Insights

Questions about how the game works, answered from the project's own event data: every pass, shot, tackle and touch of the Premier League seasons the project may read. Nothing here is a prediction.

Where is a cross worth playing from?

The half-space level with the box. A lofted cross from there is worth 0.045 goals; one from the deep wing, the second most common cross, 0.024, about half. Wide and deep is the poorest place to cross from among the places teams cross from often.

But the better places are harder to reach. The ball is on the deep wing about 106 times a match, in the half-space level with the box about 18 and level with the goal area about 6. Each zone's value says what a cross is worth once the ball is there, not that a team can choose to be there; the second chart starts from where the ball usually is.

middle third centre: too few lofted crosses to judge (0)too fewmiddle third half-space: too few lofted crosses to judge (60)too fewmiddle third half-space: too few lofted crosses to judge (60)too fewmiddle third wing: 0.017 goals a lofted cross (0.2 a match, 982 crosses); the ball is here 143 times a match0.0170.2 a matchmiddle third wing: 0.017 goals a lofted cross (0.2 a match, 982 crosses); the ball is here 143 times a match0.0170.2 a matchdeep centre: too few lofted crosses to judge (9)too fewdeep half-space: 0.032 goals a lofted cross (0.8 a match, 3,600 crosses); the ball is here 55 times a match0.0320.8 a matchdeep half-space: 0.032 goals a lofted cross (0.8 a match, 3,600 crosses); the ball is here 55 times a match0.0320.8 a matchdeep wing: 0.024 goals a lofted cross (7.4 a match, 33,690 crosses); the ball is here 106 times a match0.0247.4 a matchdeep wing: 0.024 goals a lofted cross (7.4 a match, 33,690 crosses); the ball is here 106 times a match0.0247.4 a matchbox depth centre: too few lofted crosses to judge (3)too fewbox depth half-space: 0.044 goals a lofted cross (1.0 a match, 4,454 crosses); the ball is here 18 times a match0.0441.0 a matchbox depth half-space: 0.044 goals a lofted cross (1.0 a match, 4,454 crosses); the ball is here 18 times a match0.0441.0 a matchbox depth wing: 0.030 goals a lofted cross (8.9 a match, 40,607 crosses); the ball is here 37 times a match0.0308.9 a matchbox depth wing: 0.030 goals a lofted cross (8.9 a match, 40,607 crosses); the ball is here 37 times a match0.0308.9 a matchbyline centre: too few lofted crosses to judge (14)too fewbyline half-space: 0.045 goals a lofted cross (0.9 a match, 3,914 crosses); the ball is here 6 times a match0.0450.9 a matchbyline half-space: 0.045 goals a lofted cross (0.9 a match, 3,914 crosses); the ball is here 6 times a match0.0450.9 a matchbyline wing: 0.030 goals a lofted cross (3.4 a match, 15,565 crosses); the ball is here 9 times a match0.0303.4 a matchbyline wing: 0.030 goals a lofted cross (3.4 a match, 15,565 crosses); the ball is here 9 times a match0.0303.4 a matchhalfway line

The attacking half, seen from above with the goal at the top. Each zone shows what a lofted cross from it was worth, in goals, and how many were played a match; darker is worth more. Hover a zone for the counts and for how often the ball is there at all.

Method. Every open-play cross played into the air in the Premier League seasons 2011/12 to 2022/23, 126,082 crosses in all. Each is valued by what followed it: the chance it assisted and the crossing team's other shots in the next five seconds, both priced by the project's shot model, minus any shot the opponent got in those five seconds, plus what having the ball where it ended up is worth on average over the ten seconds after that, measured from every time a ball was in that state. The pitch is cut into three lanes across (wing, half-space, centre) by four bands along it (middle third, deep, level with the box, level with the goal area), the lines being the pitch markings; left and right are folded together, having been checked to agree. The centre lane is shown as too few because almost no crosses are played from it. Goals scored directly from crosses order the zones the same way as the valued chances do.

From wide and deep, is anything better than crossing?

For the choices teams usually make, no. Every pass and the dribble are worth less than the lofted cross from the deep wing, and advancing to the byline first is worth the same. Two rare choices are worth more: carrying the ball into the box, two and a half times the cross, and shooting.

Carry into the boxCarry into the box: 0.059 goals a moment (4866 moments, 1% of them)0.059 · 1.2% of momentsShotShot: 0.033 goals a moment (2607 moments, 1% of them)0.033 · 0.7% of momentsLofted crossLofted cross: 0.024 goals a moment (23987 moments, 6% of them)0.024 · 6% of momentsAdvance down the wingAdvance down the wing: 0.024 goals a moment (24561 moments, 6% of them)0.024 · 6% of momentsLow crossLow cross: 0.018 goals a moment (1916 moments, 0% of them)0.018 · 0.5% of momentsPass forwardPass forward: 0.016 goals a moment (82751 moments, 21% of them)0.016 · 21% of momentsDribble (a take-on)Dribble (a take-on): 0.013 goals a moment (14275 moments, 4% of them)0.013 · 4% of momentsPass back or squarePass back or square: 0.008 goals a moment (200290 moments, 51% of them)0.008 · 51% of momentsLost the ballLost the ball: 0.003 goals a moment (39814 moments, 10% of them)0.003 · 10% of moments

What each choice from the deep wing was worth, in goals a moment, with the share of moments in which teams made it. The whisker is one standard error each way.

Method. A wide-and-deep moment is a team receiving, touching or winning the ball on the wing behind the penalty area's line in the attacking half: 395,067 moments in the seasons read, about 87 a match. The choice is the team's next action from there within five seconds. Each choice is valued the same way as a cross above: the five seconds that follow, plus what the ball's state at the end of them is worth. The differences between choices hold after allowing for the score, the minute and how many passes the team had already strung together. The measure counts fifteen seconds in all, so a patient choice whose reward comes later is credited only with what the ball's position is worth on average; a longer horizon would raise the passes, and by how much is not measured. The data record what a player did, not what was on, so this compares what followed each choice rather than what any team would have got by choosing differently.

The full write-up, with every table and what the data cannot show, is football_insights/cross-danger-zones/insight.md in the project.