Perfecting the art of crossing the ball

There’s something about a well-timed cross onto the head of a towering attacker that just sends fans into raptures. Why is this? One theory is that sides usually result to long balls and crosses in their moments of desperation. Over two-thirds of crosses are attempted by sides who are either drawing or are down by a single goal.

Last-minute winners scored from players rising like a salmon out of the river to bury a cross carry with them the emotion, adrenalin and euphoria and capture it in one single moment. Each of these moments galvanises a sense that crossing is the key to winning games, but how true is this?

If crossing is adored by fans, it’s equally loathed by sections of the analytical and tactical community. Deemed a wasteful and inefficient use of the ball, crossing is derided as being a hangover from a bygone era that needs to be stamped out.

A stat that is often trotted out when anyone talks about crosses is David Moyes’ Manchester United side attempting 81 crosses against Fulham. It’s also hard to think that a side now heralded for its use of analytics, Liverpool, less than 8 Years ago spent £55m on Andy Carroll and Stewart Downing with the hope that one’s prowess from crossing and the others propensity to win anything in the air would spearhead a new era for the club.

This didn’t work out as expected and a reported 1102 crosses later Liverpool finished the season in 8th place. If you followed Liverpool last season you will know that their full-backs, Trent Alexander Arnold and Andrew Robertson did produce a lot of crosses but even still Liverpool didn’t even attempt half the number of crosses as they did in 2011/12 season.

A majority of crosses come from 4 areas, as highlighted by Garry Gelade’s Opta Pro talk in 2017. Leaning heavily on Garry’s good work and using Scottish Premiership data from the 2018/19 season we can see how successful crossing is.

To some extent, the wastefulness of crossing is downplayed by most but in reality, last season, crosses in open play from these 4 locations were classed as ‘Not Complete’ 78.6% of the time. That means that nearly 4 in 5 crosses resulted in the ball going directly to the opposition. This only applies to the first phase of play, the initial cross.

If crosses so frequently result in sides losing possession then why do teams do it so much? There were 5667 crosses last season from the above 4 locations, which is not insignificant. You could argue that the uncertainty of defending a cross can result in attacking sides winning the second ball, but for now we’ll just focus on the direct results of a cross. We’ll also only be looking at crosses that have occurred from open play (no indirect free kicks or corners).

Varying types of attack and producing uncertainty in the mind of the opposition is key to producing results so as Garry Gelade points out in his article, maybe crossing still has a place in football. Maybe by looking at the data, crossing can be improved on. If sides are giving away so many crosses can we use the data to drill into which locations are the best to aim for and which is most likely to result in a chance on goal.

Using the data we have from our data partners ORTEC Sport, we can track not only the location of the next action after a cross but also if it was below head height or high in the air. These will be split into front post and back post as well as 3 principle locations within the area between the width of the 6-yard box out to the edge of the 18-yard box.

Cross Location 1 (Orange)

As crossing locations go, this is the least successful place to cross from, with only 19.1% of crosses retaining possession and 7.68% of all crosses resulting in a chance on goal. The best chance of getting a shot on goal came from high crosses to the far post aimed between the penalty spot and the 6-yard box.

Orange - Assist
Grey - Key Pass

Crosses into this location resulted in a chance on goal 18.00% of the time, more than doubling the average.

A lot of these goals appear to come from when opposing sides defenders are running back to keep up with the ball and not tracking the run of the advancing forward who is staying back. An example would be Lewis Ferguson putting away Sam Cosgrove’s floated cross for the opening goal in their 3-2 win over Livingston in December.

Cosgrove turns from provider to receiver for Aberdeen’s second, knocking in a cross by McLennan from our next crossing zone. Although it resulted in a goal on the night (and 15.60% of the time which is again almost double the average), it was not the most successful cross type from that area last season.

Cross Location 2 (Light Grey)

Crosses from here are most likely to connect with a teammate, with only 76.5% of crosses not completed (in the context of crosses this is a high completion rate). 9.43% of crosses from here were key passes (directly setting up a shot on goal).

Orange - Assist
Grey - Key Pass

Crosses aimed near-post and high, between the 6-yard box and the penalty spot created a chance on goal 17.19% of the time. Whilst this is impressive from 221 crosses there was a more successful cross type, albeit from a smaller sample size.

Orange - Assist
Grey - Key Pass

High Far Post crosses into the 6-yard area resulted in an attempt on goal 24.00% of the time from 50 crosses. An example of this would be Rangers well-worked set-piece routine from a short corner in their first post-split fixture, against Heart of Midlothian. Candeias whips in an in-swinging ball to back post that is knocked in by Katic, ironically a player better known for his aerial ability.

Whilst high crosses to the far post in the 6 yard box are the most successful cross type in the league for chance creation (for cross types that have more than 20 attempts) the worst cross type in the league is actually high crosses to the near post within the 6 yard box.

Dashboard 1 (25).png

From 53 attempts, non of these crosses even found a team mate. Rangers were the biggest culprits making up 11 of these crosses over the course of the season.

Cross Location 3 (Dark Grey)

Crosses from the dark grey zone are probably more successful than you might think, with 22.1% of them retaining possession. They however only directly result in an attempt at goal 8.40% of the time. The size of this zone means that the types of cross vary more, and more crosses are played into the area between the penalty spot and the edge of the 18.

Orange - Assist
Grey - Key Pass

Coincidently, high crosses aimed to the far post are converted into chances 15.63% of the time from 64 attempts. One such chance that resulted in a goal was Rangers opening goal in their 7-1 demolition of Motherwell last November when a man who is no stranger to a cross James Tavernier found Arfield free in the middle of the box.

Cross Location 4 (Blue)

Finally, the blue zone, is one of the lowest completion rates at only 19.90% but has the highest percentage of crosses that result in an attempt on goal at 9.84%. This is probably not a complete surprise. Crosses are more likely to be played into packed defences when taken from the half-spaces within the 18-yard box, but when they come off the receiver is more likely to be in an advantageous position to shoot.

Orange - Assist
Grey - Key Pass

The most likely type of cross to produce a shot is one that is aimed Low to the Far Post, from 89 attempts, 21.35% resulted in an attempt on goal. Flo Kamberi’s equaliser in the highly charged tie at Easter Road in March is a prime example of this kind of pass which came at the end of a quick counter-attack.

Considerations

Using data to analyse crossing, whilst may seem like a good idea, it is not without its problems. The location of a result of a cross is dependent on their being an action. The true intention of the cross is lost if the ball passes through without being touched or equally if the cross is blocked. These are things that with detailed video analysis you can parse out by looking at the movement of attacking players in the box.

A common way of evaluating how effective an attack is is to look at the xG of the resulting shot. Some xG models factor in assist type in their algorithm and as such, shots that occur from a cross are often given a lower xG score than chances from the same position set up by a short ground pass for instance. With that in mind, there wouldn’t be much use in evaluating crosses using xG.

Its important not to forget that crossing isn’t just about directly assisting chances on goal. As previously mentioned, the opportunity to pick up the second ball is always an option. Using the data we can look at how frequently a possession moment that includes a ‘Not Completed’ cross still contained a shot. From 4456 crosses 12.72% still resulted in the same side having a chance on goal. An example of this is Dundee’s opening goal in their 2-1 defeat to Kilmarnock last October. Calvin Millers original cross is poorly defended and falls nicely for Adil Nabi who still has some work to do to give himself some space to get the shot off.

Conclusion

Even the most effective crossing types are just more effective ways of producing low-quality chances, and even then, they are not that effective. That’s not necessarily a bad thing, as we have already covered, mixing up the types of attack that you utilise can have its advantages and create uncertainty for defending sides (although if it isn’t coordinated correctly you can end up confusing your own team just as much).

However, if the absolute best you can hope for is creating a chance once in every five crosses, and even then the chances are going to be at best converted at rate of one in every ten, you are still relying on around 50 crosses per goal scored (which is ridiculously high and almost certainly unobtainable) and you do have to question how effective a Plan A that is dependent on crosses really is.