Where FPL seasons are actually won and lost
Good FPL is mostly the absence of expensive mistakes. Very few seasons are made by a brilliant punt; a great many are lost by a run of small, avoidable errors — a hit taken in frustration, a chip burned in a bad week, a price drop you watched coming and did nothing about.
Five decisions carry most of the weight. They are in rough order of how much damage getting them wrong will do. Four of them are worked through below; the fifth, squad structure, is set once at the draft and is covered properly on the best FPL team page, which is linked at the foot of this one.
| Decision | The common mistake | Where it is decided |
|---|---|---|
| Captain | Following a gut feeling over the projection | Every gameweek, before the deadline |
| Chips | Playing one reactively after a bad week | Two windows — see the chip table below |
| Transfers | Taking a hit for a lateral move | Weekly, and it compounds |
| Squad structure | Paying for a bench that never plays | Set at the draft — see the best FPL team page |
| Price changes | Reacting after the rise, not before | Overnight, every night |
Captaincy: the biggest single swing you control
Captaincy is the highest-leverage decision in FPL. Your captain scores double, so the gap between the right armband and a blank is roughly the size of three average outfield returns — every week, for thirty-eight weeks.
The Onside model blends nine independent signals across four broad families into one expected-points score, and the captain shortlist ranks players by that score. How those signals are built and weighted is ours. Whether it works is not a matter of opinion, and the table below is the whole argument: the same players, the same gameweeks, the same rows, scored against FPL's own projection and against two naive baselines. Lower is better — it is the average number of points each method misses by.
Onside beats FPL's own expected-points model by about 15% on identical rows, and does so in 8 of 9 graded gameweeks. Every one of the 25,518 predictions behind that table was published before its deadline and graded afterwards; the full row-level file is open data.
| Method | Average miss (points) | What it is |
|---|---|---|
| Onside v5 | 1.57 | The model behind the captain shortlist |
| FPL ep_next | 1.86 | Fantasy Premier League's own published projection |
| Season average | 2.24 | A rolling average of the player's season |
| Recent form | 2.32 | The mean of the last three matches played |
Chip strategy: you have eight chips, not four
Fantasy Premier League gives you eight chips, not the four most managers plan around. You get two of each — one usable in the first half of the season, one in the second — and they do not roll over. An unplayed first-half chip is gone at the Gameweek 19 deadline, not carried forward.
That sentence is worth re-reading if you have been treating the wildcard as a single, precious, once-a-season decision. It is two decisions, in two separate windows, and the first one expires.
The windows below are the ones the game itself declares, not a convention. Note the detail most managers miss: the first Wildcard and the first Free Hit are not available in Gameweek 1 at all. Bench Boost and Triple Captain are.
Within those windows the principle is simple. Hold each chip for the gameweek where its mechanic is structurally worth most, not the gameweek where you feel worst. Bench Boost and Triple Captain are worth roughly double in a double gameweek because the underlying players get two chances to score. Free Hit is worth most in a blank gameweek, when a large part of your squad has no fixture at all. Plan around the double and blank calendar first and fit everything else around it.
| Chip | First half | Second half | What it does |
|---|---|---|---|
| Wildcard | GW2–19 | GW20–38 | Unlimited free transfers for one gameweek, no points hit |
| Free Hit | GW2–19 | GW20–38 | Unlimited transfers for one gameweek, squad reverts afterwards |
| Bench Boost | GW1–19 | GW20–38 | All 15 players score, not just your starting 11 |
| Triple Captain | GW1–19 | GW20–38 | Your captain scores triple instead of double |
Transfer strategy: a hit is a fixed cost, not a gamble
A points hit costs 4 points. That is the whole cost, it is known in advance, and it does not scale with how well the transfer goes. So the only question that matters is whether the player coming in will out-score the player going out by more than 4 — and critically, over how long.
That second half is where most managers get it wrong. A hit taken for one gameweek needs the incoming player to beat the outgoing one by a full 4 points in that single week, which is a big ask. The same hit, taken for a player you intend to keep for ten gameweeks, needs a margin of 0.4 points per week. Identical cost, completely different bar.
The clearest green light is replacing a player who will not play at all. Swapping a certain zero for someone returning three to five points a week pays the hit back inside two gameweeks and keeps paying after that. The worst hits are lateral moves between two players of similar output, taken in a bad mood on a Friday night.
| Gameweeks you keep him | Total margin needed | Margin per gameweek | Realistic? |
|---|---|---|---|
| 1 gameweek | 4 pts | 4.0 | Rarely — needs a return the same week |
| 2 gameweeks | 4 pts | 2.0 | Only for a clear upgrade |
| 3 gameweeks | 4 pts | 1.3 | Reasonable on a fixture swing |
| 5 gameweeks | 4 pts | 0.8 | Comfortable for a genuine upgrade |
| 10 gameweeks | 4 pts | 0.4 | Almost always worth it |
Price changes: small margins that compound
Prices move overnight, driven by how many managers transfer a player in or out. A heavily-bought player rises £0.1m; a heavily-sold one falls by the same.
The part that catches people out is the selling price: you keep only half of any rise, rounded down. Buy at £6.5m, watch him climb to £6.8m, and you sell at £6.6m rather than £6.8m. That is still £0.1m of real budget you did not have before, and across a season those margins are the difference between affording a third premium and not.
The practical rule is that you gain nothing by reacting after a rise. If you already intended to buy someone and he is rising tonight, buying tonight is free money; if you are only buying because he is rising, you are buying for the wrong reason. The price-change tracker shows net transfer direction so you can tell those two cases apart.
Mini-league tactics: when to chase, when to hold
Head-to-head mini-leagues and classic leagues require different strategies. In head-to-head, you can take more risk in weeks where you are heavily losing to your opponent — the break-even maths are better when a big week can flip a two-game losing run. In classic mini-leagues, the goal is to trail the leader by fewer points than your ceiling allows you to recover.
Check the Onside mini-league hub to see your differential coverage — how many of your players do your rivals not own. A high-differential team that underperforms one week can leapfrog a template heavy squad the next week when the differentials fire. Never sacrifice all your differentials for the template after a bad week.