SHOT ATLAS
PREMIER LEAGUE 2025 / 26

FOR FOOTBALL COACHES · SHOOTING ANALYSIS

Shot location & scoring probability

Select a pitch zone to compare where shots are more likely to become goals.

20 Teams / 380 Matches2025.08.15 — 2026.05.24

Feedback on football analysis →

ShotsOwn goals excluded
GoalsCurrent filters
Observed conversionGoals divided by shots
Players with shots537 players with appearances

GOAL SITUATIONS · MATCH HALVES

First half vs second half · goals in motion

Compare shot patterns by half, then select a goal to explore its pre-shot situation in motion.

Uses the team, player and body-part filters above. Open play, set pieces and penalties are kept separate.

Goals by pre-shot situation

Compare every category for the selected team, player, body part, situation and goal period. The selected pre-shot category is highlighted; it does not hide the other bars. Shares mean a category's share of goals, not its scoring probability.

First half (estimated)Second half (estimated)Unclassified half
Detailed counts for the full match

Counts follow the team, player, body part and situation above. This table compares both halves, including unclassified goals, regardless of the replay period or selected category.

Pre-shot categoryFirst-half goalsSecond-half goalsUnclassified goalsAll goalsShotsObserved conversion

Categories use only the recorded last action. Ball recovery does not prove a tackle or an opponent error. A clearance label does not identify a clearance error. Earlier actions in the attack are unavailable.

Compare the two halves

Both maps follow the situation and pre-shot category above. Their colors use all matching shots, including misses.

Estimated halves: the source has no period field. Some stoppage-time shots cannot be separated from early second-half shots. Unclassified records are excluded from both maps.

How halves are estimated

Shots are ordered by source ID within each match. A single clock reversal from a minute between 45 and 59 to a lower minute in the same range is treated as the half-time restart. Shots before and after it are assigned to the first and second half. Without that signal, minutes 0–44 are assigned to the first half, minutes 60 onward to the second half, and minutes 45–59 remain unclassified. This is an inference, not official period data; unusually long first-half stoppage time or source-order errors may affect it. Headline totals and the full-match map retain all records.

Both maps use the same zones and color scale. Use the control directly above each map to show or hide its shot markers independently. With markers hidden, use the goal list or half buttons to choose a goal. Animation symbols remain visible.

White dashed guides follow repeated shot locations within the selected high-rate grouped zones. Sparse gaps and edges are trimmed separately for each team and half; a larger grouped zone does not automatically produce a larger outline.

First half (estimated)

Black dots mark goals. Hiding markers keeps the color map and totals unchanged.

High-probability shooting area

Second half (estimated)

Black dots mark goals. Hiding markers keeps the color map and totals unchanged.

High-probability shooting area

Color → scoring rate
Blue · 0%Teal · 5.6%Yellow · 22.2%Coral · 50%+

Colors blend continuously between these values. The scale expands low rates so small differences are easier to see. Dark gray-green = no data (or an unavailable model estimate). Diagonal lines = fewer than 30 shots. A yellow border marks the penalty area.

Selected goal · build-up to finish

Illustrative build-up · only shot origin is recorded

Attacking direction ↑ · green ring = recorded shot · amber = illustrative build-up · cyan = illustrative shot

Only the shot origin is measured. Build-up routes and actor positions are generic diagrams chosen from the recorded last-action label; their side, distance, timing and path are unknown. The shot points to the goal center for illustration.

Situation
Shot body part
Pre-shot category
Recorded last action
Assist credited by source
Recorded position
Match period

The source last-action label and credited assist are shown separately from our category. Unknown or ambiguous actions have no invented build-up. Coordinates use a normalized 105 × 68 m pitch.

  1. 01 · Illustrative build-up
  2. 02 · Shot location
  3. 03 · Recorded outcome

Half labels are estimates. Unclassified goals remain available under Full match or Unclassified half.

Full-match map and detailed analysis

Open-play shot map

Open play only. Corners, direct free kicks, other set pieces and penalties are excluded from this map, its totals and player rankings. See the separate set-piece map below.

Colors show observed conversion. Zones with fewer than 30 shots are hatched.

Goal × No goalUse arrow keys to select a zone
Color → scoring rate
Blue · 0%Teal · 5.6%Yellow · 22.2%Coral · 50%+

Colors blend continuously between these values. The scale expands low rates so small differences are easier to see. Dark gray-green = no data (or an unavailable model estimate). Diagonal lines = fewer than 30 shots. A yellow border marks the penalty area.

High-probability shooting area

Curved dashed white outlines show the shot-supported parts of grouped zones with at least 80% of the highest estimated scoring probability. The boundary follows areas with at least 3 filtered shots within 3 m, trimming sparse gaps and edges instead of enclosing the whole large rectangle. Supported connected areas remain joined; blue edges (displayed rate 3% or lower) recede by about 1 m. Isolated outlines remain 20% narrower. Counts and rates describe the original grouped zones, not the exact curved footprint. The same rule is recalculated for every team, player, period and situation.

The dashed white guide is recalculated for the selected team, player and body part. The yellow penalty area is a fixed pitch marking; the solid white border is your selected zone.

How the guide is chosen

For the white guide only, shots are grouped into 10.5 × 8.5 m rectangles. If none has at least 30 shots and a goal, we retry with 21 × 17 m rectangles. The fine colored map remains unchanged. Eligible rectangles need at least 30 shots; estimated probability = (goals + 1) / (shots + 2), using a uniform Beta(1,1) prior. We select rectangles with at least one goal and an estimate at least 80% of the highest eligible estimate, merge shared edges and round the outline corners by up to 3 m for display. Before rounding, a 0.525 × 0.425 m display grid retains only points with at least 3 filtered shots within 3 m. This removes unsupported gaps and edges, including within the larger fallback rectangles. Blue boundaries recede by about 1 m, and isolated grouped outlines start 20% narrower. The density rule is a display choice, not evidence that 3 shots establish a reliable local scoring probability. No outline is drawn if no supported area remains. Separated areas may both qualify. The outlines can cross colored map cells, and their rates describe the whole grouped rectangle. This estimate is separate from the location color model. The displayed Wilson 95% intervals describe observed conversion; they are not posterior intervals or adjusted for selecting the best areas. The grid, cutoff and grouping rules are display choices, not evidence of an optimal tactic. No outline is shown when support is insufficient. Estimates use the selected season data and do not account for defensive pressure, shot creation or match correlations.

Explore a shot

Hover over or select a marker, or choose a shot below.

Set-piece shot map

Observed conversion only · independent of the main map’s situation and color mode

Uses the team, player and body-part filters above. Locations show where the shot was taken, not where the corner or free-kick delivery started. Own goals are excluded.

Attacking direction → · ring = goal · cross = no goal. Select a zone on the map or from the list below.

High-probability shooting area

Curved dashed white outlines show the shot-supported parts of grouped zones with at least 80% of the highest estimated scoring probability. The boundary follows areas with at least 3 filtered shots within 3 m, trimming sparse gaps and edges instead of enclosing the whole large rectangle. Supported connected areas remain joined; blue edges (displayed rate 3% or lower) recede by about 1 m. Isolated outlines remain 20% narrower. Counts and rates describe the original grouped zones, not the exact curved footprint. The same rule is recalculated for every team, player, period and situation.

The dashed guide uses only the selected set-piece type, team, player and body part. It follows the same selection method as the main map.

Color → scoring rate
Blue · 0%Teal · 5.6%Yellow · 22.2%Coral · 50%+

Colors blend continuously between these values. The scale expands low rates so small differences are easier to see. Dark gray-green = no data (or an unavailable model estimate). Diagonal lines = fewer than 30 shots. A yellow border marks the penalty area.

Highest-converting zones

Zones with at least 30 shots under the current filters

ZoneShotsGoalsConversion

Select a zone to highlight it on the map. Small differences and rankings alone do not establish superiority.

Shots by player

Totals for the selected team, situation and body part

PlayerShotsGoalsConversion

Players with zero shots are included. Select a name to filter the map to that player.

Data coverage & methodology Sources & validation

Observed data and coverage

The player list comes from Understat’s 2025/26 appearance records. Registered players who did not appear are not included. Transferred players retain the same player ID; shots are attributed to their team at the time of the shot.

Observed conversion equals goals divided by shots. Adaptive boundaries use league-wide non-penalty shots and stay fixed across filters. Starting with 21 × 17 m tiles, each tile is split into four when its width and height exceed the target: 3.5 m for recorded conversion of at least 1%, and 21 m otherwise. Each child tile is evaluated again. The finest tiles are 2.625 × 2.125 m. Zero-goal and empty tiles are not further subdivided. A recorded rate of 0% does not establish zero scoring probability. Small samples may be subdivided but remain hatched. Rates used to choose boundaries are descriptive, not independent validation. Small tiles omit text to avoid overlap; select them to see counts and rates. Coordinates are normalized to a 105 × 68 m pitch, rather than each venue’s actual dimensions. Attacks point upward in the attacking-half view and rightward in the full-pitch view. The view does not change the filters or totals. A zone crossing halfway is counted over its full area; switch to the full pitch to see the whole zone.

Confidence intervals use the Wilson method and assume independent shots. They are not adjusted for correlations within players or matches. The 30-shot threshold is a display rule, not a guarantee of reliable precision.

A model based on location only

Logistic regression uses distance to the goal center and the angle between the two goalposts. Features are standardized, with L2 regularization (coefficient 1) applied to all coefficients except the intercept. Penalties and own goals are excluded from training.

After validation, the displayed model is refitted on the full season. Estimates refer to the center of each zone. Estimates are hidden for zones with fewer than 30 training shots and are not applied to penalties.

The model does not include defender or goalkeeper positions, preferred foot, headers, pass quality or player ability. This simple location model is separate from Understat’s xG. It does not imply that moving to a location causes a goal.