FPL is a skill game — the evidence, and what not to learn from champions

Parts one and two of this series made a long argument for process: six disciplines, a weekly checklist, a hit test, captaincy by rank mode, chip planning. A fair reader asks the obvious question back: does any of it matter? FPL weeks are chaos. Deflections, VAR, a 94th-minute equaliser wiping out a clean sheet — maybe the whole game is a coin-flipping contest with better graphics.

It isn't, and we don't have to take the Hall of Fame's word for it.

The evidence that process wins

Public research on FPL has found persistent multi-season rank correlation — the same managers finishing high year after year at rates chance can't produce — with long-term planning and consistent decision quality identified as the differentiators. That's the statistical fingerprint of a skill game: noisy on any given Saturday, decisively non-random over 38 gameweeks, and cumulatively non-random over several seasons.

The forecasting literature agrees from a different angle. Work on public-data projection models found that open ensembles can compete with commercial services over one-to-three gameweek horizons — and, more instructive for a manager, that the features doing the predictive work are the unglamorous ones: recent returns, involvement metrics, and above all playing time. Papers on team selection as an optimisation problem and player performance forecasting land in the same place: minutes and role carry the signal; narrative carries almost none. One study's polite finding that news text alone is a weak predictor is the academic version of what part one said about tilt — the story around a player is the least reliable thing about him.

Notice how neatly the research matches the Hall-of-Fame behaviour it never studied directly. The elites plan over horizons (the skill papers say long-term planning separates them), pay for minutes (the forecasting papers say playing time is the dominant feature), and ignore narrative (the text-signal finding). Two completely different methods, one conclusion.

That's also why we're suspicious of anyone selling certainty. Weekly FPL outcomes are extremely noisy; a good process loses plenty of individual weeks — sometimes embarrassingly. What it doesn't do is lose the season. If a tool or a pundit promises you a rank, they're describing variance with a straight face. What honest analysis promises is better decisions, made repeatably — the same reason our expected-points numbers come with their uncertainty attached.

What NOT to learn from champions

This series started by calling champions partial lottery winners, so let's be precise about the traps — because champion interviews are the most-read strategy content every summer, and they're genuinely full of useful material if you filter them correctly.

• A world-number-one season usually contains a lucky captaincy run, a bench block that broke kindly, and rivals fluffing their own run-ins. Skill got them to the table; variance picked them from it. A Hall-of-Fame manager with ten elite finishes is a better teacher than a first-time champion with one.
• "Gut feeling" is not an input you can use. When a winner says they "just felt" a haul coming, translate it into things you can observe — matchup, role, minutes, game script — or discard it. Untranslatable intuition is indistinguishable from luck that got interviewed.
• A successful anti-template punt is not evidence that anti-template is right. And a successful template hold is not evidence that template is right. One season is one data point, whichever way it points.
• The better habit: extract the decision context, not the final player name. Ask why the call made sense with the information available at the time — then steal the why. The who expires; the why compounds.

The context rule that ties it all together

Ask the very best about risk and they give the same structurally strange answer: it depends what you're trying to do.

Protecting a lead, they cover rivals' captains and high-owned threats, and skip clever punts — Ali Jahangirov went into his title's final day with a literal sheet of rival danger players, and Lovro Budisin protected his lead by studying rival squads rather than his own. When projections are close, they match the field on purpose.

Chasing, they accept lower floors for real leverage — a captaincy the rivals don't hold, one or two transfers toward genuine differentials. But the differentials must have projection behind them; being different is a means, never the thesis.

In between — which is most of the season — they largely ignore rivals and maximise expected points while keeping chips and transfers flexible, switching to rank-aware play only when the run-in makes it matter. Jon Ballantyne describes risk as exactly this kind of dial: something you turn up or down against your objective, at a known cost in expected value.

In other words: the same gameweek has three different correct answers depending on where you stand. That's not a flaw in the advice — it is the advice, and it's why our recommendations change with your mini-league position instead of handing everyone the same list.

None of this series has been glamorous, and that's the point of it. The champion's season makes the better story; the Hall of Fame's Tuesday-night checklist makes the better template. Steal the process, not the picks.