How Team Ratings work

The Pascal Rating summarizes a team's strength from completed games up to the date shown. Each sport has its own method: MLB ratings describe season run scoring and run prevention without opponent adjustment; NFL and college football ratings estimate opponent-adjusted team strength; NBA ratings use opponent-adjusted offensive and defensive efficiency; NHL ratings estimate opponent-adjusted strength from goals and shots on goal. Men's college basketball ratings, when available, publish overall opponent-adjusted strength only.

Every sport uses a 100-centered scale: 100 is that sport's league reference, and higher is better. Overall is the headline rating; each sport page explains how it relates to the Offense and Defense columns.

Compare teams within one sport and one view: the same season and the same date or week. Rating points are not comparable across sports, and a difference is not a percentage. In football we convert the difference into a win probability with a separate calibration, described in the next section.

A larger gap indicates a larger estimated strength difference. In the studied samples, larger pregame rating gaps were associated on average with larger later scoring-margin advantages, but single games vary widely and many higher-rated teams lose.

In the NFL and college football, we publish a win probability for each game before kickoff. The rating gap between the two teams, plus a home edge that is dropped at neutral sites, gives an expected scoring margin. The probability is how often a margin that large holds up given the typical error of past margins. The conversion was fitted on earlier seasons and is versioned (rating-win-probability-v2), so a published probability keeps its meaning.

Each football ratings page scores those probabilities against the closing line, game by game, in a season record that includes the weeks where the rating did worse than the market.

For MLB, the NBA, the NHL, and men's college basketball, we do not publish game forecasts yet. On those pages the rating describes strength only and is not itself a win probability.

Teams are ranked by Overall rating at full stored precision. Equal values share a rank and the next rank is skipped (1, 1, 3). Displayed values are rounded, so two teams that look equal may still hold different ranks. Within a shared rank, list order is only a display tie-break.

Each table caption shows the date its data runs through: results of games completed after that point are not included. It is not a promise that the ratings were refreshed today, so check it before relying on a view. Choosing an earlier season, week, or date shows the ratings stored for that point, not a recalculation that uses later results.

Updates are driven by events, not a fixed clock. Ratings change after newly completed games are processed from source data, so they do not move on days without new games or during the off-season. Processing depends on source availability and validation checks, which can delay an update.

MLB ratings appear only after a batch passes source-coverage review and is published.

• Comparing the relative strength of teams in one sport at a point in time.
• Adding context to records, schedules, and strength of schedule.
• A starting point for your own research and model inputs.

• Outside football, they are not win probabilities or forecasts of any single game.
• They are not betting advice, and we do not claim an edge over the betting market.
• They do not explain why a team wins or loses; association is not causation.
• They do not guarantee future results, and early-season ratings rest on small samples.
• They are not comparable across sports.

Each sport's unchanged method was replayed over past seasons using only games completed before each target game, then checked on a later chronological holdout the method had not seen. In every sport, higher-rated teams won more often and rating gaps were positively associated with later scoring margins.

NFL, college football, and NBA ratings exceeded the simple standings baseline on both correlation and AUC. MLB was mixed: slightly higher AUC but lower correlation, and its relationship is modest. A men's college basketball research method showed similar results; it is not in this table, and its ratings are published only once an approved data source and rating build are in place. The NHL rating was validated separately and is not in this table either: on its 2024–25 and 2025–26 holdout, the team favored by rating and home ice won 56.8% of 2,624 games. Hockey results are close to a coin flip, so its ratings separate teams less sharply.

Limits of this evidence: most of it is retrospective replay of the current methods, not a record of ratings published before those games. Only MLB has a sample of ratings actually published before games, and at 118 games it agrees in direction but is too small for season-level validation. The standings baseline is deliberately simple, and small differences are not proof of material superiority. This holdout did not compare the ratings with betting markets.

The football records do. Each season, every completed NFL and college football game with a pre-kickoff rating and a closing line is scored on the result, and the record is published on the sport's ratings page. As of October 2026, the NFL rating was ahead of the closing line on a sample of games too small to separate skill from luck, and the college football rating trailed it. Each page shows the current figures. The closing line is a strong forecast, and we report where the rating falls short of it as plainly as where it does better.

• NFL record against the closing line
• College football record against the closing line
• Track Record: every season, study, and published forecast