Sports ratings, win probabilities and team strength ratings
SharpModels provides benchmark team ratings and matchup probabilities across the NFL, NBA, NHL and MLB.
The models are designed to provide a consistent, comparable view of team strength, matchup difficulty and expected performance within each league.
SharpModels is a benchmark model. It is not designed to include every injury, lineup change, weather update, rest announcement or piece of team news. Those factors can be compared against the model separately.
Team ratings
Each team is assigned a dynamic Elo-style rating that updates as results are added. Ratings are calculated independently within each sport and are intended to measure relative team strength within that league.
Ratings are designed to measure underlying team strength relative to the rest of the league. Beating a strong opponent improves a team's rating more than beating a weak opponent, while losing to a weaker opponent has a larger negative impact.
Sport-specific models
Each sport is modelled separately because leagues behave differently.
Home advantage, scoring patterns, schedule structure and result volatility are not the same in the NFL, NBA, NHL and MLB. SharpModels uses separate calibrated settings for each sport rather than applying one universal formula.
Comparing ratings across sports
Ratings should be compared within a sport rather than across different sports.
Each league uses its own calibrated model parameters, including home advantage, rating updates and probability conversion settings. As a result, the rating scale naturally develops differently in each sport.
For example, the difference between the strongest and weakest teams is typically much larger in the NBA than in MLB, while the NFL and NHL tend to sit somewhere between the two.
This does not mean one sport has "better" teams than another or that a 1700-rated NFL team is equivalent to a 1700-rated NBA team. Ratings are designed to measure relative strength within each league.
The most meaningful comparison is the rating gap between teams in the same sport, as this is what drives the matchup probabilities and projected margins shown throughout SharpModels.
Historical data coverage
SharpModels ratings are built using historical game results and are maintained across multiple decades of data.
Current historical coverage begins with:
- NFL: 1970 season
- NBA: 1976 season
- MLB: 1969 season
- NHL: 1968 season
Historical ratings, matchup probabilities, standings, results and trend analysis are calculated using data from these seasons onwards.
Home advantage
Home advantage is included in the matchup calculation and is calibrated separately for each sport.
Historical results show that home advantage varies by league. For example, NBA home court advantage has historically been stronger than NHL home ice advantage.
Neutral-site games are treated separately and do not receive standard home advantage.
Win probabilities
For each matchup, the model compares the two team ratings, applies any relevant home advantage, and converts the rating difference into a win probability.
These probabilities are calibrated against historical results for each sport.
A team shown as a 60% chance is not expected to win every game. It means that across many similar matchups, the model would expect that team to win around 60% of the time.
Projected margins
SharpModels also shows projected average margins where appropriate.
The projected margin is the model's average expectation for the matchup. It is not a predicted final score and it is not designed to represent every possible outcome.
For example, a projected margin of Team A by 4.5 means the model's average expectation is that Team A is around 4.5 points stronger in that matchup.
Projected margins should be viewed as average expectations rather than expected winning margins in every individual game. Actual results can vary significantly around the projected value.
Season tracking
Matchup pages include charts showing how model probabilities have changed through the season.
These charts allow users to compare how team strength and matchup expectations have moved as new results have been added.
Rating pages also track changes over time, showing which teams are improving or declining relative to the rest of the league.
Model calibration
SharpModels ratings and probabilities are tested against historical results.
Home advantage values, probability conversion functions and other model parameters are periodically reviewed to ensure that rating differences continue to produce realistic probabilities.
The goal is not to predict every game perfectly, but to maintain a well-calibrated benchmark that performs consistently over large samples of historical results.
What the model does not include
SharpModels does not currently make manual adjustments for team news, injuries, lineup changes, weather, travel disruption, player availability or late tactical information.
This is intentional. The model is designed to provide a stable benchmark that can be compared with those factors, not to replace them.
Users may choose to incorporate additional information into their own analysis, but the published model outputs remain objective and rules-based.
Future development
Season simulations, playoff forecasts and championship probabilities are planned for a later stage.
The current version focuses on team ratings, matchup probabilities, projected margins and season-to-date probability tracking.
Future enhancements may include conference projections, playoff simulations and additional team performance metrics built from the underlying rating systems.
Important notes
SharpModels ratings, probabilities and projected margins are statistical estimates, not guarantees.
Sports outcomes are inherently uncertain and individual games can differ substantially from model expectations.
The purpose of SharpModels is to provide a transparent, consistent and data-driven benchmark for evaluating teams and matchups over time.