Finding the next Trent Alexander-Arnold with event-level data scouting.

Trent Alexander-Arnold has had a remarkable start to his career. 

His record of assisting goals is exceptional, yes it helps to be playing passes to Salah, Mané and Firmino, but the quality of his passing would benefit any attacker.

So, from a technical scouting perspective what can we pick out from his data profile that might help us find players that could potentially play in a similar way?

Our basic data profile gives us some very interesting information. All these figures are in comparison with other full-backs playing in the Premier League.

The standout stats on his attacking profile are his ability to create chances (expected assists), through balls (passes that turn defenders) and his volume of passes into the penalty area.

On the defensive side, his performance is well below average. He hardly challenges for the ball and when he does he loses the challenge far more often than other fullbacks. The only area he excels in his how few fouls he concedes and his very high level of interceptions. And this highlights exactly why you need to be careful when searching for players using just raw statistics. He doesn’t need to defend much in the Liverpool system. With their high pressing style he intercepts rather than tackles, and is positioned so high up the pitch that he doesn’t get involved in battles against the opposition winger.  

Where Alexander-Arnold really excels is in his Passing and Progression metrics. He doesn’t carry the ball much, his ball retention and passing accuracy levels are fairly average. What he does do is pass the ball into dangerous areas, a lot. He plays long passes into dangerous areas, time and time again. When he gets into the final third he will pass the ball forwards, progressing the ball towards goal. And he gets into position to do this a lot, playing around 60 passes a game.

So what do these basic measures tell us? That if you want to find an Alexander-Arnold you need to concentrate on a player’s ability to play dangerous (high xA) progressive passes into the opposition penalty area.  

So if you just have Wyscout you could do this, put the filters on and see who comes out of an advanced search.

No surprises to see Kimmich there. Quite a few midfielders appearing as well as we’ve restricted it to players operating on the right side of defence and midfield.

If we age restrict it to players U23 we get

This list throws up some interesting names. Tom Edwards was regarded as a more defensive full back early in his career but seems to be showing up well as an attacking force. Definitely some names I’d be checking out if looking for an attacking right back.

But these are just basic statistics.

The great thing about working directly with clubs is access to more advanced data. Obviously we can’t share everything publicly but I’ll explain our process. Using an Opta data set with details on pass locations we can isolate specific types of passes.

We want our Alexander-Arnold type player to be playing the same type of passes. Long, raking through balls into the penalty area from between the halfway line and the 18 yard box on the right-hand side of the pitch.

Using our Pass Score system we can do just that. We isolate the area of the pitch we want the passes “played from”, and the area of the pitch we want the passes played in to. This allows us to pick up all the players playing these type of passes regularly and whether the passes were successful or not.

But that is not all. We have also developed a method of sequencing event data that allows us to “fast forward” play. There is obviously a quality difference between a lofted pass from the wing into a crowded box, and a delicate, curling pass into the path of an onrushing forward.   By looking at the events that follow a pass we can estimate the quality of the pass received. Again we need to consider the quality of the player being passed to (a short pass to Messi versus a short pass to a non-professional would have very different outcomes) but it is a good proxy in general.

We can also produce our own versions of metrics to better refine them. I am a huge fan of the progressive passes metric on Wyscout, it consistently highlights players who play forward passes. But with further refinement it can show much more. With our Alexander-Arnold player we want a player who plays long progressive passes within the opposition half but also that these passes are aimed into dangerous areas not just into channels. And we wouldn’t want to exclude players with low pass accuracy as goal creating passes generally are played into crowded areas where completion rates are low. It is not about metrics, it is about the right metrics. And creating your own if none exist that capture what you need to capture.

And what about those cross field passes Alexander-Arnold plays? Well we can measure both X gain (depth of pass) but also Y gain (width of passes). Combine that with our sequencing data and you can see players who successfully play passes with high Y gain that leads to their teams progressing further down the pitch.

The main difference between the player profile data (such as the bar charts or radars) and the event data (whether supplied by Opta or Statsbomb) is that profile data is cumulative, it reflects the volume of actions a player has performed over time. It is great for picking out players who are producing good output but are undervalued.  

The advantage of technical scouting with detailed event data is that you can identify players who perform actions you like, but who are playing in systems that don’t maximise their talents. 

Again take Alexander-Arnold. He was played generally as a box to box midfielder in the youth system. He had energy, a good range of passing, and could finish well. From that position, his ability to strike accurately from distance was put to use scoring goals rather than crossing. He was, and still could be, a very good box to box player but as a “lateral” right-sided player is world-class.

So can we use event-level data (with sequencing) to identify players with the potential to play different roles to a higher level? I’d say, when combined with an overall technical and physical assessment it takes you a long way towards it.

Event level data is expensive but within the overall budget of a club relatively affordable to Championship level clubs and above. Once you have the data you need people who know how to use it. The great thing about an organisation like ours, who only work with non-competing clubs, is that we can share data insights throughout our customer base. We work to create unique templates to fit with how you want your players to play within your system. Trent Alexander-Arnold is great, but he wouldn’t suit the role of a defensive right back. So our data science team use event-level data to find precisely the sort of players who will suit your system. We can turn this into great visuals you can view in Tableau and non-technical staff can easily navigate.

In the above example, we looked at PL players who run with the ball, how often they do so in central attacking areas and their success rate at it. We mark out player types, then apply the metric to all the leagues our clients have data for. This system allows early identification of players showing good ball retention ability in crowded areas.  

Once we have our short list we scout them. We pull up their data profiles and watch them for physical and technical details that are important to their position, role, and level they will be playing at. We have 6 scouts watching footage every week, working closely with clubs to ensure our reports match their requirements.

So can the system find the next Trent Alexander-Arnold? You will be able to follow our progress over the next few transfer windows and find out!

We are currently working with clubs in the EFL Championship and USL. If you are interested in our services please get in touch.