Power Ratings Explained

A power rating is a single number representing a team's strength on the scoreboard. Subtract two ratings, add adjustments, and you have a projected margin — which is a spread. Nearly every betting line in existence starts life this way.

The Basic Mechanic

Suppose Team A rates 6.0 and Team B rates 1.5. The difference is 4.5, so on neutral ground you project Team A by 4.5. Add two points for home field and Team A at home projects as a 6.5-point favourite.

Compare that with the posted line. If the book has it at -4.5, your rating says the favourite is undervalued by two points, and that gap is the basis of a bet — assuming your ratings are any good.

The scale is arbitrary but conventionally centred so that an average team is 0. What matters is that the differences are meaningful, not the absolute values.

Building Them

The core input is margin of victory adjusted for opponent strength, which is circular and therefore solved iteratively: rate everyone, use those ratings to adjust each result, rate everyone again, repeat until stable.

Refinements that matter:

Cap blowouts. A 45-point win contains less information than the margin suggests, because the loser stopped competing. Most systems cap the contribution.

Weight recent games more. Teams change. But not too much, or the ratings chase noise.

Use predictive stats rather than results where possible. Yards per play, expected points added, efficiency measures and turnover-adjusted margins predict future games better than raw scores, because turnovers and field goals are noisy.

Regress towards the mean early in a season. Three games of data is not enough to abandon a preseason estimate.

The Adjustments

Home field has declined across most sports over the past decade, and using an outdated three-point figure is one of the more common amateur errors. Quarterback value dwarfs everything else in football, which is why NFL lines move more on one position than on all others combined. See injury report betting and rest and travel spots.

Where Power Ratings Fail

Matchup specifics. A single rating cannot express that one team's strength attacks another's weakness. Ratings treat teams as scalars; games are not.

Personnel changes. A rating built on ten games with one roster does not describe a different roster.

Motivation and situation. Ratings do not know a team has already clinched or is looking ahead to next week.

Style effects on totals. Ratings project margin, not pace, so a separate model is needed for totals. See pace and totals.

Small-sample sports. In baseball and hockey, team-level ratings matter far less than the day's starting pitcher or goaltender.

Which is why ratings are a starting point rather than an answer. The professional workflow is ratings to generate a number, then situational judgement to decide whether the number is trustworthy in this specific game.

Using Them Honestly

The main discipline is knowing your own error bars. If your ratings are accurate to within two points — which is respectable — then a one-point disagreement with the market is not a signal, it is noise.

The way to find out is to record your projected number for every game alongside the closing line, then measure the distribution of differences. That tells you both how accurate your ratings are and whether disagreements above a certain size actually predict anything.

Most people never do this, which is why most people bet one-point disagreements with a market that is better informed than they are. See closing line value.

Frequently Asked Questions