Why the market picks a better captain than form does
Nearly every captaincy tool ranks players on what already happened: last season’s numbers, a five-game form window, a fixture-difficulty rating from one to five. The sharpest read on this weekend is published every week in plain sight — the match prices — and almost nobody in FPL uses it.
Backwards data, forwards data
Form is a rear-view mirror. A striker’s five-match haul tells you what he did against five specific defences, some of which were poor, one of which parked eleven players behind the ball. It says nothing about Saturday’s opponent, and it reacts to team news roughly never.
A match price is the opposite object. It is a forward-looking estimate of this exact fixture, updated continuously by people staking real money on being right — which means it already contains the injury list, the new manager, the cup fatigue, the tactical matchup, and plenty you haven’t read yet. When the market moves a team from 1.60 to 1.45, something happened; the form table finds out three weeks later.
FPL’s own fixture-difficulty rating grades a match from 1 to 5. A price grades it to two decimal places, and re-grades it every few minutes.
From prices to expected goals
topb.in takes five prices per fixture — home, draw, away, and over/under 2.5 goals — and strips the bookmaker’s margin so they sum to sensible probabilities. A Dixon–Coles goal model is then fitted to what remains, which yields two numbers per team:
- Expected goals (λ) — how many the market thinks each side scores in this fixture, on average.
- Clean-sheet probability— the chance the opponent scores nothing, read straight off the model’s score distribution. What that number really tells you.
Those two numbers are the match layer of a captaincy projection, and they are the part form-based tools get wrong most often — because they are about the fixture, not the player.
The player layer, honestly labelled
Team-level expected goals still need dividing among the players. topb.in uses each player’s share of squad expected goals and assists, minutes patterns, defensive work, and FPL’s availability flags — and, where the market prices individual scorers, those prices sharpen the shares further. Every component of the projection is shown, not hidden.
This layer is softer than the match layer, and we say so in the tool rather than pretending otherwise. Rotation is the biggest error source in any projection, ours included. The difference is that our match layer starts from the sharpest public estimate that exists, so the soft layer has less work to do.
What this is not
The model re-expresses what the market thinks; it does not beat the market, and does not claim to. That distinction matters. In FPL you are not competing against the market — you are competing against several million people picking on instinct, highlights, and groupthink. A market-sharp read, doubled on the right player, is a real edge over them. That is the entire claim, and it is a provable one.