Every score on a player page comes from ratings: fans’ ratings plus, for many players, one machine-made starting estimate. This page walks through how those ratings become the scores you see, where the first numbers on a page come from, and what the site does not do.
It is the long version, with a worked example, of the short summary on the About page. Read the About page
What a score is
A score on this site is not one person’s opinion, and it is not a number the operator picked. It combines many fans’ ratings of the same ability of the same player. Each ability is pooled on its own, so ratings for a player’s finishing are never mixed with ratings for their passing or pace.
There are two ways to rate. Fan mode uses broader abilities such as shooting, crossing, heading and physical, on a 0–100 scale. Scout mode splits the game more finely, into abilities such as finishing, defensive awareness, aerial duels, decisions and anticipation, on a 0–20 scale. The two modes keep separate ratings and separate scores, so a Fan-mode rating never changes a Scout-mode score, and the other way round. Goalkeepers have their own set of abilities in both modes.
Why rate abilities one at a time instead of giving one overall number? Because a single number hides the interesting part. Two players can share the same overall score and have little else in common: one a quick winger who hardly ever tackles, the other a slow centre-back who wins everything in the air. Rating abilities separately also makes disagreements specific. "He’s overrated" is hard to argue with. "His passing is a 12, not a 16" is something two fans can check against the matches they have actually watched.
If you are signed in, rating the same player again in the same mode replaces your earlier rating for that season instead of adding a second one.
The trimmed mean, step by step
The score shown for each ability is a trimmed mean. The ratings for that ability are lined up from lowest to highest, a few are cut from each end, and only the rest are averaged. The number cut from each end is 10% of the number of ratings, rounded down, and the result is shown to one decimal place.
Example: 12 Scout-mode ratings for one ability
- 1
- 10
- 11
- 12
- 12
- 13
- 13
- 14
- 14
- 15
- 16
- 19
- 10% of 12 is 1.2, which rounds down to 1, so one rating is cut from each end: the 1 and the 19.
- The 10 that remain add up to 130: 10 + 11 + 12 + 12 + 13 + 13 + 14 + 14 + 15 + 16.
- 130 ÷ 10 = 13.0. That is the score shown.
- Trimmed mean (shown)
- 13.0
- Plain average of all 12
- 150 ÷ 12 = 12.5
The difference comes down to one rating. Eleven people put this ability between 10 and 19, and one person gave it a 1. In a plain average, that single 1 drags the result down to 12.5; the trimmed mean leaves it out and lands on 13.0. The 19 at the top is left out too, even though it may well be honest. The rule treats both ends the same. It doesn’t decide who is right; it only stops the most extreme ratings from setting the number.
When trimming starts
| Ratings for one ability | Left out at each end |
|---|---|
| 1–9 | none |
| 10–19 | 1 |
| 20–29 | 2 |
| 30–39 | 3 |
Below 10 ratings, 10% rounds down to zero, so nothing is left out and every rating counts. At 25 ratings, 10% is 2.5, so two are left out at each end.
The trimmed mean makes manipulation harder; it doesn’t make it impossible. If many people pile in with the same extreme score, those ratings are no longer at the edge of the list, and the score will move. That is one more reason to read a score as where opinion sits right now rather than as a final judgement.
Daily settlement
Scores are not recalculated every time someone rates. They are settled in daily batches: the trimmed mean for every ability is worked out again from all the ratings so far, and that result is what player pages and the ability rankings show until the next settlement.
This means a new rating can take a while to show up in a player’s public score. If the latest settlement hasn’t reached a player yet, the page works out the same trimmed mean straight from the ratings, so the scores are not left blank.
Right after you submit, the rating window compares your scores with a quick plain average of everything submitted so far, yours included and nothing trimmed, so it can differ a little from the public score after the next settlement.
The overall score
A player’s overall score is the plain average of their ability scores in that mode. Every ability counts the same. There is no weighting by position, and no ability is treated as more important than another.
So read the overall as a rough summary: a centre-back’s overall includes their finishing, and a striker’s includes their tackling. Compare players through their ability scores; the ability rankings rank players one ability at a time.
Starting estimates
A player page with no ratings at all looks empty, and an empty page gives nobody anything to push back on. So many players start with a machine-made starting estimate. It is a rough first number, not a judgement, and it is there to be argued with.
In Scout mode, the starting estimate is worked out by a script from public career numbers: appearances, goals, age and height, mostly taken from Wikipedia, together with the player’s position and the level of the league they play in. In Fan mode, the estimate is the player’s Scout-mode score converted to the 0–100 scale (multiplied by five); Fan-mode abilities with no direct Scout-mode match are filled in from the closest related ones.
Public career numbers can only tell you so much. They show how often a player has played and scored, not how well they pass or how they read the game. Expect the estimate to be wrong in places, sometimes badly. That is exactly why fan ratings are built to take over from it.
In each mode, the estimate counts as exactly one rating per ability, the same weight as one fan. It is never counted twice and never given extra weight, so it fades quickly as fans rate:
- With no fan ratings, the score is simply the estimate.
- With four fan ratings, the estimate is one fifth of the average.
- With eight fan ratings, it is one ninth.
- From the tenth rating on, counting the estimate, the trimming rule applies to it like any other rating. If the estimate is the highest or the lowest rating for that ability, it is left out of the score altogether.
Estimates are only added in a mode where the player has no ratings yet. Once fans have rated a player in a mode, no estimate is slipped in there later.
Editor ratings
Esper, who runs the site, rates players too. When the operator has rated a player in the mode you are viewing, the player page shows a small "Editor rated" label next to the rating count.
An editor rating is one rating, counted exactly like any other signed-in user’s: it goes through the same trimmed mean, can be left out at either end, and is replaced if the operator rates again. The operator’s scores are not shown separately and are never used to override anyone else’s. The label tells you the operator’s view is in the mix; it doesn’t give it more weight.
One caveat: before the site opened to the public, the operator entered test ratings. Those don’t carry the label, and until a cleanup is finished some early scores still include them.
How FM tactic cards are built
FM Manager Styles uses different data from the player ratings, and the two never mix. Each card turns a real team’s season (most are club seasons) into a Football Manager 26 tactic: a formation, team instructions and a role for each position.
Most cards start from FBref (Sports Reference) team statistics for one league season. A team is compared only with the other teams in its league that season, and each style axis is a rank within that group. Ranks are used instead of a fixed scale because leagues play differently: the same number of crosses means one thing where everyone crosses and another where hardly anyone does.
The five style axes and the stats behind them
- Pressing Intensity — combines the team’s league ranks for tackles won, interceptions and fouls, to read how hard it goes after the ball.
- Defensive Line — how often opponents were caught offside per game. Teams that hold a high line catch more.
- Tempo — shots per game.
- Width — crosses per game.
- Build-Up — share of possession.
On generated cards, the figure and the league rank are shown next to each axis.
On a generated card, the formation is the shape the team used most in the league that season. Team instructions follow from the axes: only instructions with a clear signal in the data get a value, and the rest stay at the FM default, shown faded. Roles and attribute targets are suggestions for each position. They are not the real players’ attributes, and they are not scores from this site.
Not every card is generated. Cards marked "Hand-researched" were researched one by one from public sources, and national-team cards also draw on material such as UEFA and FIFA technical reports and match reports. Every card lists its sources. When a hand-researched card covers a club season that also has generated data, its five axes use the same league ranking.
Generated cards have not all been checked by a person, so cards with a known problem carry a notice, and each card says how much match data sits behind it. FM26 also can’t express everything a real team does. Chained rotations, movement triggered by where the ball is and automatic changes by score or time have no setting in the game, so a card holds only what the game can take. Your squad is different and the match engine is its own system, so a card is a starting point to test in your own save, not a promise of the same results.
What we don’t do
- No hidden formulas for players: A player’s scores come from ratings, plus at most one starting estimate per mode, worked out exactly as described above. There is no secret adjustment, no weighting by reputation and no FBref formula behind player scores. FBref data is only used for the FM tactic cards.
- No paid placement: Nobody can pay to raise a player’s score, to get into a ranking or to be featured on the home page.
- No answers: A score is where the argument stands today, not a verdict. We won’t tell you what a player’s "true" rating is, and the legend notes give hints on what to look for rather than numbers. If you think a score is wrong, rate the player and say why in the comments on their page.
Limits and plans
Many players still have few or no fan ratings, so for them the starting estimate still carries most of the weight. The legend notes are written with AI assistance; some legends also have an English note checked against Wikipedia, while the rest are only in Korean for now. This page will be updated whenever the way scores are worked out changes.
Further reading
- Ability Guide — what each ability measures
- About — who runs the site and where the data comes from
- Legends — rate the all-time greats
- FM Manager Styles — tactic cards for FM26