How a rating moves

What the ratings are

The 442 Power Rating is one number per club that measures how strong its results have been, allowing for who it played and where. 442ball calculates it from the results of finished matches since 2010. It is our own measure, and the data provider's ratings are never an input.

Ratings are published once a week, on Monday, after the weekend's matches. Pages show the weekly value and its history. No rating change is ever attributed to a single match.

How a rating moves

Before each match, the two clubs' ratings give an expected result. After it, each rating moves by the gap between the result and that expectation. Beating a higher-rated side earns more than beating a lower-rated one, and losing to a lower-rated side costs more. The two clubs move by the same amount in opposite directions.

The weight of a match depends on the competition: 20 for a league match, 20 for a domestic cup and 40 for a continental match. A clear winning margin adds some weight. Every value here was fitted on the results of 2010 to 2018.

Home advantage is measured for each league from its own results rather than as one number for all. Across the 32 leagues measured it ranges from 25.8 to 105.2 rating points. A match at a neutral venue carries none.

Season starts

At the start of each season, every rating is pulled 50 percent of the way toward a target made of 25 percent the league's average and 75 percent the club's own long-run level over up to five previous seasons. This allows for squad changes without forgetting what a club has been.

A promoted or relegated club keeps its rating. A club with no rating enters at the average of the clubs relegated from that league over the previous three seasons.

The 2010/11 season is a settling season. Every club started level at 1500, so its ratings carry the widest band, are shaded on every chart and are left out of each club's peak and lowest.

Comparison groups

Leagues are linked only by matches between their clubs: continental competitions, cross-league cups, and clubs moving between divisions through promotion and relegation. Leagues linked by enough matches form one comparison group.

There are four groups: Europe, South America, MLS and Saudi Arabia. Ranks are given within a group. No list ranks clubs or leagues across groups, because too few matches link them for the comparison to be fair.

League Strength is the average rating of a league's top six clubs, or top four in a league of 12 clubs or fewer, ranked within the league's group.

EuropeLeague Strength

Confidence

Every rating carries a band, widest for a club with few matches: 44.4 points either side for a club with fewer than 10 matches, narrowing to 30.4 once it has played 20 or more.

For a league weakly connected to the rest of its group, the band is widened 1.5 times. That multiplier is a judgement, not a measured value, and we state it as one.

Attack, defence, home edge and the long run

Attack and defence indices come from the same model run on goals scored and conceded. 100 is the average club in the same league that week, and above 100 is better for both.

Home edge is how many points higher a club rates at home than overall, with the same number lower away.

The Long-run Rating is the average weekly rating over a club's last five league seasons, or fewer when it has fewer.

The fitted values

A win or a defeat moves a rating by its weight times a margin factor: (ln(goal difference + 1) + b) × a ÷ (a + 0.001 × the winner's rating lead before the match, home advantage included). A draw has a factor of 1. The divisor is never less than a quarter of a, so a win by the lower-rated side counts at most four times what the same margin earns between equal sides. Fitted: a = 50, b = 0.

How the ratings were validated

The method was fitted on 2010 to 2018 and tested on 2019 to 2025, seasons the fit never saw. The test asks how well the ratings before each match anticipated its result, measured by log loss, where lower is better. The ratings are compared with two simple baselines: always favouring the home side, and last season's league table.

The ratings beat both baselines in 7 of 7 test seasons. Overall log loss was 1.0141, against 1.0726 for the home-side baseline.

Final tables: across 408 league seasons, the median rank correlation between end-of-season rating and final league position is 0.9293. 8 league seasons fall below 0.75, and each is a flag for review rather than a failure.

Champions League winners: every winner since 2011/12 was in its group's top 9 at the time of the final.

Stability: 99 percent of weekly moves are 23.8 points or smaller. One percent of weekly moves are larger than 23.8 points.

The adopted method

In October 2026 five changes to the original method were tested. One was adopted: the season-start pull toward each club's own long-run level, called Variant E. It lowered the test log loss from 1.0155 to 1.0141 and made no test season worse.

Rulings

• A change to the method is adopted only if it improves the test result overall and makes no single test season worse.
• Home and away are shown as one home edge.
• League Strength is ranked within each comparison group.
• The winning margin is not reduced for strong favourites, because the data did not support it.
• A final-table correlation below 0.75 is a flag for review, not a failure.
• Champions League winners are expected near the top of their group. Europa League and Conference League winners are shown as information.
• The 1.5 times band for weakly connected leagues is a disclosed judgement.
• Promoted clubs keep their rating.
• No rating, value or rule is ever changed to move a club toward an expected order.

What the ratings are not

• They are not predictions. The test above measures the method; it is not a forecast of any match.
• They are not the data provider's ratings, and no provider rating is an input.
• Player ratings and similarity are not published yet. This page will describe them before they appear.

Change log