Expected points is the number that lets you compare a goalkeeper to a forward without lying to yourself.
Everything else in fantasy is measured in its own units. Goals, clean sheets, minutes, bonus, cards — all real, none comparable. Expected points translates every one of them into the only unit that actually settles an argument: points.
What the number means
Think of it as an average. If this exact match were played a hundred times, how many points would this player score on average?
That framing does two useful things. It tells you the number is not a prediction that he *will* score 5.2 — he almost certainly won't, because 5.2 isn't a score you can get. And it tells you what a difference between two players actually is: not "this one will beat that one", but "across many versions of this weekend, this one comes out ahead more often".
A forward on 5.2 and a defender on 5.2 are equally good picks. They'll get there in completely different ways — the forward through a goal or nothing, the defender through a clean sheet and a couple of bonus points — and both will return two points more often than you'd like. Same expectation, very different shapes.
What goes into it
An expected points figure is several projections rolled into one:
- the chance of scoring, and of assisting
- the chance of a clean sheet, and the goals likely to be conceded
- expected minutes
- bonus potential
- cards and other deductions
- and the scoring rules themselves, which decide what each of those is worth
That last one matters more than people expect. The same clean sheet probability is worth twice as much to a defender as to a midfielder, so the *same* match projection produces very different expected points depending on where a player lines up.
The four ways managers misread it
Treating small gaps as real. A 0.3 difference between two players sits comfortably inside the model's own margin of error. It is not a reason to make a transfer. If two players are within a few tenths, the tiebreaker should be something the model doesn't capture — fixture run, price, your team's structure, your own read on minutes.
Reading one gameweek instead of a run. A single-gameweek number is the noisiest version of the projection. Expected points across the next several fixtures tells you far more, and it's the horizon transfers actually operate on.
Ignoring the minutes underneath it. This is the big one. A high projection resting on a player you don't believe will start is the most common way to mislead yourself with a model. Minutes are the hardest component to forecast — rotation genuinely resists prediction — which is why our tools let you edit any player's expected minutes yourself, and his expected points update automatically to match. If you're convinced someone starts, or doesn't, you don't have to agree with the model.
Expecting it to pick the captain alone. Expected points ranks by average. Captaincy is a decision about the *upside* half of the distribution, because you're doubling it. The highest-xPts player is usually the right captain, but when two are close, the one with the higher ceiling is the better double — and that's a judgement the average conceals.
Where it's genuinely strong
Comparing across positions, which nothing else does honestly. Finding bargains — the players whose expected points are high relative to their price, which is where value actually lives. And narrowing a captain shortlist from ten names to two, which is most of the work.
It's also the fastest way to sanity-check a decision you've already made emotionally. If your intended transfer is well behind the alternative on expected points over the next several weeks, that's worth knowing before you press the button, not after.
Where it isn't
It can't know that a manager said something in a press conference an hour ago. It can't price a player's first game back from injury properly. It doesn't know your team's structure, your chip plan, or the fact that you're 40 points behind in a mini-league and need variance rather than expectation.
And it's a probability, not a promise. Projections are right on average and wrong constantly — that's what an average is.
The live numbers are on the expected points board, with editable minutes for every player. If you want to know how the projections are built, what they're anchored to, and the out-of-sample results I hold them to, that's in how my expected points model works.