Glossary
What the numbers on this site mean, in plain language. For the math behind them, see how it works.
- Power ratings
How good each team is right now, judged by its results and who it played.
A team that beats strong opponents by a lot rates higher than one with the same record against weak ones. Ratings update after every game, and recent games count a little more.
- Margin
How many points (or goals, or runs) a team would beat an average team by, on a neutral field.
+7.0 means a touchdown better than average; −3.0 means a field goal worse. It is the sum of the offense and defense numbers beside it.
- Off and Def
The two halves of margin: scoring more than average, and allowing less than average.
Both are positive when good. A team at +4.0 Off and −1.0 Def scores four more than average but gives up one more, for a +3.0 margin.
- Elo
A running score that goes up when a team wins and down when it loses. 1500 is average.
Beating a strong team earns more than beating a weak one, and bigger wins earn a bit more. The 7d column shows how far it moved in the last week.
- Strength of schedule
How tough a team's opponents have been.
Margin already accounts for it: results are judged against each opponent's own rating, so a win over a top team counts for more.
- Luck index
Whether a team has won more or fewer games than its play has earned.
Positive means results have run ahead of performance (close wins, good bounces), which tends not to last. Negative means the team has been better than its record.
- xG (expected goals)
In soccer, how many goals a team's chances were worth.
A tap-in from six yards is worth far more than a shot from distance. Over a season, xG says more about how well a team is playing than the goals themselves.
- Win probability
Each team's chance of winning, from thousands of simulated plays of the game.
70% does not mean a sure thing: a 70% favorite still loses about three times in ten.
- Season odds
Each team's chance of making the playoffs or winning the title.
We play out the rest of the season 10,000 times using current ratings and count how often each outcome happens.
- Projections
What we expect a player to produce in an upcoming game.
We simulate the game thousands of times and share each team's output among its players by their share of the season so far. They are a middle estimate with a range, not a promise, and they don't model specific defensive matchups.
- Impact
How many points per game a player adds compared with an average player at the same position.
It only counts what the box score records, so some positions are a partial view. Small samples are pulled toward average until a player has played enough.
- Score (1 to 99)
A player's impact turned into a simple grade within their position. 50 is the average regular.
Each 15 points is one standard step above or below average, so 65 is clearly good and 80 is elite.
- Depth charts
Our read of who plays first at each position, from how much each player has actually played.
They are not a team's official chart. Players listed as out drop below healthy ones.
- Model accuracy
How often our forecasts have been right, checked against real results.
We publish it so you can judge the forecasts for yourself. A good model is right about as often as its percentages say.
- Pick'em
A free game: pick the winner of upcoming games and see if you can beat our model.
Picks lock when a game starts. You are scored against the result and against the model's pick for the same game.