Research note

Attribute projection, tested against what players actually became

Every Football Manager tool will show you a potential ability. Very few players ever reach it, and a number that most players miss is not a forecast. Attribute projection is our answer to "what will this player actually turn into?" This page shows what it does and how close that answer came: across nine twenty-season careers and 600,000 players aged 21 and under on both editions of the game, and then on a career nobody looked at until the projection was finished.

Nine twenty-season careers 600,000 young players followed FM26 and FM27 Scored once on a locked career

The feature

The whole player at his expected best

Attribute projection works out a player's full attribute panel at his expected peak, assuming he gets his football. Turn it on and the screens you already use (search, the player page, the squad planner, Value rank) read the player he becomes rather than the player he is. Any tool can put a number on a player's future. What matters is how close that number lands, so the rest of this page is the test.


The question

Potential ability is a ceiling, not a prediction

A player's potential ability says how good he could become. It does not say how good he will become, and the gap between the two is the whole of scouting a youngster. Assume every player 21 and under reaches his potential, and on our careers you would be wrong by about 16 points of current ability on average, and wrong by more than 10 for about two players in every three. Even among the youngsters who went on to play regularly, more than half missed it by more than 10.

So the claim we wanted to make was simple and testable: the projected player is close to the player he actually becomes. Close enough to plan a squad around, with an honest range around it, and measured on careers the projection was not built from.

Testing that takes two decades per career. To know what a 16-year-old became, you need him at 16 and again at his peak, years later, in the same save. Each career here was played through twenty full seasons, then followed player by player from his first season to his last.

How it was tested

Projected at one season, checked against the rest of his career

01

Project from one season

Each player was projected from one season exactly as the app would project him that day, with nothing from later seasons, and through the same code the app uses.

02

Find what he actually reached

His peak is the best current ability he reached from that season on, and his attributes at that peak. A player who left the game early is counted at what he reached before he left, and one still playing when the career ended, at the best he had reached plus a measured allowance for what was still to come. The busts are part of the test, and so are the youngsters who stay.

03

Hold careers out

A projection can flatter the careers it was built from. So each career was also scored on a version built without it, and both sets of figures are below.

04

Beat the obvious answers

A projection has to beat the free guesses: "he reaches his potential", and "he grows like a typical player his age". If it cannot beat both, it is not worth a column.

05

Keep one career locked

Every career above was looked at many times while the projection was built. So a further twenty-season FM27 career was kept locked: the pass marks were written down, the projection was finished, and only then was it scored, once, with nothing changed afterwards. That is the headline figure below.


The result

For young players who go on to play, nearly nine in ten land within 10

The projected peak is the player he becomes if he gets his football, so it is judged first on the players who went on to play regularly: 21 and under when projected, averaging 2,000 or more senior minutes a season over the next three. Everyone else is shown beside them. Each player is counted once, from the first season his career could be followed to its peak, with the projection as it ships. These figures are from the locked career, scored once.

Locked career, players 21 and under who went on to play regularly (6,694): how often each answer landed within 10 of his real peak
Within 10
Attribute projection 87 %
"He grows like a typical player his age" 62 %
"He reaches his potential" 44 %
Locked career: the projection for everyone, by how much they went on to play
Players 21 and under Players Within 10 On average they finished
Played regularly (2,000+ minutes a season)6,69487 %1 above the projection
…of them, potential 150 or more15087 %2 above
Played some (1,000–1,999)11,97985 %2 above
Rarely played (under 1,000)177,03878 %level with it
Everyone195,71179 %level with it
Locked career: everyone, by age when projected
Age when projected Players Within 10
14–16130,66576 %
17–1836,61279 %
19–2128,43492 %
22–2532,84897 %

Across all nine careers

On the nine careers the projection was built and checked on, 84–94 % of young players who went on to play regularly landed within 10, career by career, and 78–91 % of everyone 21 and under. The locked career sits where those figures said it would.

The attributes, not just the number

The projected player is a full attribute panel, not one ability figure. On the locked career, more than four in five of a youngster's projected attributes landed within a point of what he actually reached.

It ranks the future too

Sorting youngsters by their projection orders them by what they became better than sorting by today's ability or by their potential does, on both editions. That is the search a "who should I sign for the future" scout runs.

Some attributes go down

Real development is not all upward. The projection can lower an attribute as well as raise it, rather than growing everything, because that is what the game does.


The range

An expected outcome, with the spread around it in the app

No projection can be exact, because a career is decided by things that happen after the save. The app shows the projected player, and on hover it shows how far from that figure real players of the same age finished. The number is meant to be read as a centre, not a promise.

We also measured what moves a player within that range once the save is made:

Playing time

The one thing that moves a youngster's outcome consistently. Young players who went on to get their football finished a point or two of ability above the projected peak on average; those who rarely played finished level with it or a little below, and spread more widely. Set his playing time on the player page to see what it is worth season by season.

And the game's own dice

Even knowing in hindsight what happened to a player after the save barely tightens the range. Most of what is left is Football Manager's own randomness, which is part of why you play it.


The development score

Will he get there? The other half of the answer

The projection assumes a player gets his minutes. Whether he gets close to his potential at all depends on more than minutes, and that is what the development score is for. Every player gets one from 0 to 20. It reads the personality the game uses when it develops a player, with weights fitted on who actually reached their potential across our careers. Ability and potential are left out of it on purpose, because it answers "will he get there", not "how far is there".

Read the two together. A strong projection with a high development score is the youngster most likely to arrive close to his potential. A strong projection with a low score is a player who needs his football and still may not collect what he has.

Young players with 10 or more points of headroom: how many reached 90 % of their potential
Development score Seven test careers Locked career
Below 12almost nonealmost none
1419 %22 %
1645 %47 %
1753 %56 %
18+88 %73 %

Every step up the score delivers more. A 16 or 17 means about half go on to reach 90 % of their potential; 18 or above means most do.


The limits

What it does not claim

This test is of the peak

The figures here are for the player at his expected best, which is what search and the squad planner use. The season-by-season view on the player page is a separate estimate that widens the further it looks: about nine in ten land within 10 two seasons ahead, and four in five at five.

It assumes he gets his football

The peak is the player he becomes with his football. Young players who played regularly finished a point or two above it on average, so it is a little cautious for them; one who barely plays can fall short of it, and injuries and your own management can move it either way.

Past 25 there is little left to project

By then most players are at or near their peak, so the projection and "he stays as he is" agree to within a point or two. That is the right answer, not a weakness.

The youngest and keepers are the hardest

A 15-year-old is ten seasons from his peak: about three in four aged 14 to 16 land within 10. Goalkeepers 21 and under land within 10 for 71–87 % depending on the career, and about seven in ten on the locked one.

The panel can undersell a young regular

For a youngster who goes on to play regularly, the projected attributes add up to about a tenth less growth than he really gains (about a fifth less for keepers), even when his projected ability lands close. Which attributes grow is the part it reads best.

Some FM26 careers read regulars low

On three of the five FM26 careers the projected peak sat 2–3 points below where young regulars finished. It is a cautious miss, but a miss.

AI management

Every career was played under AI management, so the figures describe how players develop in the world at large, not under any one manager's training and selection.

One shipped database per edition

Each edition was tested on its own shipped database. The locked career took a game update halfway through, and development did not change across it. A major data update or match-engine change is still a reason to test again, and we will.


What is not here

The test is here. The model is not.

This page shows how the projection was tested and how well it did, including where it is weakest, because that is what you need to decide whether to trust it. How it is built stays with us, which is why the projection runs on our servers rather than in the app.