Ahead
Protect
Template captain
Lock in the obvious pick. Avoid needless variance.
02 / Model audit
Accuracy metrics, without the hand-waving.
Every 2025/26 player-gameweek with actual points is scored against the model prediction. Lower error is better; higher correlation means the model ranks players more correctly.
Standard ML metrics
29,722 player-gameweeks
Spearman included for ranking quality.
MAE
1.02
avg absolute miss
RMSE
1.96
penalizes big misses
Spearman
0.70
rank correlation
Pearson
0.57
linear correlation
R2
0.32
variance explained
Plain-English check
33% lower MAE
Compared with always guessing the same average score for every player. SmartPlay misses by 1.02 pts instead of 1.53.
Season-long player pool
73 of 100
Model top-100 by total xPts that also finished in the actual top 100 by points.
Calibration sanity check
4.03 -> 3.80
When the model makes higher-upside calls, actual average points move with the prediction instead of staying flat.
Use xPts as a ranking and risk signal, not a promise that one player will score exactly one number in one gameweek. The simple comparison above is just a sanity check; MAE, RMSE, Spearman, Pearson, and R2 are the audit metrics.
Read the methodology03 / Season recap · XI audit · 2025/26
Did the model spot the right players?
We added up SmartPlay's weekly xPts for every player across all 38 gameweeks, then picked the highest-scoring legal 3-4-3 XI from that list. Below, we compare it to the XI you'd build with perfect hindsight — the players who actually scored the most.
This shows whether the AI got the long-term player calls right. It isn't a recommendation to start the same 11 every week.
Headline result
95%
of the points a perfect-hindsight XI scored.
SmartPlay's XI banked 2,034 real FPL points across the season — 107 short of the best possible 3-4-3 (2,141 points).
The numbers
Gold ✓ on a jersey means the player is also in the perfect-hindsight XI.
SmartPlay XI scored
2,034
real points
Perfect XI scored
2,141
real points
Gap to perfect
107
points (5.0% short)
Players also in perfect XI
7 of 11
same names
04 / Mini-league lens
Made for your mini-league.
Most FPL tools give every manager the same advice. SmartPlay’s picks adapt to the specific rivals in your league.
Strategy
The mode you pick reshapes every call.
Each mode rewires the captain, transfer, and chip recommendations — without you touching the underlying model.
Close
Balanced
Best EV pick
Highest expected points without forcing risk.
Behind
Chase
Differential captain
Lower-owned pick for rank separation.
Mini-league toolkit
Track rivals. Catch up. Lock in the lead.
Rival captain coverage
See which players each rival is captaining this gameweek.
Differential edge
How much rank you stand to gain from your low-owned picks.
Template alignment
How much of your team matches the league consensus right now.
Live transfer activity
Watch every rival's transfers in and out as they happen.
Rival watchlist
Pin up to 3 rivals to follow their decisions closely.
Catchup analysis
Gameweeks left × points-per-GW you need to close the gap.
The mini-league lens is what SmartPlay does that nobody else does.
Open the mini-league lens05 / Founders
FPL Expertise combined with Artificial Intelligence
A top 0.4% all-time FPL manager and an AI/ML engineer trained at Oxford and Berkeley. Both use SmartPlay every gameweek.
Alish
Product & FPL strategy
- FPL manager since 2014 — twelve seasons playing.
- Finishes: 1× Top 10k, 2× Top 30k, 3× Top 50k, 4× Top 100k worldwide.
- Best season: rank 9,046 — top 0.1%, roughly 1 in 1,000.
- Season 2025/26 rank 49,078 — top 0.4%.
- All-time rank 38,405 — top 0.4%, roughly 1 in 250 managers who have ever played.
- ACCA-qualified — the same analytical instinct that audits a balance sheet picks apart an FPL squad.
- Shapes the FPL methodology and strategy behind every SmartPlay recommendation.
All-time rank 38,405 (top 0.4%)4× Top 100k finishes10 of 12 seasons in top 10%ACCA-qualified
Qazy
AI, Design & Engineering
- UC Berkeley School of Information — focus on Artificial Intelligence applications.
- University of Oxford AI Programme — top 5 in cohort with 89%, across deep learning, NLP, and ethical AI.
- 2nd place at the UC Berkeley MIMS Quantitative Finance hackathon — applied financial ML built head-to-head against the cohort.
- 5 years working in football, including professional football club FC Kairat.
- Builds production AI systems for a living — the same engineering rigor SmartPlay's prediction model is built on.
- Built the prediction engine, the solver, and the product behind SmartPlay.
AI/ML engineerUC Berkeley · MIMSOxford AI · Top 5 of cohortEx-FC Kairat
06 / Pricing
Two plans. Pick your season.
Pay once for season access. Secure checkout runs through Lemon Squeezy, and card details never touch SmartPlay.
Full Season
Most popular · Limited deal$80$49
Until the end of the 2026/27 season
- AI captain & transfer recommendations
- Protect, balanced, and chase strategy modes
- Mini-league lens — rival-aware picks
- Custom-budget squad optimizer
- Faceted search across every player
- Updated every 2 hours during the gameweek
- Full ML model methodology
Half Season
Limited deal$45$29
Half-season access
- AI captain & transfer recommendations
- Protect, balanced, and chase strategy modes
- Mini-league lens — rival-aware picks
- Custom-budget squad optimizer
- Faceted search across every player
- Updated every 2 hours during the gameweek
- Full ML model methodology
One-time payment only. Half Season customers can upgrade to Full Season for $29 from Account. No recurring payments or auto-renewal. Taxes may be calculated at checkout.
07 / FAQ
Common questions, honest answers.
How is SmartPlay different from other FPL tools?
Three things you won't find together anywhere else. First: the mini-league lens — recommendations change based on the specific managers you're competing with, not a generic top-1k consensus. Second: three one-click strategy modes (Protect, Balanced, Chase) that reshape captain, transfer, and chip decisions around your actual rank position. Third: a fully documented model — xPts, OFR, FRR, all explained at /under-the-hood. Most tools either pick for you in a black box or hand you a spreadsheet to interpret alone. SmartPlay does neither.
What does SmartPlay actually do for my mini-league?
When you link a mini-league, four things kick in. Captain calls adapt — safer when you're ahead, differential when you're chasing. Transfer plans factor in who your rivals already own, so a 'good buy' becomes a 'great buy' or a wasted move depending on your league. A live rival watchlist tracks each opponent's transfers, captains, and ownership. And a catchup view tells you how many points-per-GW you need to overtake the names in front of you. This is the wedge — everything else flows from it.
Will SmartPlay actually improve my rank?
Honest answer: it depends on whether you use it. SmartPlay gives you a model trained on 8 seasons of Premier League data, rival-aware recommendations, and a documented framework that the recent FPL champions all describe as part of their own process. What you still bring: discipline, patience, and the willingness to occasionally override the model. We don't promise top-10k — anyone promising that is lying. We promise tools that make better decisions easier.
Do I still play, or do I just follow the AI?
You still play. SmartPlay is a co-pilot, not autopilot. Every recommendation comes with the reasoning behind it — you can agree, override, or use the tool only on the decisions you actually want help with (captain only, transfers only, or all of it). The 2023/24 and 2024/25 FPL champions both describe analytics as one input alongside form, fixtures, news, and gut feel. SmartPlay is built to be the smartest input, not the only one.
Can I join mid-season?
Yes — and mid-season is often the smartest time. The model uses past-season data plus the current season's live data, so it's never starting from scratch. By the time you're halfway through, you know where you stand in your mini-league, your free hits are spent, and the run-in to GW38 is where most rank movement actually happens. The limited-deal price applies whether you join in GW1 or GW28.
How does payment work? Is it recurring?
Checkout runs through Lemon Squeezy. SmartPlay plans are one-time payments: Full Season is $49 until the end of the 2026/27 season; Half Season is $29 for the rest of the current season. Half Season customers can upgrade to Full Season later for $29. There are no recurring payments and no auto-renewal. If something goes wrong, email us from the footer link — we'll figure it out.
08 / Make your call
Managers who outsmart their league don’t guess.
Pay once for full access until the end of the 2026/27 season. No recurring payments.