What type are you?

On drawing evocative pictures to describe players

Over the summer, I’ve been enjoying the content put out by Scouted. They write a lot - most days see something worth a read from their newsletter. They cover everything from finding the ‘next big thing’ to trying to explain what makes the best players what they are. They’ve had some great guest writers over the last few months (including Ashwin Raman, who was the original data wunderkind, turning his online content into a role with Dundee United when he was barely out of nappies).

One of their current projects (organised by Jake Entwistle) is the “Archetypes” - a way of profiling players that’s evocative and easy to get your head around¹. How do you compare strikers like Brian Graham and Tony Watt? They play in the same rough area of the pitch, but … they’re not the same. Harry Milne and Paddy Reading? Same position, yet not the same. Myles Roberts and David Mitchell? You get the idea.

Most of us can’t really look at a number and know what it means in the real world. 8.7 progressive passes? Duel win rate of 43%? OBV of +0.7? xG of 0.24? Nice on a spreadsheet. Not so much on the grass. As Homer once sang to Flanders, you don’t win friends with numbers.

On the other hand, most of us are ‘story’ people. We relate to stories, to pictures, and to ideas. We’re much better at comparisons: This is like that. It tastes like chicken. The Maryhill Messi. He’s a poor man’s Haaland. Watching him is like the sound of nails on a blackboard. Comparison brings ideas to life.

Jake is developing a description of players that uses clear picture words (and a bit of geekery) to segment players. There’s a great primer on their methodology here.

Basically:

• All players want to maximise their team’s production - basically ending up with more goals than the other guys

• Production of this sort can be achieved in a variety of ‘primary means’ - possession (gaining and keeping the ball), progression (moving the ball towards the goal) and penetration (getting the ball into the most dangerous zones)

• A player’s style is defined by which of these they really major in

• They can achieve these things by doing one of several activities: passing, dribbling, ball carrying, movement and pressure

• The archetype a player fits into is defined by how they contribute to the eventual production of goals - which primary means they focus on, and which tools and activies get them there.

So they have pictures like ‘defuser’ - a player who is good at evading pressure from opponents, and so is good at retaining possession and relieving the attentions of the other team on his teammates, a release valve who turns danger into security (think peak Banzo). Or a ‘power forward’ who majors on penetration into the box, by good vertical movement. Even off the ball, they’re driving to the danger zone, making opponents fear to leave him unmarked (Viktor Gyökeres or Ollie Watkins in England are good examples - maybe Scott Tiffoney in his prime, running at defences consistently is a good example for Thistle fans).

Anyway, I’ve found their descriptive picture that evokes a clear idea a really useful way of thinking of players.

So over the next few months, I’m going to try to do something similar here at Jagalytics Towers. What are the championship Archetypes? Some of them will be the same, or at least analogous, to what the Scouted lads have found. But I’d like to work a bit on what defines a profile in our league.

First: a bit of groundwork.

How do you classify the different types of attacking players you find in the Championship?

To help us out, I’m going to enlist someone else whose work I find thought-provoking - Spencer Mossman. Spencer is the king of scatter-chart-as-segmentation-tool. He often picks pairs of metrics, and then plots groups of players using them. Players who cluster together are similar in some way - it’s not just about ‘who is the best’ but also: ‘what groups of player types can we find’?

This week, he had two charts in particular that I decided to recreate:

• One on the usage and expected output of strikers

• One on the athleticism and end product of young attackers

I’m going to use these two scatter charts as a starting point for profiling and categorising attacking players².

Expectations and involvement

The first way of profiling attacking players is to look at how much they touch the ball in the attacking third, and what they do with those touches.

I’ve focused for now on players with at least 50 touches in the attacking third of the pitch so far this season, according to Sofascore (which has the interesting effect of filtering out players who don’t get as many touches but use them very well, like Tony Watt and the boy Fitzpatrick). I’ll probably revisit this in a few weeks to see what’s changed.

Players who touch the ball a lot in the final third of the pitch are those who are the conduit for a team’s attack - the players teams want to get on the ball most because they’re the key to unlocking final production. Or, in other words, players likely to be involved in producing goals.

They can produce goals either by scoring or by creating good chances for team-mates. Thus, the other dimension we’re going to look at is how many ‘expected goal involvements’ the players have generated. This is a combination of:

• xA (expected assists) - a measure of how many goals team-mates would be expected to score as a result of the chances the player has created for them. It’s related to xG … effectively it’s measuring ‘this pass created a shooting opportunity that was equivalent to 0.24xG.’ Making passes that create good chances, or making lots of passes that create any chances, maximises this number

• Non-penalty post-shot xG, or npPSxG³ - a measure of how good the player’s actual shots are (not just how good the chances are). I’ve excluded penalties, because they’re a bit different to regular shooting, and a bit more subject to the vagaries of Championship grade refereeing. And I’m using my own post-shot model that takes into account the quality of a player’s shooting, not just the chances they get.

In the chart, the further right you are, the more you’re used in the final third by your team. The further up you are, the more expected goal involvements you have (either by your own chances, or the chances you create for others). Mossman has a ‘continuum’ of attacking roles (the diagonal line I’ve recreated here, though in reality it’s probably more of a curve) - it’s ok to be high usage and have fewer goal involvements (Lamine Yamal appears in that zone in one of his charts) - it could mean you’re the guy that calls the plays and runs the attack, even if you don’t do the final ‘thing’.

At this point, let’s just note a few things:

• Logan Chalmers is the only Thistle player who meets the threshold. Fitzy and Watt (and others) are not far away from joining the chart (probably near the top left). In a week or two they will probably join the chart in pretty healthy places

• On the continuum, there are clearly groups or clusters or players. Andy Tod sits in a different zone than Dylan Easton, for instance.

• Mossman calls the high touches but lower actual expected involvements ‘point guards’ - a basketball reference. This is the player who runs the attacking plays, to whom the first ball into the zone is passed, and who creates the patterns that define how his team attacks. As the point guard goes, so goes the team’s attack. Easton and Sidibeh at St. Johnstone are firmly in this zone. They’re high involvement, if slightly lower-end production for now

• There are also highly efficient focal players - Tod at Dunfermline (as a goal threat), Mallan at St. Johnstone (as a creator), and Chalmers at Thistle (as a bit of both) do so with a lower number of touches needed to generate a lot of expected goal involvements.

• Josh McPake is … out on his own. He’s had about 40 more touches than anyone else in the league in the final third, and he’s generated more expected goal involvements too. Maybe this will regress and become more ‘normal’ over the season. But at this point, he’s the most compelling attacking player the league has seen in a long, long, long time.

If all you care about is efficiency, then that gives you a slightly different view. Who generates the most expected goal involvements on that chart per touch in the final third? A more efficient player doesn’t need many touches to generate an expected goal or assist.

Interestingly, Logan Chalmers is top 4 for this measure - he generates around 4 expected goal involvements per 100 touches in the final third. That’s almost twice as good as the average for these top attacking players. Only three others are more efficient - Dow at Arbroath is the most out-there on the list, and most likely to be under fans’ radars. Mallan and Tod are both very efficient players, who generate a lot of expected involvements with not many touches needed.

Physicality and chance creation

The second lens to use on attacking players is how well they can impose themselves on the game physically, and what their volume of end product looks like.

This time, rather than including the quality of chances, Mossman just looks at the volume - how many shots does the player get and how many shot-creating passes do they make? And to measure their physicality, he uses how many ‘ground duels’ the player wins. This covers every activity where the player has to fight with at least one opponent for the ball - either because they’re in possession, dribbling, and so need to beat their man, or because, for instance, they are out of possession and tackling a defender as they carry the ball out.

The idea is that physical players can impact the game by forcing their opponents out of position to cover them, or through a desire to damp down the attacker’s threat, and in so doing, they create space ‘passively’ for team-mates⁴ and discomfort in the defence. More ground duels won is an indication that the player is good at making defences work hard, pulling them out of position and committing them to mark. Look at me! says the physical attacker. I’m the captain now!

End product is a measure of how often this player is the final (or penultimate) brick to be laid down in a team’s attack.

• Once again, Josh McPake is out on his own (this time, largely as a result of the sheer volume of dribbles and ground duels he managed against Morton). Let’s leave him be for now

• Five players are elite in terms of end product - including Logan Chalmers. Hale, Easton, Cooper and Mallan are the others.

• 10 players are particularly strong in terms of physical output - including Chalmers and Fitzy

• Three players, therefore, are top-tier in the league for both final product and physical output - Easton, Chalmers and Josh Cooper. These are good players to explore for our archetypes.

So what?

If we put this all together, Logan Chalmers (and to a lesser extent, Josh Cooper) is:

• One of the most efficient final third players in the league

• One of the most physically proficient attackers

• One of the best for ‘final product’

On the other hand, you have Dylan Easton, who:

• Is very physically proficient

• Is high for final production volume

• Is high usage, but less efficient at turning that into expected goal involvements

That gives us the seed of a couple of attacking profiles (ignoring McPake who is just a bit hard to quantify other than ‘too good for the league’ right now).

I’ve not finished thinking about it yet but perhaps you have a good idea how you’d classify these types of attackers? Suggestions on the back of a stuck-down envelope…

1. See here for example: https://scoutedftbl.com/noni-madueke-arsenal-nottingham-forest-stats/

2. Just as an aside, I’m going to replicate Spencer’s approach of going with absolute volumes of activity rather than normalised to ‘per 90 minutes played’ - I think for now it shows some interesting patterns. It does mean that some of the charts are weighted to players who play more minutes (impact subs don’t have as much time to put in the numbers, for instance), and - as we’ll see - it biases towards players who have had one or two blowout games (Josh McPake and the boy Tod for instance). I’ll shift to p90 metrics soon, but for now, I think it’s interesting, whilst acknoweldging that this approach has shortcomings

3. Ok, ok, I know it’s a mouthful. Just run with it for now.

4. I’ll probably talk about this again in future … but not right now