What are expected minutes in FPL?
Expected minutes are the average number of minutes a player is projected to play in a fixture. The word average matters. Nobody actually plays 63 minutes. He starts and sees out the game, or starts and comes off, or comes on late, or never appears. Expected minutes fold those outcomes into one number, weighted by how likely each one is.
SquadPilot models minutes first and apart from points, because most points error is minutes error. Get the minutes wrong and every downstream number goes with it: goals, assists, clean sheets, bonus. That is why the public scorecard reports minutes error on its own line rather than burying it in the points figure.
How is start probability different from expected minutes?
Start probability is the chance a player is named in the starting eleven. Expected minutes is what that chance is worth once you account for how long he stays on and how often he appears as a substitute when he does not start.
A lot of FPL mistakes live in that gap. A nailed centre-back and a rotation forward can share the same start probability and still be very different assets: the centre-back plays ninety when he starts, the forward comes off on seventy. So the app shows four things rather than one:
- Start probability: how likely he is to be in the XI.
- Minutes if he starts: the ninety-minute defender versus the sixty-minute striker.
- Cameo probability and cameo minutes: what a bench role is actually worth.
- Rotation risk and role confidence: how far the model trusts its own read.
Why are goalkeeper minutes winner-take-all?
A club plays exactly ninety goalkeeper minutes per fixture and starts exactly one keeper. That sounds too obvious to write down, and naive models break it every week. Give the first choice an 80% start and the backup a 40% start and the club is suddenly projecting more keeper minutes than a match contains.
So the keeper start gets assigned rather than shared. One goalkeeper carries the club's start and the other carries the remainder. The evidence for who that is already sits in the data the pipeline holds: the market prices the starter, and managers own starters. That read is sharpened toward the number one and capped below certainty, because a keeper who looks "nailed" can still be rested, injured or dropped over a season. Where two keepers are genuinely competing for the shirt, the answer comes back contested, and the app says so rather than inventing a favourite.
Why must a club's minutes add up to 990?
Eleven players, ninety minutes each: 990 minutes per club per fixture. Every club's expected minutes are constrained to that total, alongside exactly eleven expected starters and exactly ninety keeper minutes. These three invariants are tested, not assumed.
That constraint is what makes an individual estimate worth trusting. If a winger's minutes can only rise when a team-mate's fall, the model cannot quietly rate everybody as a starter. This is the classic August failure, where rate-based projections happily rank a backup goalkeeper above a first-choice striker because nobody has played a minute to prove otherwise.
What feeds the estimate?
- A persisted, structured role prior for each player: nailed starter, rotation risk, impact substitute or backup.
- Historical starts and minutes, and current-season evidence weighted in as it accumulates.
- Official FPL availability flags: injury status and chance-of-playing percentages from the official API.
- Team news in the final hour. Inside the last hour before a deadline the read refreshes every five minutes as press-conference and line-up information lands.
External team-news sources are graded against the line-ups that actually took the field rather than taken on trust. Only a graded source is allowed to change what the model does.
How should you read a "prior only" stamp?
Every expected-minutes figure carries an evidence freshness stamp. Early in a season, or just after a transfer, a player's estimate may be marked prior only. That is the model telling you it is running on history and role rather than on observed minutes, which is exactly what you want to know before you spend a transfer on it. A fresh observed estimate and a prior-only estimate are different grades of evidence.
What does this mean for your transfers and captain?
Expected minutes are the first input to expected points, which are the first input to every recommendation in the app. In practice:
- Buy minutes before form. A 90-minute player on modest underlying numbers usually out-scores a 60-minute player on brilliant ones over a run of fixtures.
- Read the bench as minutes rather than names. Bench Boost wants a bench that actually plays.
- Never captain a doubtful starter. The captaincy options exclude anyone below a credible start probability for exactly this reason.
- Check the freshness stamp before the deadline. A number that moved in the final hour moved because team news landed.
Every frozen forecast records its expected minutes, and the Decision Ledger grades them after the whistle. When the minutes model is wrong, the scorecard says so.