Fantasy Premier League: what a model can tell you about captaincy

Every gameweek, the same question lands in every WhatsApp group: who is the captain? It is asked as though there is one right answer waiting to be found, and the tipster economy is built on pretending to have it. A model will not play along. It will not hand you a captain. What it hands you instead is more useful: expected points, and the spread around them. That is a better way to think about the armband than any tip, and it is worth understanding why.

Expected points, not a name

Expected points (xP) is the average FPL score a player would post if you could replay the same fixture hundreds of times. It is a blend: the chance of a goal, of an assist, of a clean sheet, plus the near-certain appearance points, each multiplied by what it is worth and weighted by how many minutes the player is likely to get. A striker nailed on for 90 minutes against a leaky defence scores highly. A rotation risk, however talented, gets marked down because half the outcomes involve a cameo or a bench.

The number is honest precisely because it is an average. It is not a promise that a player will return 6.4 points on Saturday. It is the centre of a range, and the range is where the real decision lives.

Why the spread matters as much as the number

Here is the part the “just captain him” crowd skips. Two players can share almost the same expected points and still be completely different bets, because the shape of their outcomes differs. A defensive midfielder who reliably banks appearance points and the odd clean sheet is a nailed-on 6: narrow range, few surprises. A mercurial forward who either blanks or hauls is a boom-or-bust 12: same average, far wider spread.

Same expected value, different risk. Which one you want is not a fact the model can settle for you, because it depends on what you are trying to do this week.

Fixtures lift the whole curve

The single biggest lever on expected points is the fixture. An easy game raises the chance of goals and clean sheets, so it lifts every attacking return at once and shifts the whole range upward. This is why the fixture ticker matters more than form for a one-week call: a good player in a poor match-up can carry a lower expected points than a decent player in a soft one. The model bakes the opponent in; a hunch usually does not.

Differentials are a leverage play, not a hack

A low-owned captain is the classic high-variance move. If most managers captain the obvious pick, backing someone else is a bet on the field being wrong, and the maths only favours it when you are chasing. Behind in your mini-league, you need the swing, so the wider spread of a differential is a feature: it is the only thing that closes a gap. Protecting a lead, the same spread is a liability, and the safer high-xP pick is the sounder call.

None of this is a tip. It is a way of reading your own position honestly, then choosing the risk that fits it. The model supplies the numbers and the ranges; it does not tell you whether you should be brave this week. Only your table does that.

What it cannot do

Be clear about the ceiling. A model cannot guarantee captain points, and it never should claim to. It puts a defensible number and an honest range next to each option so your decision starts from something better than vibes. A nailed-on pick will still blank sometimes; a punt will occasionally return a hat-trick. That is not the model failing, it is football being football.