Today's Bankers
Predict IQ ranks every market on every fixture by its own confidence (the blended model probability), then builds accumulators that grow from the same core: the 3-fold is the 2-fold plus the next-best selection, and so on. Use the league chips above to filter competitions.
βοΈ Banker settings
One selection per fixture is always enforced.
Ask Edge
Type what kind of bet you want and Predict IQ builds it from today's fixtures using its own probabilities and Sky Bet's prices.
What can I ask? β tap for examples
| You want⦠| Type something like⦠|
|---|---|
| An acca of a set size | "treble" Β· "5 fold acca" Β· "safe treble" Β· "long shot acca" |
| An acca of one market | "btts acca" Β· "over 2.5 acca" Β· "win only acca" Β· "win or draw acca" |
| Full-time results only | "result treble" Β· "result 4 fold" (legs are home win / draw / away win) |
| A target payout | "Β£20 to win Β£200" Β· "Β£10 to win Β£300 results acca" Β· "treble at 5/1" |
| Correct scores | "correct scores" Β· "correct scores for Newcastle v Liverpool" |
| Best value singles | "value singles" Β· "under 2.5 picks" |
| Your own selections | "Liverpool to win, Southampton to win, Everton v Palace 1-1" β add "@ 6/4" after a leg to use the price you've been offered, and "Β£10 to win Β£100" at the end to have it topped up to a target |
What you get back: the legs with Sky Bet's price, the model's chance for each, the edge vs the bookmaker, combined odds, the chance of the whole bet landing, returns on your stake and whether it's positive value β plus a second "scores on the same matches" bet, an "Open on Sky Bet" link per leg, and buttons to copy the slip or record it as placed. Only fixtures in the selected leagues (chips above) are used.
Favourite bets
Save the kinds of bet you like. Each has a target β the odds you want, or a stake and the amount you want to win β and Edge finds the best way to hit it from the current fixtures (respecting the league filter).
Detailed analysis
Professional-analyst view of one match: every independent signal we can measure, a confidence scorecard, and the question that matters β where has the bookmaker got this wrong? Premier League & Championship (where FotMob data is richest).
Performance stats
What would have happened if you'd put Β£ on every bet Predict IQ suggested, at Sky Bet's price. Predictions are logged automatically when you open the app and settled automatically once results are in.
How Predict IQ works
Every morning Predict IQ gathers the numbers, rates every team, turns those ratings into match probabilities, and compares them with the bookmakers' prices. You see the result as predictions, value bets and bankers β with the reasoning shown, not hidden.
1 Β· Gather the data
Results, goals and shots (football-data.co.uk) Β· expected goals (xG), form, league tables, injuries, predicted line-ups, head-to-head, referee and weather (FotMob) Β· live prices from six bookmakers including Sky Bet and Bet365 Β· odds movement tracked day by day.
2 Β· Rate every team
Each team gets an attack and defence rating for home and away, built from real xG and results. Promoted and relegated sides are converted to their new division. Ratings update as the season goes on, weighting recent games more heavily.
3 Β· Turn ratings into a match prediction
Attack vs defence, home advantage, squad value, recent form, injuries (weighted by the market value of who's missing) and fatigue produce an expected-goals figure for each side. A Poisson / Dixon-Coles model then works out the chance of every scoreline β and from that, win/draw/win, over/under, BTTS, correct scores and more. Early in the season the model deliberately widens its uncertainty.
4 Β· Blend in the market
Bookmakers are very good at pricing football. Predict IQ blends its own view with the six-bookmaker consensus (50/50 by default) β so you get an independent opinion, anchored to reality.
5 Β· Find the edge
Edge = Predict IQ's probability β the probability implied by Sky Bet's price, in percentage points. +8 means the model thinks the outcome is 8 points more likely than the price suggests. Positive edge is value; negative edge means the price is short β avoid. Every card shows the model, the bookmaker and the gap side by side.
6 Β· Keep score
Every prediction is logged the moment you open the app and settled automatically against the final score. The Stats tab shows hit-rates, simulated profit at level stakes, and the categories the model is best and worst at. Tap "I've placed this bet" and it scores its picks against your real money too.
Practise first β use the app without risking a penny
The best way to learn whether Predict IQ (and you) can beat the bookmaker is to keep score for a few weeks before staking real money. Everything below is free and automatic.
- Make your own calls. Results tab β Logged predictions: type your score for each match (e.g. 2-1) and tap Lock before kick-off. Locked calls can't be changed, so it's an honest test.
- Compare after the final whistle. Results settle automatically the next morning. The You vs Predict IQ scoreboard shows how many results and exact scores each of you got right.
- Paper-trade the bets. On any banker or Ask Edge acca tap π Track only (paper) β it's logged and settled exactly like a real bet, with no money involved. Try different styles: safe doubles, value singles, 4-folds.
- Read the Stats tab weekly. It shows what Β£10 on every suggestion would have returned, split by bet type, league and confidence band, plus the best and worst categories. Give it 3β4 weeks (50+ settled bets) before drawing conclusions β short runs are mostly luck.
- Use the "Why?" panel. Before each match, read the reasoning and decide whether you agree. Over time you'll learn which signals you trust (injuries, H2H, market moves) and which you don't.
- Graduate slowly. Only when the paper record is positive, move to real stakes of 1β2% of a fixed bankroll, and record them with π· I've placed this bet so the app scores its picks against your actual money. If real results fall behind the model, the Stats tab will tell you β cut back.
A good weekly routine: Friday β open the app, load fixtures, lock your own calls, note the value matches. Saturday/Sunday β place (or paper-track) only bets with a real edge. Monday β check Results and Stats, and the AI analyst write-ups on anything surprising.
How it's designed to help you win
- It hunts for value, not winners. Profit in betting comes from being paid more than the true chance justifies β not from picking favourites. A 35% shot at 3/1 is a better bet than an 80% shot at 1/5.
- It tells you when NOT to bet. A grey badge or "Sky Bet agrees" means the price is fair β most matches are. Skipping those is where most of the saving is.
- It shows its working. The "Why?" panel lists every factor. If the reasoning doesn't convince you, don't bet.
- Bankers are graded by real chance of landing, not hype β a 6-fold at 8% is labelled very high risk for a reason.
- It measures itself honestly. Wins and losses both count. If a category keeps losing, the Stats tab will say so β use it to cut what doesn't work.
- Stake discipline. Treat your stake as a percentage of a set bankroll (1β2% on a single, less on accas). The AI analyst suggests sizing; never chase losses.
FAQ
Does it guarantee winners?
No. Nothing can. It produces probabilities and value estimates from data. Good bets lose often; the aim is to be paid more than the risk over many bets.
Why does "Predicted score 1-1" appear when a team is favourite to win?
A win is dozens of scorelines added together; 1-1 is just the single most likely one. The card now shows the likeliest score within the predicted result, plus the draw risk.
What does "+8 point edge" mean?
The model rates the outcome 8 percentage points more likely than Sky Bet's price implies. Edge is only as good as the model β check the Stats tab to see if its edges are paying.
How fresh is the team news?
Data refreshes every morning (plus late morning at weekends). Line-ups are "predicted" until about an hour before kick-off β check the official team news before placing.
Why do promoted teams look weak?
Because last season's numbers came from a lower division. They're converted and blended with a typical promoted-side profile, and real results take over after ~10 games.
Can it place bets for me?
No β bookmakers don't allow it. Use "Open on Sky Bet" and "Copy slip" to place quickly, then "I've placed this bet" to track it.
Where is my data kept?
On your own device only (your browser's storage). Nothing you log is sent anywhere. Use Settings β Export to back it up.
Which leagues?
Premier League and Championship, with the richest data. Others can be added.
No guarantees β please bet responsibly
Predict IQ is an analysis tool, not financial advice and not a tipping service. Every figure is an estimate; losing runs are normal even when the model is right about the odds. Only bet money you can afford to lose, set a budget and stick to it, never chase losses, and never borrow to bet. If betting stops being fun, stop.
18+ only. Help and support: BeGambleAware.org Β· GamCare β National Gambling Helpline 0808 8020 133 (free, 24/7) Β· GAMSTOP self-exclusion.
Results log
Tracks how the model actually performs. Log predictions snapshots the current call for every fixture (result, best-value market, predicted score). Track this bet on any Ask Edge / Favourites bet adds it with its stake. Settle marks everything won/lost from the final scores in the daily gather (or type scores in below).
Tracked bets
Logged predictions
Add your own call for each match (e.g. 2-1) and tap Lock before kick-off. After the final whistle the app scores you against Predict IQ.
Fixtures
Add the fixtures you want predictions for. League is taken from the home team.
Teams & Stats
Per-game expected goals for/against, split home & away. Att/Def adj let you factor in injuries, suspensions and form the stats don't know about yet (1.00 = no change; 0.90 = 10% weaker; 1.10 = 10% stronger) β add a note so the "Why" summary can quote it. Sources: FootyStats (xG tables / CSV), plus the multi-source web gather below.
Gather from the web
node scripts/gather.js locally) into data/ratings.js; this button applies them. It blends: ClubElo team strength ratings, football-data.co.uk results (goals, shots on target β xG proxy), and FootyStats xG if you add an API key in Settings. Each team shows which sources fed it.Or import a CSV
SkyBet Odds
Enter SkyBet's prices (fractional like 11/10 or decimal like 2.1). Edge = model probability β bookmaker implied probability, in percentage points. Leave blank to use the model's fair odds.
Model settings
AI analyst
The analyst runs through a Netlify function that holds your Anthropic key server-side (set ANTHROPIC_API_KEY in Netlify β Site configuration β Environment variables). If you also set ANALYST_PASS there, enter the same passphrase here so only you can use it.
Other keys
Data
Predict IQ trial build Β· Model: Poisson / Dixon-Coles on xG Β· Stats: FootyStats, ClubElo, football-data.co.uk Β· Prices: SkyBet Β· 18+ BeGambleAware
