How can we unearth a gem?

One advantage of using data scouting is the ability to potentially find unearthed “gems” especially those who are playing for teams that aren’t necessarily fighting it out at the top of the table and therefore going under the radar. Sure, most good scouts or judge of a player might have noticed these players before, but how many are we missing? Data in itself cannot give us sure-fire results however it can be used with further analysis as an indicator to highlight some potential stars of the future.

So how can we look at data in a different way? None of what I’m about to say is ground breaking or even an exact science, there’s variables and pitfalls within any “model” or “analysis” however I believe it is a good way of analysing things with an alternative perspective. One question I think analytics can help to answer is “It’s not fair to compare players stats who play for a league at the top league compared to those at the bottom”.  I present to you Relative Metrics. Relative metrics aren’t new. Hockey analysts have used them for time so I am not attempting to reinvent the wheel more bring it to the EFL!

What are “Relative” metrics?

In essence, any stat such as goals, expected goals, chances created etc, all can be assessed on a team level. What I mean by that is if we compare each player in a single team’s performance on a given metric, take the average for the team and assign a +/- number for each player based on how above or below the team average they are we then get a comparable stat for all teams in the league. Lets face it comparing Fleetwood Town’s 310 shots to Barnsley’s 509 shots is going to massively skew the stats to Barnsley in any creative or shot metrics we look at. To show what I mean I’ve chosen to look at expected assists for EFL L1. I’ll be looking at midfielders (DM, CM, CAM and Wingers) only to help balance the search.

Here’s the current Top 10 for each L1 in Expected Assists p90 based on those who have played 900 mins or more:

l1 top 15 xA

Won’t spend too much time on this list but some names you’d expect and one or two performing exceptionally for less fancied teams but generally a list I think most could put together based on knowing a bit of data and who are the creative players in the league.

Here’s the Relative xA top 15:


So the first observation, lots of these names match the “normal xA” list. However, many have changed positioning and their ranking. Mitch Pinnock for example at AFC Wimbledon, the bottom team in the league looked very good in the previous chart as 8th highest xA in the league, When you compare his relative stats, he is 4th highest in the league. A seriously creative player. Ash Hunter is very much the same case, 12th in the original list and now as a high as 6th in relative xA. I won’t expand on analysing these numbers at this stage as there’s various ways to exploit the data but as many of the names above are similar from 1st to 2nd list the following shows the biggest movers in terms of xA to Relative xA.

Player Team Minutes played xA per 90 Rank Rel xA Rank Diff
C. Lines Bristol Rovers 1330 0.15 32 0.082 22 10
A. Barcham AFC Wimbledon 1794 0.14 39 0.052 30 9
R. Charles-Cook Gillingham 1014 0.15 33 0.061 26 7
D. Parrett Gillingham 1105 0.14 40 0.051 33 7
A. Hunter Fleetwood Town 2099 0.22 12 0.14 6 6
J. Payne Bradford City 2752 0.21 16 0.125 10 6
S. McConville Accrington Stanley 2949 0.2 19 0.11 13 6
J. Ginnelly Walsall 1632 0.16 27 0.083 21 6
S. Finley Accrington Stanley 2307 0.15 34 0.06 28 6
M. Pinnock AFC Wimbledon 1850 0.23 8 0.142 4 4

The biggest mover shown was Chris Lines. Lines is a key cog for struggling Bristol Rovers side and this is reflected by fact in terms of expected assists Lines rates as 0.082 better than the average of all other Rovers midfielders (8.2% better than the others in the midfield for creativity).


Another way to use relative metrics like this is to say how a player in a certain team is performing in 1 stat comparable to others. For example I’m going to take Ben Reeves because why not! Reeves is currently running at 0.13 xA p90 which ranks at 49th from my database. Nothing special. Now we look at his relative xA and he has a rating of 0.064 which ranks at 25th. That’s a 24 place jump when we compare his performance against his team mates and rank that across the league, a much clearer indication of the qualities he may possess in a more creative team. Positional responsibilities and tactics aside as we are looking broadly here players of a similar (just above or just below) rating to Reeves come out as Sam Finley (Accrington), Josh Morris (Scunthorpe) and Jamal Lowe (Portsmouth). Knowing the names there and how impactful they can be on their teams I think we now have a better understanding of how good Reeves could be.


As I said nothing ground breaking here, certainly other metrics like key passes, or even something as simple as shots or dribbles but I feel like xA isn’t directly based on positioning. I’ll certainly be revisiting this and adding to it but the analysis has thrown up some names that if I were a championship club I’d be looking further into!


%d bloggers like this:
search previous next tag category expand menu location phone mail time cart zoom edit close