaws with this analysis (one outlier can have a significant effect on the results given such a small sample, it doesn't take into account injuries or oddities about the games, it doesn't account for team trends like if a team is improving each week, etc.), but it is one of the better simple ways to evaluate a "predicted score." This analysis for the Tennessee game gave a predicted score of, I believe, 49-10.