The centre-back receives: which way is he facing?

The ball comes across the back line from the other centre-back or from a full-back. The receiving centre-back is in his own half with the whole opposition in front of him. The claim from the touchline is that if he is already open he carries or plays forward, and if he is closed the ball goes back where it came from. Body pose lets us check it.

Brisbane Roar v Perth Glory, 2024-12-21 · Sydney FC v Adelaide United, 2025-02-01. Complete open-play passes from a centre-back or full-back to a centre-back, received in his own half, with at least 9 of the ten outfield opponents in front of him. 147 of 151 had all ten.

ONE ANGLE, MEASURED BEFORE ANYTHING HAPPENS

What "open" and "closed" actually mean

"Before any of the numbers: what is the angle measuring?"

Everything below rests on one measurement: the angle between a centre-back's shoulders and the direction of the goal he is attacking, at the instant the ball reaches him. Zero means he is square to the pitch and can see it all. Ninety means he is side-on. Past ninety his shoulders point back towards his own goal, and the pitch is behind him.

Four real receptions, one per band. Every dot is a measured joint from the pose file at the frame the ball reached him, drawn relative to his own hips. The thick green line is his shoulders, the dashed line is the direction of the goal he is attacking, and the angle between them on the ground is the only number this study uses. Drag the left panel to turn it.

SAME PLAYER, 46 MINUTES APART, SAME DISTANCE TO THE NEAREST OPPONENT.

The square and closed examples are both R. Grant, in the same match, with an opponent essentially the same distance away (6.7 m and 7.6 m), so neither can be explained away as the one who was being pressed. At 28° the pitch is in front of him and the ball finishes +14.7 m further up. At 106° his shoulders point back towards his own goal and it finishes -10.5 m, behind where he got it. The other two tabs fill in the middle of the range. That is the whole study in four frames; the rest of the page is whether it holds over 151 of them.

How square is he when the ball arrives?

"Is he already open, or does he have to turn?"

Openness is the angle between his shoulders and straight up the pitch at the moment the ball reaches him. Zero is square to the opponent's goal, 180° is his back to it. This is the shoulder line from the pose skeleton, not his direction of travel.

SHOULDERS OFF STRAIGHT-UP-THE-PITCH, AT THE MOMENT OF THE TOUCH

One column per 15° band. 0° is square to the opponent's goal, 90° is side-on, beyond 90° his shoulders point back towards his own goal.

THE SITUATION HE IS IN

AND WHERE THE BALL CAME FROM

Measured from the receiver: a feed at 90° is the square ball across the back line, a feed near 0° came from a team-mate further up the pitch playing it back. The condition that the whole opposition is in front of him is not a filter we had to force: in deep build-up it is simply true.

THEY MOSTLY DO ARRIVE OPEN.

The median reception is at 35°, so the typical centre-back in this situation is already fairly square. But the spread is wide: 69 receptions inside 30°, 43 between 30 and 60, 26 side-on between 60 and 90, and 13 beyond 90°, which is shoulders pointing back towards his own goal. That spread is what makes the question answerable. Where the ball comes from is part of it. 102 of these are the square ball across the back line, and those arrive at a median 26°. The other 49 come from a team-mate further up the pitch playing it back, and those arrive at 45° and go forward 33% of the time against 61%. A ball played back into a centre-back tends to leave him facing the wrong way. It does not explain the effect though: the two are only weakly related, and the shoulder angle keeps its size in a model that has the feed direction in it (section 04).

And does it decide what he does?

"Open means he carries or plays forward. Closed means it goes backwards."

His release is the carry and then the pass he plays. Forward means the ball finished more than 5 m further up the pitch than where he received it.

FORWARD AGAINST BACKWARDS, BY HOW HE ARRIVED

THE HYPOTHESIS IS RIGHT, AND IT TURNS ON ABOUT 60 DEGREES.

Inside 60° he plays forward 61% of the time and the ball finishes a median +7.35 m up the pitch. Beyond 60° that falls to 26% and -4.2 m, i.e. the ball ends up further back than it started. Backward releases go from 27% to 62%. The gradient is there across all four bands, but it is not smooth: the first two bands behave alike and so do the last two. Being turned 25° costs nothing. Being turned past square costs the attack.

Or is he just being pressed?

The obvious objection: he is closed because someone is on top of him, and he plays backwards because of the pressure, not because of his shoulders. So split by how close the nearest opponent is, and then put everything in one model.

PLAYS FORWARD, SPLIT BY HOW CLOSE THE NEAREST OPPONENT IS

If pressure were doing the work, the gap would shrink in the pressed row. It widens.

EVERYTHING IN ONE MODEL

AND ON THE CONTINUOUS OUTCOME

Least squares. The shoulder angle and where the ball came from both separate from zero; pressure, depth and the length of the feed do not.

NO. UNDER PRESSURE IT MATTERS MORE, NOT LESS.

Split at the median distance to the nearest opponent (10.9 m) and the effect survives both halves. Among the pressed receptions, arriving inside 60° gives 71% forward against 20%; with more space it is 54% against 44%. The gap is wider when he is pressed, which is the opposite of the objection: when there is time you can get away with being turned the wrong way, and when there is not, the shape you arrived in decides it. Two things survive a model that holds everything else constant, and they are the two the page has been about. The shoulders: each degree costs 0.1321 m of progress (p = 0.0003), so 60° of extra turn is worth about 7.9 m. And where the ball came from: the squarer the feed, the more likely a forward release (p = 0.0017 on that outcome, though only 0.0984 on the metres). Distance to the nearest opponent, how deep he receives and the length of the feed are all indistinguishable from zero. So the honest version of the coaching point is two sentences, not one: a ball played back into a centre-back starts him facing the wrong way, and from wherever he starts, the angle he has his shoulders at when it arrives is what decides where it goes next.

Where the touchline version is wrong

Two parts of the touchline version do not hold up, and one question about scanning has a blunter answer than expected.

BACKWARDS, BUT NOT BACK TO THE PASSER

DOES CHECKING BEFOREHAND PUT HIM IN A BETTER SHAPE?

A look away from the ball means his head more than 60° off it for at least 0.2 s, in the three seconds before it reaches him.

NOT BACK TO THE PASSER, AND NOT SLOWER.

He does not give it back to the man who fed him. That rate is flat across every band: 13%, 14%, 15% and 8%. What changes is backwards in general, not backwards to him. He finds a different escape, usually the goalkeeper or the other side. He is not slower on the ball either. A closed centre-back releases in a median 1.19 s against 1.54 s when square. He is not labouring over the turn, he is getting rid of it. The cost is direction, not time. And they barely check. With the ball travelling 15 to 30 m across to him, his head stays on it 72% of the last three seconds, and 90 of 114 receptions contain no look away from the ball at all. Whatever puts these players in an open shape, it is not a shoulder check in the final seconds. It is where they were standing before the pass was played.

The receptions, frame by frame

Every reception, replayed. The green arrow is his shoulder line, the dashed line is straight up the pitch and the arc between them is the number. Red is the opposition, all of it in front of him.

Player by player

Centre-backs with at least six receptions, ordered by how square they arrive. Two matches, so this shows the measure separates players, not that any of them is good or bad.

How this was built

The data. SkillCorner published body pose for two matches as a testing release: Brisbane Roar v Perth Glory (2024-12-21) and Sydney FC v Adelaide United (2025-02-01). Twenty-nine joints per player at 25 fps. We joined it to our own event data for the same matches, which is where the passes come from. Two matches is the whole public sample, so treat the size of every effect here as indicative even where the direction is clear.

The situation. A complete open-play pass from a centre-back or a full-back to a centre-back, received in the receiving team's own half, with at least 9 of the ten outfield opponents further up the pitch than him. That last condition barely bites: 147 of 151 receptions had all ten, which is what makes this a build-up problem rather than a transition.

The moment the ball arrives. The event clock is whole seconds, which is useless here, so the reception is taken from the feeding pass's own sub-second end timestamp and then refined inside ±0.6 s to the frame where the 25 fps ball is closest to him. Median ball-to-player distance at that frame is 0.67 m.

Openness is the angle between his shoulder line and straight up the pitch, from the pose skeleton: the vector perpendicular to the left-to-right shoulder vector. Not his direction of travel, which for a receiving centre-back is often nothing at all. The convention is validated the same way as in the earlier reports: over every player-frame above 4 m/s the shoulders sit within 90° of the direction of travel about 98% of the time, and beyond 135° about 1%, which is what rules out a flipped sign. Coverage is not an issue in this situation, because the broadcast camera is on the ball and the ball is arriving at him: the shoulder yaw is present on the reception frame for every one of the 151 receptions.

Checks are his head more than 60° off the bearing to the ball for at least 0.2 s. Here the ball is 15 to 30 m away while it travels, so that bearing is well defined and the on-the-ball / off-the-ball split is a horizontal question. Head elevation was computed too and is reported for completeness, but it separates nothing in this situation (-27.4° with his head on the ball, -27.2° with it turned away), which is exactly what you would expect: an incoming ball 20 m away and a team-mate 30 m away are both roughly at eye level, so the difference between them is left and right, not up and down. Elevation only earns its keep when the ball is at a player's feet.

What we cannot claim. This is an association, not an experiment. A centre-back may arrive open because the situation was already easy, and the same judgement that puts him in an open shape may be what finds the forward pass. The model rules out the three confounders we can measure (pressure, depth, length of the feed) and the effect is unchanged, but it cannot rule out the player himself. What it does establish is that the shoulder angle at the moment of reception is strongly and consistently attached to whether the ball goes forward, and that it is measurable before the outcome happens, which is the part a coach can use.