Planning FPL transfers with a model over a four-gameweek horizon

Plan the horizon, not the week

A single gameweek is mostly noise. The signal in transfer planning appears over three to six gameweeks, where fixture swings, expected minutes and price changes compound. A four-gameweek horizon is the practical sweet spot: long enough to capture a fixture run, short enough that injuries and form drift have not made the projection fictional.

Weight the near gameweeks more heavily. A player who is excellent in GW+3 and ordinary now may not survive in your squad to enjoy it.

The hit maths

A transfer beyond your free allocation costs four points. So the move must gain more than four expected points over the period you intend to hold the incoming player, not over one gameweek. A 1.2 xPts weekly edge clears a hit comfortably across four gameweeks; a 0.6 edge does not, once you account for model error.

There are two legitimate exceptions. Fixing a squad that cannot field eleven players is worth a hit at almost any margin, because a non-playing asset scores zero rather than his projection. And selling a player whose price is collapsing can be worth it if you would otherwise be priced out of the squad you want.

Rolling is an option, not a failure

Banking a free transfer buys flexibility: two moves next week lets you execute a paired swap that fixes a structural problem rather than papering over a single slot. Roll when no move clears the threshold clearly, when injury news is still unresolved, or when a known price-change or fixture pivot is one week away.

Constraints the planner must respect

• —Your actual fifteen players, not an idealised template
• —Bank balance and the selling price of outgoing players, which is not their current price
• —The three-players-per-club limit
• —Squad structure: you still need two goalkeepers, five defenders, five midfielders and three forwards
• —Free transfers available, including rolled ones
• —Whether you intend to wildcard soon, which makes short-horizon patches pointless

Sequencing multi-week plans

The planner should output a path, not a single swap: this week move A out for B, next week move C out for D, with the price and fixture assumptions stated. That lets you see whether the plan survives contact with reality, and which step is the one that actually matters.

Treat any step beyond GW+2 as provisional. Its purpose is to stop you making a move this week that blocks a better move next week.

Common transfer traps

Chasing last week's haul is the most expensive habit in FPL: you pay a risen price for output that has already happened and is partially unrepeatable. Reacting to a single blank is the same error in reverse. And taking a hit on the final day of a fixture run wastes the entire justification for the hit.