Super Bowl Predictions Summary

LAST UPDATE: Saturday, 1/26

Based on a point spread of New England -12, analysis at default settings shows New England as the strong favorite to win the game, with win odds of 88.5% and an expected margin of 13.5 points. Similar games analysis indicates that the Vegas spread is efficient, as neither team has greater than 55% odds to cover. (There is a 3-5% margin of error at play in the GameZone calculations.)

ALERT: The GameZone tool's analysis methods assume this is a home game for New England. As there is not enough relevant data to base predictions solely on historical neutral site NFL games, one must use other data to try and assess the relative impact of both teams playing away from home, and adjust the baseline predictions accordingly.

Notes from the Nerds: Analytically, we can investigate the potential impact of a neutral site using the TeamRankings.com home/away power ratings for this season. As we all know, the Giants have excelled on the road, especially recently; their away power rating is 44.2, good for a #2 ranking. However, guess who is ranked #1 in away power rating for the season? The Pats, of course (rating of 51.5). Mathematically, considering the entire season, the advantage is still with New England.

At the same time, the Giants, who have a horrible home power ranking (#20) and an abysmal home advantage of -9.7 points (last in the NFL), nearly upset the Pats at home in Week 17. Which begs the question...was that game just a wild anomaly, or the start of a longer term trend of overperformance? And if so, just how irrelevant are the Giants' results from the earlier part of the season, today?

This is not an easy problem, but the good news is that the GameZone tool gives you a quick way to investigate the matchup from several angles. If you believe the Giants' recent momentum is of strong importance, you can set the time filters for factors like Points Scored and Average Margin of Victory all to "Last 3" (games), and increase the importance of those factors using the sliders. Then see how the predictions change based on the historical matchup data.

Here are some thoughts on potential changes to make to the default settings:

• DECREASE importance of Distance Traveled, since both teams are going a similar distance to Phoenix, and have many days to acclimate
• INCREASE the importance of Week of Game to the highest level if you want to focus the tool on historical playoff results
• SET time filters for factors to "Last 3" where applicable and increase their relative importance if you want to adjust for momentum
• DECREASE the importance of season power ratings one notch to adjust for momentum; just remember, these are powerful stats, so we can't advise ignoring them!

One more piece of data you should know about, and that is not yet included in the GameZone tool, is our Last 10 power rating. You can find this data toward the bottom of the Patriots and Giants team pages in the NFL section of our site. Filtered for away games only, the Last 10 Power Ratings for the Giants and Pats are extremely close, and by this measure New England is only the #4 ranked team in the NFL. So the Pats' relative advantage in away games diminishes when only the last 10 games are considered. However, remember that the sample size of games included in this measure is smaller, so this is a less "accurate" statistic than season-long measures.

(By the way, you will also notice that the prediction listed at the bottom of each team page favors New England at 76.4% win odds by 9.9 points. This prediction is 100% based on power ratings and not on the more diverse range of data included in the GameZone predictions tool. Subtract the Patriots' home advantage of 1.4 points, and the power ratings only method produces a prediction of New England by 8.5 points. That is the "old school" part of our site, but another interesting data point nonetheless.)

As you experiment, you should also scroll down the list of historically similar games identified by the tool, and look to see how the team like the Giants did specifically in games where the underdog faced a spread in the range of +12 (there aren't many) and how that compares to the predictions at the top. At the tool's default settings, the team like New York has covered more than twice as many times (9-4) as the the team like New England in historically similar games with spreads of 10 or greater. However, the team similar to the Giants only won two of those games straight up.

So what is the Nerds' final call? First, there is no way in heck we'd ever pick the Giants to win. They do have a chance for an upset -- across various angles, the tool has New England at 80-85% odds to win. So the Giants will win one out of every five or six of these matchups in typical scenarios, maybe a little more than that on netural turf. Still, we'll play the odds on that. Even if the momentum and situational aspects not measured by the tool are in the Giants' favor, it is hard to imagine them making New York an expected winner. If they pull it off, more power to them.

Regarding the spread, we would not bet this game. We simply don't have the level of relevant data we need to make a very educated decision. Regarding our $20 fun wager, however, right now we'll take the Giants to cover. We're apt to give them some extra credit for their recent performance in ways not measured by our data, and for their Week 17 performance against the Pats. We'll say slight favorites at +12, a strong pick at +13, and if you got them at +14...we may even have laid down our pocket protectors and protractors on that one. This assumes no injury to Tom Brady, of course, although we love all the theories and gossip surrounding that situation.

We'll be updating this analysis every few days as the game nears. Remember that injuries, weather, and a whole bunch of other stuff aren't considered by the GameZone predictions tool. That last 10%-15% of making a prediction is often where the most fun lies...

If we had to make a final score prediction based on the scores of similar historical games, we'd say Patriots 31, Giants 21.