The MrPredictions methodology page is the full breakdown of how our daily football picks are made. Every pick on the daily list runs through the same statistical model, the same data inputs, and the same confidence filter before publication. This page explains exactly how that process works, who runs it, and why our daily list stays short on purpose.
If you want a quick summary: we use Expected Goals modelling, compare our model probability to the bookmaker’s odds, and only publish picks where the gap is meaningful. The full version, including the four-pillar confidence framework and the team that runs the daily process, is below.
How Our Football Prediction Model Works
Every MrPredictions football pick starts with Expected Goals, or xG for short. xG is a statistical measure of the quality of every shot a team takes and concedes, weighted by shot location, angle, defensive pressure, and shot type. It tells you what a team should have scored or conceded based on the quality of chances created, rather than the lucky or unlucky final score.
A team that wins 1-0 from a deflected free kick may have actually conceded 2.3 xG in clear chances. The scoreline lies. The xG tells you what the game actually looked like. We feed the last 10 to 20 fixtures of xG data for both teams into the model, then layer in form trends, head-to-head history, fixture context (rotation, travel, midweek European football), and team news (injuries, suspensions, manager changes).
The model produces a probability for each available bet type on the fixture: 1X2 match-winner, BTTS Yes and No, Over and Under 2.5 goals, correct score clusters, HT/FT outcomes, Asian Handicap lines, corner totals, and yellow card counts. We then compare each probability to the bookmaker’s implied probability (calculated from the published odds). If the gap between our number and the bookmaker’s number is at least 5 percentage points, the pick goes on the daily list. Below that threshold, the match gets skipped.
The Four Confidence Pillars
Every pick on MrPredictions earns its place by clearing four structural checks. These pillars apply differently to each bet type (BTTS picks lean on clean-sheet rates, corner picks lean on crossing rate and referee tendencies, card picks lean on team discipline averages) but the underlying logic is the same:
- Statistical Edge. Our model probability must exceed the bookmaker’s implied probability by at least 5 percentage points. No edge, no pick.
- Data Quality. The teams in the fixture must have enough recent fixtures (at least 10) for the model to produce a reliable xG baseline. Newly-promoted clubs in their first month, or clubs returning from a long international break, get treated with lower confidence.
- Fixture Context. Rotation patterns, midweek European football, travel fatigue, weather forecasts, and team news must all be checked before the pick locks. A model number that ignores a key injury is worthless.
- Bet Liquidity. The bookmaker line must be liquid enough to actually place a meaningful bet on. Thin betting lines on obscure fixtures get filtered out because the line is unreliable and the spread eats the edge.
A pick that clears all four pillars goes on the daily list. A pick that fails any one of them gets skipped. The discipline is the methodology.
Meet the MrPredictions Team
The MrPredictions daily picks are produced by a small editorial and analytical team based across the UK, Nigeria, and continental Europe. Every team member has direct experience in either professional football data analysis, sports journalism, or quantitative betting modelling.
Olumide Adekunle, Editor-in-Chief
Olumide leads MrPredictions editorial direction from Lagos. He spent 12 years covering African and European football as a sports journalist before moving into football prediction publishing in 2019. He sets the daily editorial priorities, oversees the picks publishing schedule, and reviews every methodology change before it goes live. Olumide also reviews every utility page on the site, including this methodology page, to keep the writing honest and the reader-facing claims accurate.
Sarah Chen, Lead Data Scientist
Sarah built the MrPredictions xG model and runs all statistical modelling for the site from London. She spent six years in quantitative finance before moving to football data in 2020, drawn by the analytical similarities between match-result probability modelling and the financial market modelling she trained on. Sarah maintains the xG database, runs the back-testing pipeline that validates every methodology change, and reviews the daily model output before publication.
Marco Russo, Senior Analyst, European Football
Marco covers الدوري الإيطالي الدرجة الأولى, الدوري الإسباني, دوري أبطال أوروبا, and الدوري الأوروبي for MrPredictions from his base in Milan. A former football scout with experience in both Italian and Spanish football, Marco translates tactical patterns into picks. His specialism is the squad-rotation effect, how clubs juggle midweek European fixtures and weekend domestic football.
Jamie Whitfield, Senior Analyst, English Football
Jamie covers the الدوري الإنجليزي الممتاز and English football pyramid for MrPredictions from Manchester. He spent eight years inside the UK betting industry before moving to MrPredictions in 2022, with deep experience in how bookmaker pricing works and where the gaps between bet type lines and statistical fair value tend to emerge. Jamie’s daily focus is the Premier League picks and the cross-league fatigue patterns that emerge from midweek European competition.
Tunde Bello, Senior Analyst, African Football
Tunde leads MrPredictions coverage of African football, including the NPFL, the CAF Champions League, the CAF Confederation Cup, and the African Cup of Nations. Based in Abuja, Tunde brings 10 years of African football scouting experience and direct connections across the continent’s domestic leagues. His coverage often spots value the European-focused prediction sites miss entirely.
Diego Alvarez, Pick Specialist, BTTS and Goal Bets
Diego owns the توقعات تسجيل كلا الفريقين و توقعات أعلى/أقل for MrPredictions. From Madrid, he runs the daily goal-based analysis and maintains the team-level BTTS rate database that drives our daily Yes and No picks. Diego’s specialism is mid-table BTTS Yes spots and Serie A Under 2.5 picks, where the bookmaker line consistently misprices the structural defensive culture.
Hans Müller, Pick Specialist, Asian Handicap and Cards
Hans covers Asian Handicap, corner totals, and yellow card bet types for MrPredictions from Hamburg. With a background in German sports data journalism, Hans built the referee-tendency database that powers our yellow card predictions و توقعات الركنيات. His daily focus is finding fixtures where the referee assignment changes the fair line beyond what the bookmaker has priced.
Why We Skip Matches
The single biggest difference between MrPredictions and most other free football prediction sites is what we do not publish. Most prediction sites publish a pick on every fixture on the daily calendar, sometimes hundreds of picks per day across hundreds of leagues. We do the opposite.
If the data does not show a clear edge over the bookmaker’s odds, the match gets skipped. About one day in five, no match on the entire daily card clears our مصرفي موثوق confidence threshold, and we publish no banker that day. That is not a quiet day; it is the system working. Forcing picks to fill space is how most prediction sites bleed their long-term hit rate. Discipline is the only sustainable edge.
How We Update the Model
The MrPredictions methodology is not static. Every month, Sarah Chen runs the model output against the actual results from the previous 30 days and tracks where the model under-performed against expectations. If a league changes tactical character (the Bundesliga shifts from gegenpressing to possession football, for example), the model has to adapt or its picks drift out of calibration.
Quarterly, the full editorial team reviews the model assumptions against the previous quarter’s actual hit rate per bet type. If a specific bet type is drifting (BTTS picks under-hitting expectations, for example), Diego or Hans investigates whether the structural inputs need reweighting. The model has been recalibrated seven times since launch, with the most recent recalibration in February 2026.
Every methodology change is logged with a timestamp and a reason. Olumide reviews every change before it goes live. We do not chase fashion; we update only when the data shows the current methodology is missing something real.
Honest Limits of Our Methodology
No statistical model can predict football with certainty. Red cards happen. Penalties get awarded for soft fouls. Goalkeepers have career nights. Stars get injured in warm-up. Our headline accuracy on heavy-favourite picks sits around 60 to 65% across a full season; on closer matches with tighter xG profiles, accuracy is lower because the variance is genuinely higher. We publish the model confidence on every pick so you can size your stake accordingly.
The methodology is the best framework we have for finding the fixtures where the statistical edge is real. It is not magic. We do not promise 100% sure wins, we do not sell fixed matches, and we do not have insider information on referee decisions or player injuries. Anything that sounds too good to be true is. We bet smart over the long run rather than chase miracles on a single weekend.
Questions About Our Methodology
What is Expected Goals (xG)?
Expected Goals is a statistical measure of the quality of a shot, weighted by location, angle, defensive pressure, and shot type. A team that creates many high-quality chances will outperform a team that creates few low-quality chances over a long run, even if their immediate scoreline does not reflect it. xG is the foundation of every MrPredictions pick.
How accurate are MrPredictions picks?
Headline accuracy on heavy-favourite picks sits around 60 to 65% across a full season. Tighter matches with closer xG profiles have lower accuracy because the variance is genuinely higher. We publish the model confidence on every pick so you can size your stake accordingly. We do not promise 100% accuracy and we do not sell fixed matches.
Is the model open source?
The model itself is proprietary, but the methodology behind it (Expected Goals, the four confidence pillars, the league-by-league rate baselines) is fully disclosed on this page and across the prediction pages. People who want to dig deeper into how a specific pick was made can read the dedicated prediction pages for كلا الفريقين يسجلان, Over/Under, corners, or yellow cards.
How often is the model updated?
The xG database refreshes daily with the previous day’s results. Model calibration is reviewed monthly by Sarah Chen against actual hit-rate output. Methodology changes are reviewed quarterly by the full editorial team. The model has been formally recalibrated seven times since launch, with the most recent recalibration in February 2026.
Can I contact the team about a specific pick?
Yes. Reach the team through the contact page with your specific question. Methodology queries go to Sarah Chen. Pick-specific questions on a particular league or bet type go to the relevant senior analyst. We respond within 24 to 48 hours during the active football season.