A defensive stats personal library

Historians sometimes get very worried about floppy discs. Also DVDs, USB sticks, most social media sites. They're collectively haunted by that one big library that burnt down and, like, people reading some old books being such a big deal in Europe that they called it the Renaissance.

I don't know if football analytics has ever been great at passing on knowledge. There's often an incentive not to, when people are working professionally. And also, to be honest, the old forget what it's like to be young and the people who gained knowledge first-hand just assumed it got passed on.

This newsletter post will be a web article that I'll keep as a mini-library of useful and formative works on defensive statistics. It's an area that's notoriously awkward, and an area I have a moderate specialism in. The posts in this library aren't necessarily the invention of the steam engine or the metric system, but they'll be a collection of ideas or research that are good to refer back to, for inspiration or for warning signs.

Let's avoid having a Renaissance. It means the old knowledge died.

Original PPDA (2014), Colin Trainor - https://statsbomb.com/articles/soccer/defensive-metrics-measuring-the-intensity-of-a-high-press/

PPDA (Passes Per Defensive Action) is a way to gauge pressing intensity and a rare example of a metric that's stuck around. The post is also a good example of that era of blogging. It digs into event data; it notes that the metric may be a work in progress, and notes possible limitations; it isn't a whole and complete thought (very blog-y), building off a previous post that Trainor wrote. Reading the whole thing won't be for everyone, but for those interested it's an invaluable bit of stats history.

PATCH (2016), Thom Lawrence - https://deepxg.com/2016/02/07/defending-your-patch/

PATCH, as a metric, hasn't really stuck around*, but drawing convex hulls around a player's defensive actions sure has. This being blog-era posting, Lawrence notes other examples of similar work being done at the time at the bottom of the piece. There's also a lot of considerations in here of what went into the metric, designed to quantify ball progression through a player's zone of defensive responsibility, which will be useful for anyone approaching defensive stats.

*I think? Please correct me if I'm wrong

Centre-back shot suppression (2017/2019), @TheGersReport - https://www.modernfitba.com/blogs/2019/6/1/the-best-shot-suppressing-centre-backs-in-the-scottish-premiership

This is a 2019 post although the metric was originally devised in 2017*; it's a similar concept to PATCH, and counts chances in zones designated to central defenders. Again, it's worth taking a look at this, and the 2017 posts, to see what things were considered.

*The 2017 link is now dead. Wayback Machines do exist, though for the sake of this post I will stick to things I don't need to use a digital seance for

1v1 ratings (2018), Garry Gelade - https://www.statsperform.com/resource/a-new-metric-for-evaluating-1v1-ability/

Duels are 1v1s and there are methods to rate participants in head-to-head contests that can be taken from elsewhere. This one uses something called the 'Bradley-Terry' model, but other similar posts have used Elo ratings (which you may have heard of if you are familiar with chess). Besides the use of a model, there's some useful straightforward data analysis too.

Defensive duels and possession adjusting (2021), Me - https://www.getgoalsideanalytics.com/duels-position-possession-adjusting/

Like part of the previous article, this notes that tackles are won at different rates on different parts of the field. It also looks at various correlations between the possession share of a team and the number of defensive actions that players make, in an attempt to examine whether it makes sense to 'possession adjust'* these stats. Part of this breaks down the correlations by different player positions (i.e. defender, midfielder, forward) to see if there are differences. Another feature of blogging though: working with available data; this piece only used two seasons of freely available StatsBomb WSL data.

*Adjusting statistics in an alternative way to 'per 90 minutes played' is possible in a number of ways (e.g. 'per X team touches' or 'per X player touches' or 'per X team possessions'). There's a slot in the library waiting for a piece that examines any of these.

Please get in touch with any pieces that may make good additions. You can do so on Twitter (@get_goalside), Mastodon (@getgoalside@mastodon.skrimmage.com), or email (getgoalside.newsletter@gmail.com)

Get Goalside is just about the only place you can get this kind of work. If you like it, you can become an official supporter for just £2 a month.