About Premium Predictions

Team Rankings' premium numbers go far beyond what's on the standard site. We devised our power ratings to be "fair" - they are an estimate of how good teams are, based on what they've done so far this season. They only take into account the score, whether the game was home/away/neutral, and when the game was played. The non-premium predictions are essentially the difference between teams' ratings.

The premium predictions are based on a wide array of historical information. The system relies on unique predictive modeling techniques we've developed, taking into account many underrated factors. How much travelling has a team done recently? What have they done in their last ten games? Their last game? How have these teams fared against one another in the past? These are all important questions in trying to forecast a game, and the key is combining them in the right manner.

We've come up with four separate predictions:

The first is the probability of each team beating the spread. This model was originally developed on data from the 1994-2001 seasons, and it did very well the following three years. It was 69-36 against the spread for games above a certain threshold. We redeveloped the model using data from 1994-2004, and it is 35-17 in 2004-05 through Feburary 7 on games with a probability over 60%. If you take out November games, that record goes to 34-13. Another key stat: in games where the final outcome was at least ten points different from the Vegas line, the system is 17-4. 60% games tend to occur 3-5 times per week.

The second value is the score, without using any information from Las Vegas. This prediction doesn't use the spread in coming up with an expected result, so anything like injuries, trades, coaching changes, etc., will not factor in. We see it more as a tool than an accurate predictor

The third prediction is of the final score, using the spread and the over-under line. The additional information from the spread is quite useful, and you'll note that this number usually tends to be pretty close to the Vegas prediction. There's a good reason for that: Vegas generally does quite a good job at attaching numbers to a game, and it's pretty unusual to find a situation where they're way off.

The fourth prediction we call the "spread score". If this number is high, it indicates a likelihood that one team will beat the spread by a significant margin. Anything above 52 qualifies as relatively high, and anything above 53 very high. A spread score of 53, with a 50% ATS number, would indicate that a team is only 50-50 to beat the spread, but that if they beat it, it's likely to be by a lot. If there are too many numbers on the page for you, ignore this one and concentrate on the ATS number :)

Good luck! Hopefully, you won't need it.
Copyright 2000-2008 Team Rankings, LLC. All rights reserved. All content on this site is produced for entertainment purposes only and is not intended to be used in direct or indirect violation of any local, state, federal, or international law(s). By using this site, you agree and consent to our Terms of Service and Privacy Policy.