Finding the bench to mark against
In this season of resolutions, one that has not yet failed for me is this: if you don't like a film/TV show/band, then look up what that artist likes. At best, you start appreciating aspects of something you didn't like; at the very least, you understand their choices. It's easier to judge something in context than in isolation.
Similarly, - go with the segue - if you read a research paper on running data, then at some point you'll get a line like: "This research could be used to design more match-relevant training sessions for X, Y, Z." Because designing the preparation for athletes depends on the context of what they actually do in-competition.
With that in mind, if you have some time off over January and want some juicy reading material, I'd recommend checking out this Google Scholar list (of works that cite the 2018 paper, 'Are Current Physical Match Performance Metrics in Elite Soccer Fit for Purpose or Is the Adoption of an Integrated Approach Needed?'). Save that Waterstones voucher for another time.
The notion of 'needing to know what players are required to do' is the same for 'event/on-ball' data too, and the results can be quite surprising. Recently, the Opta Analyst recently wrote about Martin Ødegaard and some of the areas he excels in. At time of writing, he had more line-breaking passes into the box than any other Arsenal player, and averaged more possession regains in the final third of any Premier League player. However, those figures were only around 2 per 90 minutes and 1 per 90 minutes: two skills that coaches (and fans) desperately want in their players, but which happen only a little more often than fans getting hitting in the face by a stray shot.
High-impact events in football are rare - that's not revelatory news, but the extent of it is more far-reaching than high-quality shots alone. Here's a selection of stats from the current Premier League season (only counting players who've played a decent amount of minutes, time of writing 01 January after 18-19 games):
- Only two goalkeepers face 5 shots on target per game, on average
- Zero players are taking four shots per game, on average
- Only a little more than a dozen players are making three or more tackles per game, on average
- Only three players are making two or more interceptions per game, on average
- Only seven players carry the ball into the penalty area at least twice per game, on average
Name a consequential football action, and the chances are that you can count the tally for the league's third-best player on your fingers. Maybe that's just the nature of something being 'consequential', but it's a strange kind of situation where looking for a 'midfield destroyer who can break up the play' means someone who makes a particular action less than half a dozen times in a 90-minute game.
Part of the solution to this 'problem' is to give people context of when a number is big and when it is small. But is this, in a real sense, a 'solution'?
Players who make 'consequential' actions on a regular basis are unicorns. If you're looking for a striker who takes lots of shots, you might think of Erling Haaland... who takes a shot once every 20-25 minutes. And so, maybe the search should be for slightly-less-consequential but slightly-more-frequent actions, or traits, that players show.
Martin Ødegaard may 'only' average two line-breaking passes into the box per 90 minutes, but his control places the ball in the right spot for him to make improvised passes without being dispossessed, his weight of pass is astonishingly well-controlled, his frame of mind is towards reward rather than risk-avoidance. Some of these will show themselves in other statistics.
Using event data is often about working out proxies for valuable skills. There is, clearly, a literal value to players who make lots of passes into the final third or penalty area, but there's also implicit information about technical ability, vision, composure, et cetera. Generally speaking, the further you get from goal, the more that the value of implicit information will outweigh the value of the literal information.
At the moment, this post is skating pretty close to gate-keeping; making it seem like data is Aramaic, able to be deciphered only by a special, learnéd few. Yet the aim is the opposite - maybe data is Aramaic, but how can we bring Aramaic to the masses?
One of the most efficiently thorough lessons I learnt about football statistics came through typing stats into a spreadsheet: despite the fuss made by TV commentators, players who made the most clearances were generally on bad teams, and the best defenders did not tend to make many clearances. The reasoning is simple: bad teams come under the type of pressure that leads to clearances being necessary; good teams, where the good defenders play, do not. But how do you signal that to people, many of whom approaching stats for the first time, without requiring the (useful but tedious) time spent in Microsoft Excel?
It would be tempting to try and link stats to success-based outcomes like 'winning', but probably a mistake: there are far too many factors at play. And people have a tendency to see causation in correlation, no matter how many times you ask them politely not to.
But knowing that Brandon Flowers listened to Depeche Mode and Roy Orbison growing up makes The Killers' back catalogue make a lot more sense. There's something in that, I think. Maybe.
Hey: if you didn't like this blog, try and work out who Get Goalside's influences are. That'll make it better.
Cutting room floor
Bits of internet research I did but didn't end up putting in the main post.
The Pablo Picasso museum in Málaga is, from memory, quite good for putting Picasso's work, and his successors, into context of the art trends and global events that they were experiencing.
The difference between the first half of Hot Fuss and the majority of The Killers' other work always kinda surprises me. But knowing that he was big into Depeche Mode's Songs of Faith and Devotion as a teen, but that Oasis was a deliberate influence for Hot Fuss makes it make more sense.
UK school examinations body AQA have a document outlining their current policy/thoughts around comparative judgement as an alternative to rubric-based assessment. It can kinda be boiled down to 'we like it but it's not viable at scale right now'.