Is xG accurate for football predictions?

Over a single match xG is noisy, but across rolling 10-match samples it predicts future goals far better than league position or recent form. Pro syndicates and bookmakers both build their models around it.

Where can I find free xG data?

Understat (top 6 leagues), FBref (Opta xG across most European leagues), and the EuroPicks fixture pages — every match preview ships with rolling xG and xGA for both sides.

What is a good xG per match?

Elite attacking teams average 1.8–2.4 xG per match in domestic play. Mid-table sides cluster around 1.1–1.4. Anything under 0.9 across a 10-match window is relegation-level output.

Does xG include penalties?

Most public xG values include penalties at ~0.76 each, which can distort small samples. Look for 'npxG' (non-penalty xG) when judging open-play quality.

How does EuroPicks use xG?

Our model blends rolling xG and xGA, momentum-weighted form curves, and closing-line value. Picks publish only when the modelled edge exceeds the bookmaker margin by at least 4%.

xG Football Predictions: How Expected Goals Find Betting Value in 2026

# xG Football Predictions: How Expected Goals Find Betting Value Expected Goals — xG — is the single most important number in modern football analytics. It tells you how many goals a team *should* have scored from the chances they created, rather than how many actually went in. For bettors, that gap between expected and actual is where value lives. This guide breaks down what xG is, the math behind it, how to use it for football predictions, and exactly how EuroPicks blends xG with form curves and market odds to surface profitable picks. ## What is xG (Expected Goals)? xG assigns every shot a probability between 0 and 1 of becoming a goal, based on dozens of variables: shot location, body part, assist type, defensive pressure, goalkeeper position, game state, and more. Sum every shot in a match and you get the team's xG for that game. - A tap-in from 4 yards out: ~0.85 xG - A header from a corner: ~0.10 xG - A 25-yard speculative effort: ~0.03 xG A team that creates 2.4 xG but scores 0 has been unlucky. A team that scores 3 from 0.8 xG has been clinical — and is almost certainly about to regress. ## Why xG Beats Raw Form Football tables lie in the short term. A team on a 5-match winning streak might be riding hot finishing and shaky opposition keepers. Their underlying xG can tell you whether the run is sustainable or about to collapse. Three things xG catches that form tables miss: 1. **Finishing variance** — over-performance vs xG regresses to the mean inside 8–12 matches. 2. **Defensive luck** — conceding few goals from high xGA (Expected Goals Against) means a keeper is bailing the defence out. 3. **Chance quality** — a team taking 20 long-range shots per game is not creating, even if the volume looks impressive. ## The Math: How an xG Model Is Built Modern xG models are gradient-boosted decision trees trained on hundreds of thousands of shots. The model learns the probability a shot becomes a goal given features such as: - Distance to goal and angle - Body part (foot, head, other) - Type of pass leading to the shot (through-ball, cross, set-piece) - Pattern of play (open play, counter, set-piece, penalty) - Defenders between shot and goal - Goalkeeper position A simple logistic regression on shot distance alone already explains ~60% of the variance. Production models from StatsBomb, Opta, and Understat push that past 80%. ## Using xG for Match Predictions Three practical workflows for bettors: ### 1. Spot regression candidates Pull each team's last 10 matches. Compare goals scored to xG and goals conceded to xGA. Teams with the biggest negative gaps are over-priced by the market and are the strongest "back" candidates over the next 4–6 matches. ### 2. Over/Under markets Sum each team's average xG and xGA over the last 10 matches at home/away splits. The combined number is a fair estimate for total goals. If the model says 2.9 and the line is 2.5 Over at 1.95, that's value. ### 3. BTTS (Both Teams to Score) Multiply each team's per-90 xG by the opponent's xGA. If both teams project above 1.1 expected goals, BTTS Yes is mathematically favoured. ## How EuroPicks Uses xG Our pick engine blends three layers: - **Rolling xG and xGA** from the last 10 matches, weighted by recency. - **Form curves** — momentum-adjusted ratings that decay older results faster than a flat average. - **Closing line value (CLV)** — comparing our fair odds to the market's closing price to flag where books are slow to react. Picks only publish when the model edge exceeds the bookmaker margin by at least 4%. That's why our [track record](/track-record) shows positive ROI even in losing months — the math doesn't care about narrative. ## Common Mistakes Bettors Make with xG - **Using a single match.** xG is noisy over 90 minutes. Use rolling 8–12 match windows. - **Ignoring match state.** A team trailing 2-0 will rack up xG with low-percentage shots. Filter or weight by game state. - **Treating penalties as normal shots.** Penalties are ~0.76 xG every time and distort small samples. Strip them when judging open-play quality. - **Forgetting opposition strength.** 1.5 xG vs Manchester City means more than 1.5 xG vs a bottom-half side. ## xG vs Other Football Metrics - **PPDA (Passes per Defensive Action)** — measures pressing intensity. Complements xG; doesn't replace it. - **Expected Threat (xT)** — values every on-ball action by how much it improves scoring probability. More granular than xG; harder to source. - **Field Tilt** — share of attacking-third passes. Useful style indicator, not a value signal on its own. ## Free Tools to Get Started - **Understat** — free xG data for top 6 leagues. - **FBref** — full Opta xG for most European leagues. - **Our [Smart Acca Builder](/accumulators)** — pre-filters legs by combined xG edge. ## FAQ See common questions about xG predictions below, or jump straight to today's [data-backed football tips](/picks).

FAQ

Explore more on EuroPicks