A razor-sharp FPL model
with nothing hidden
Sideline Quant is a Fantasy Premier League planner built on a deep expected points model with all its inner workings accessible. Use it to build your team, forecast transfers and win your mini-leagues.
No account needed. Public FPL data only — we never ask for your login.
Crystal Palace 1.7 goals expected · 32% chance of clean sheet
10% chance this player scores in the match
How does the model hold up?
Before every deadline the model names the best squad it could legally have bought with £100.0m and a max of three players from one club. After the results have all come in, that squad is then scored on what actually happened.
The model’s own entry has scored +24 against the average manager over 3 gameweeks (206 to 182: +9, +10, +5; Triple Captain played in GW1). In GW3 it predicted 680.7 points for its 199 likely starters; they scored 685.
Averages 64.0 a gameweek, with the captain included, over the first 3 gameweeks. That is 41% of the 418-point ‘hindsight ceiling’, which is the XI you could only pick after the whistle, which obeys no budget and no club limit and which nobody reaches.
A fresh £100.0m squad, bought from scratch every gameweek: it pays no transfer hits, carries nothing over from last week and banks no price rises. It is a weekly mark on the model's picks under the real squad rules — not a strategy any manager can run, and not comparable with a managed entry on equal terms.
Each row links to that gameweek if you want to take a closer look.
These are judged in season, as well as against last season's points per game — the baseline, a naive competing forecaster — over the first 3 gameweeks. Points-error figures are scored on the 414 likely-starter rows both forecasters cover (83 model rows the baseline cannot reach are excluded from both sides). Rank correlation and the haul AUCs are averaged across gameweeks, and the AUCs only cover weeks whose run kept haul probabilities.
- Spearman rank correlation (higher is better)
- 0.3100.188
- Haul AUC, 6+ points (higher is better)the model put a player who went on to haul above one who didn't 69% of the time
- 0.6910.611
- Haul AUC, 10+ points (higher is better)the model put a player who went on to haul above one who didn't 70% of the time
- 0.7010.627
- MAE (lower is better)
- 2.4572.549
- RMSE (lower is better)
- 3.1963.283
Sample population: reference arm's predicted E[minutes] > 60, imposed on every arm (n=414 of 1202 player-fixtures); players with a prior-season appearance only
The model separates most clearly on ordering players and finding the tail — i.e. identifying the sub-population of ‘haulers’. It has also had the lower points error in every settled week that has a comparator, though by a small margin each time. 3 gameweeks are enough to start to form a picture, but not enough to conclude from; these figures will move, and the weekly pages are where you can watch them move — including GW2, where the baseline beat us on a measure.
One payment. Pro through 2026-27.
71p a gameweek on the pass. Same Pro limits either way — six-gameweek planner, chips, drafts, rivals. Monthly cancels whenever you like; the pass has nothing to cancel. 14-day refund on both.
Upcoming gameweek
Max xP is the strongest fifteen the model can name for the coming gameweek, three per club as always. The Differential XV chases points the crowd hasn't spotted yet, and Elite Ownership shows what the top 10k are actually fielding, captaincy included.
A forward projection, spending nothing and owing nothing: the strongest fifteen by expected points, three per club, with no £100.0m budget to respect. Not the same object as the judged legal squad, which was bought inside the budget before a past deadline and has since been marked on what actually happened — that one is on each settled week’s own page.
A weekly toolkit — with the workings shown.
Sideline Quant turns match odds, player shares, and minutes risk into expected points for every player, every gameweek — refreshed for every deadline, with nothing behind the curtain.
- 01Read this week's board
Expected points, minutes risk, and haul chances for the gameweeks ahead.
Browse predictions → - 02See the real fixture map
Every match priced from odds-based expected goals — find the genuinely soft runs.
Browse fixtures → - 03Open any number
Every xP figure unpacks into four layers — odds, shares, minutes, sims.
How it works →
What makes the numbers different
- Minutes are a distribution — start, sub, cameo, unused — not expected-minutes ÷ 90.
- Fixture difficulty comes from de-margined bookmaker odds while the market is posted, then our own team ratings — both finer than FPL's 1–5 FDR.
- Every player carries haul probabilities from the same simulations, not just an average.
- Last season's run is published in full — every gameweek, transfer, and chip.
Built to out-plan your mini-league.
One good deadline doesn't win a league; thirty-eight might. The engine behind the board solves every one of them.
The optimiser runs a mixed-integer solve over the whole player pool and up to six gameweeks in one pass — transfers, captain, bench order, all four chips, and whether a −4 pays for itself. Every plan reports the probability it beats simply holding.
Open the optimiser →Import your mini-league and the same simulations play out your run-in: expected finish, chance of first, who is realistically catchable, and which differentials swing the most points against the managers above you.
Score your rivals →The board, the optimiser and the rivals view all read the same simulations, so the numbers never disagree with each other — and when they surprise you, you can open them and see why.
See how it fits together →The season the model never played
A replay of 2025-26 run after the fact: the model re-deciding every transfer and captain across a season already in the books. Different exhibit, different question — the live record above was predicted before each deadline and marked after it. A sceptic should weigh this one second, and we would agree.
+991 vs the best set-and-forget baseline, over 38 gameweeks with hits netted off.
Every backtested gameweek, transfer, and chip →- This model242763.9/GW
- Set-and-forget template143637.8/GW
- Set-and-forget by price108428.5/GW
Baselines are set-and-forget reference teams that never transfer — not a typical active manager.
Before the 2026-27 season we ran this model through 2025-26 gameweek by gameweek, letting it see only what was known before each deadline, and let it pick a team, make transfers and play chips. This is what it would have scored. It is a simulation, not a live record, and past performance does not guarantee future performance. The replay used last season's bonus-point rules; this season's changes are in the live model.
The whole season, free. Pay for the fixture-by-fixture detail.
Every player’s projected points to the end of the season are open to anyone. You pay when you want the model fixture by fixture — the transfer plan, chip timing, drafts and rivals scoring.
A wider board and more solves than browsing signed out.
- Every player, ranked by xP
- 2 gameweeks of predictions and planning
- Haul probabilities — P(≥6) and P(≥10) for the next gameweek
- Clean-sheet, goal and bonus chances on every player
- 10 solves a day · 1 synced draft
- Full performance archive
Try a month. Same Pro limits as the season pass.
- Every column: elite top-10k EO, price changes, floor and ceiling
- 6-gameweek prediction horizon
- 6-gameweek planner, chips and hits
- 4 synced drafts · 60 solves a day
- Rivals scoring and layer inspector
One payment, nothing to cancel. Saves 30% on monthly.
- Every column: elite top-10k EO, price changes, floor and ceiling
- 6-gameweek prediction horizon
- 6-gameweek planner, chips and hits
- 4 synced drafts · 60 solves a day
- Rivals scoring and layer inspector
What the model is reading into this season
I assumed the answer was 'as long as you can afford to wait'. It isn't, or at least not in ten seasons of data.
Single-week expected points for opening weekend, off the 10 August model run. Mbeumo tops the board, Haaland is fourth at 74% ownership, and two of the biggest names carry minutes doubt.
Every club’s first six fixtures, ranked by the model’s expected-goal difference rather than FPL’s 1–5 difficulty ratings.
Transfers, captain, and chips from one model — so you stop guessing.
Open any expected-points number and see how it was built.
Check the season track record — receipts, not vibes.