How to score goals, statistically

The FIFA World Cup 2026 is ongoing and — besides all the scandals — there are people actually playing football. Even though (or maybe because) I'm not really that much into football, watching a match makes me wonder: Where should one shoot from, ideally? And is the top corner actually the best spot to aim for?

For my analysis of the best shooting positions and goal targets, I used StatsBomb Open Data — a free subset of an extremely rich football event dataset that StatsBomb distributes commercially. (Shoutout to StatsBomb for making such a gem available to the public!)

The dataset covers 80 competition seasons across 24 competitions — from several men's and women's World Cups and Euros to the Champions League to entire league seasons of La Liga, the Premier League and the FA Women's Super League. It records every pass, dribble and tackle, but what we’re going to focus on today is the shots: 101,227 of them, 11,328 of which ended up in the net.

So what is the ideal position to shoot from?

To answer this question, we divide the pitch into equally sized squares. For each pitch cell c , we can calculate the empirical conversion probability:

where "all shots" really means all of them - saved, blocked, and the ones that ended up as far as the last row of the stands.

The ideal plot for this type of data? A heatmap. Note that this visualization only considers regular shots happening during open play. It is intentionally missing penalties and shootouts, direct free kicks, corner shots, and headers. Also note that all cells with fewer than 10 shots are plotted as 0, since at such low sample sizes a single lucky screamer would light up a red cell. Here’s a heatmap that includes all shots for the curious.

We can confirm the obvious: shooting from right in front of the goal has the highest chance of scoring. It's also reassuring that the map looks pretty symmetrical — there is no reason the left wing should outscore the right.

In general, an open-play shot converts about 10% of the time.

Around 8 yards, or 7 meters, in front of the goal, where the chances of scoring drop from almost guaranteed to 50/50, there's a semi-circular hot zone where the probability hovers between 10% and 30%. Beyond that, it fades into single digits quite quickly. Interestingly, the map doesn't just keep fading: there's a belt left and right farther out where the odds tick up again, and even some long-distance shots beyond 46 yd (~42 m) that beat spots much closer to the goal.

But players don't shoot from 40 yards under normal circumstances. Those attempts mostly happen when an unusual opportunity presents itself. It is an example of selection bias: the red cells far from goal don't tell us that shooting from there is always a good idea; they only tell us that when players do decide to shoot from there, they tend to have a good reason.

And what's the best place to aim for?

Answering this question isn't as easy. We first have to solve a problem. Once a ball reaches the goal line, the dataset gives us its position — the x, y, and z coordinates of the center of the ball. But what about shots that don't reach the goal, because the goalkeeper catches or parries them on the way?

Fun fact: This is a form of statistical censoring. If we only considered shots that made it (almost) all the way to the goalmouth, we would fall prey to survivorship bias. The ones we'd keep would be exactly the unsaveable ones, skewing our analysis.

Fortunately, there is a workaround. The data does tell us where the keeper caught or deflected the ball. We can use this to extrapolate the trajectory and make an educated guess about the position at which the ball would have crossed the goal line, hadn’t the keeper interfered.

Obviously this is a somewhat rough estimate. Balls don't fly in straight lines — there's spin, wind, drag, and gravity involved, and bending the ball is a common tactic. It's the best we can do without major effort, though, and it's less bad than it seems at first glance: for a heatmap we aggregate within cells anyway, so small errors don’t matter at all. And the law of large numbers helps too: some balls curve left, some curve right, so the individual errors cancel out in the aggregate.

Besides penalties, direct free kicks, direct corner shots, and headers, two more kinds of shots are excluded here compared to the map above: those whose (extrapolated) path misses the frame and shots blocked by defenders.

Once again we can confirm the obvious: aiming for the left or right side of the goal is far better than shooting straight at the center, where the keeper usually stands.

And the success rate is nearly symmetrical: shots aimed at the left half of the goal went in 30.1% of the time, and 30.3% for the right half.

You can also see the shape of the goal posts and crossbar along the edges of the heatmap. That's because we track a single coordinate — the center of the ball. A football's diameter is 22 to 23 centimeters, or roughly one and a half cells in the heatmap, so when the ball hits the frame, it has a good chance of bouncing back out.

Beyond open play: free kicks and penalties

To spice things up a little, we can also look at scenarios other than open play. Since there's far less data, the cells have to grow from 0.5 × 0.5 ft to 2 × 2 ft to keep a meaningful number of shots per square.

Direct free kicks first. Even if a free kick is on target and not blocked by a defender, it only converts to a goal in 20% of cases, compared to the 30% above. (Still, direct free kicks are more successful than open-play shots from the same distance.) For free kicks, it’s clear that aiming for the corners is vital, especially the top corners.

Penalties are a different sport altogether: 74% of all penalties score a goal, and 78% of those that hit the frame. Hit anywhere near the top, and they're practically unstoppable. Shooting straight up the middle seems to work great too.

So the top corners are where to aim, right?

In open play, probably not. A ball that arrives just inside the top corner is nearly unsaveable, and the heatmap of the goal above glows accordingly. But that visualization doesn't show the attempts that sailed just over or hit the stanchion, and those failures are the tax on aiming high.

The heatmap below, on the other hand, shows the distribution of attempts: which parts of the goal did players actually aim for?

Turns out that players actually keep the ball low (and safe) in the majority of cases. There is a clear trend to aim for the corners, though.

In the end, a century of coaching wisdom survives contact with 100,000 shots: close beats far, low beats high, corner beats center. Some might be disappointed by how predictable and by-the-book the best shots are — but now at least there are heatmaps to prove it.

That’s all from me for this week. I hope you enjoyed the read. Watch out for another Weekly Chart next week, this time by my colleague Luc!