How The Algorithm works

Most FPL decisions get made one week at a time: who is in form, who has a nice fixture, who is everyone else buying. But the game's real difficulty is that every decision touches the next six. Selling a defender this week changes which transfers are affordable in three weeks. Rolling a free transfer now might unlock a double move before a fixture swing. Playing your Bench Boost in the wrong week can cost you the right one.

Humans are good at judging players. We are bad at holding a hundred interacting constraints in our heads at once. That second part is exactly what computers are good at.

Constraint programming, briefly

The engine behind KnightManagers is a constraint-programming solver (specifically CP-SAT, the same class of technology used for airline crew scheduling and factory planning). You give it the rules of FPL — budget, squad limits, three players per club, valid formations, transfer costs — plus projected points for every player over the next six gameweeks, and it searches an astronomically large space of possible transfer sequences to find plans that are provably strong, not just plausible-sounding.

It is not guessing or sampling a few ideas. Within the rules and projections it is given, it systematically finds the best plans available.

What the solver actually weighs

• A six-gameweek horizon. Every transfer is judged by what it does over the whole planning window, not just the next deadline. A move that wins this week but wrecks your structure for a fixture swing gets marked down.
• Free transfers have real value. Rolling a transfer is not a wasted week — it is an asset. The solver prices free transfers and money in the bank properly, so it will tell you to hold when holding is genuinely the better plan.
• Your bench is scored honestly. Bench points are not zero and they are not face value — they depend on who actually plays. We run a Monte-Carlo simulation of the real auto-substitution rules, thousands of times per plan, so a reliable 4-point bench player is valued differently from a risky 6-point one.
• Chips are scheduled, not spent. Wildcard, Bench Boost, Triple Captain and Free Hit are placed where they are worth most across the horizon, taking your actual squad and the fixture calendar into account.
• Multiple ranked plans, not one command. You get several distinct plans, ranked, with the trade-offs visible — because a good tool should show its reasoning, not issue orders.
• Edit and re-solve. Refuse to sell your captain? Fine. Lock the players you believe in, adjust the squad, and the solver re-plans around your decisions.

Why not just use gut feel or a single-week tool?

Gut feel is genuinely valuable for judging players — and useless for combinatorics. Single-gameweek tools have the opposite flaw: they answer "what is the best team for Saturday?" while silently spending your free transfers, ignoring your bank, and walking you into blank and double gameweeks unprepared. A points hit that looks bad this week can be correct over six; a "free" transfer that looks free this week can be expensive by October. You only see that with a solver that plans the whole horizon.

What it does not do

No optimiser can see the future. The solver is only as good as the projections it plans over, and football keeps its variance. What it removes is the unforced errors — the structural mistakes, the mispriced hits, the chips burned in the wrong week. You still make the calls; you just make them with the maths done properly.

See it on your own team

The fastest way to understand it is to watch it plan with your actual squad. Try the demo — or rate your current team free at Rate My Team.