Football AI Coach

## Coach notes: 30-second clip

Headline: White controlled the ball for 11.2 s against 5.4 s for lime (2.1 times as long), but 48.2% of white's control was the goalkeeper holding the ball. No shots in the clip.

White (defends the left goal, attacks right)

• Shape: average width 48.5 m, depth 32.2 m, spread 16.3 m; the wider side, 11.6 m wider than lime.
• Position: 6.2% of tracked time in its own third, 76.8% in the middle, 17.0% in its attacking third.
• Passing: 4 pass candidates (3 backward, 1 forward, 0 sideways), 0 of them low confidence; counts are minimums.
• Goalkeeper: stood a median 18.7 m from his own goal line (range 17.0 to 23.0 m), at or beyond the penalty-area edge (16.5 m), and held the ball for 5.4 s.
• Pace: W05 covered 20.2 m of the team's 22.3 m of high-speed running (90.6%), top speed 26.6 km/h.

Lime (defends the right goal, attacks left)

• Shape: average width 36.9 m, depth 29.3 m, spread 13.8 m; the more compact side, 11.6 m narrower than white.
• Position: 29.3% of tracked time in its own third, 69.3% in the middle, 1.3% in its attacking third.
• Passing: 1 pass candidate (0 backward, 0 forward, 1 sideways), 1 of them low confidence; counts are minimums.
• Pace: L06 covered 17.4 m of the team's 30.0 m of high-speed running (58.0%), top speed 27.3 km/h.

Key events (high and medium confidence)

• 2.3 s: pass W01 to W02, 7.4 m (backward 6.9 m) [high]
• 5.0 s: pass W02 to GKW110, 20.7 m (backward 16.9 m) [high]
• 12.0 s: pass GKW110 to W05, 40.6 m (forward 40.6 m) [medium]
• 25.2 s: L04 knocked the ball on 2.9 m and regained it [medium]
• 26.9 s: contested ball, it passed from L03 to W04 [medium]
• 27.1 s: pass W04 to W08x, 11.6 m (backward 11.6 m) [medium]
• 28.8 s: contested ball, it passed from W08x to L05 [high]
• 5 further events are low confidence and are not listed.

Both teams

• Lanes (as seen in the video): near side 54.4% of white's and 57.9% of lime's tracked time; far side 10.7% and 4.9% (partly a camera effect).
• Shots and out-of-play: No shot, goal or out-of-play moment in this clip. The ball was never near a goal mouth (0 steps); it came within 1.5 m of a boundary line for 5 steps (0.6 s, still in play); and it was inside a penalty area for 12 steps (1.4 s).

Recommendations (hypotheses to test on video, not verdicts)

• Goalkeeper as the passing hub (white). Evidence: he held the ball for 5.4 s (48.2% of white's control) while standing a median 18.7 m from his own goal line. Suggestion: review that spell on video: was he waiting for an option or being pressed? If the aim is to play out from the back, work on a quicker release and defined passing options around him; if the long ball is the plan, make that choice earlier rather than after a long hold. Confidence: high for the facts, medium for any interpretation.
• Progression with the ball (white). Evidence: 3 of white's 4 pass candidates went backward, and white had 17.0% of its tracked time in its attacking third; it was 11.6 m wider than the other side. Suggestion: look for short forward options through the middle third, and check on video whether the width is being used to open lanes. Confidence: medium (very few passes).
• Getting out of its own third (lime). Evidence: 29.3% of tracked time in its own third and 1.3% in its attacking third, with 1 pass candidate against 2 carry events (counts are small). Suggestion: check how the team progresses after winning the ball and whether L06's runs get support. Confidence: medium (zones are solid, events are few).
• Support around the long ball (white). Evidence: the 40.6 m ball from GKW110 to W05 was followed by a change of possession (that loss is a low-confidence event). Suggestion: check on video who was near W05 and whether the contest was in the air; if long balls are part of the plan, rehearse the second-ball positions. Confidence: low.
• Reliance on one runner. Evidence: W05 produced 20.2 m of white's 22.3 m of high-speed running (90.6%); L06 produced 17.4 m of lime's 30.0 m of high-speed running (58.0%). Suggestion: over a full match, track whether pace depends on these players and what that means for their load and for how predictable the team is; a 30-second clip cannot show whether it is a pattern. Confidence: low (high-speed running is a lower bound).
• Far side of the pitch. Evidence: the far-side lane holds only 10.7% of white's and 4.9% of lime's tracked time. Suggestion: check on video whether the far side is genuinely unused or just hard to see from this camera; a second camera angle would settle it. Confidence: low.

What this cannot tell you

• Short clip (about 30 s): all numbers are illustrations of the pipeline, not match statistics.
• Distance and high-speed running are lower bounds: they count only stretches where a player was tracked.
• Player positions are good to roughly 1 m (calibration 0.1-0.8 m depending on the area, plus foot-point error); speeds are less affected than positions.
• Top speed is reported only for players with enough clean tracking; others are not reliable.
• Players are tracked as chains, which can be fragments of the same person; events are attributed to chains.
• Ball speed and position are rough while the ball is in the air; pass lengths between touches are better than speeds.
• Passes and duels are minimum counts; events resting on a one-step touch are low confidence.