It’s been a long time since I updated the long-term trend graphics and I thought it was also worth giving them an upgrade. The previous versions are explained in full here (and below) but this time around I’ve added two additional charts and formatted them to fit a phone screen like I did with the recent squad graphics.

Explanation

These are adapted from a very similar design by the excellent Swedish blogger Zorba138 intended to track a club’s long-term performance and also where they’ve been under or over-achieving based on the quality of chances created in their matches.

There are three charts:

1. Overall, which compares a club’s goal difference (on a rolling 10-match basis) to its expected goal difference (more on that below);
2. Attack, which compares the average number of goals scored to expected goals scored;
3. Defence, which compares the average number of goals conceded to how many they were expected to concede.

Previously I only produced the overall chart, but as this combines the attacking and defensive data it wasn’t always clear what was driving the trends and differences.

Each chart has two lines:

1. The blue line shows the rolling average of actual goals (be it goal difference, goals scored or conceded) over the previous 10 matches;
2. The red line shows the rolling average of expected goals, based on my model’s assessment of shot quality.

Comparing these two allows us to see how a club’s performances have changed over time and whether there were any interesting differences between chances created and goals scored.

Each chart is shaded as follows:

• Blue shaded areas show an overachievement, where the rolling average of a club’s actual performance outstrips their expected goals data;
• Red shaded areas show an underachievement, where the quality of chances (according to the expected goals model) looked healthier than the number of goals which resulted from them.

Over the long term we’d expect the two lines to converge unless there’s a significant difference in a club’s attacking or defensive skill compared to the average for the division (or perhaps that my data doesn’t pick up something unusual that they’re doing). We can’t tell from the data alone whether skill or luck is the cause, but the longer a difference persists the more I’d suspect the former.

Other changes

I’ve also indicated on the charts when a new manager has taken over (using a dark dashed line) as this can provide useful context. I’ve included caretakers in this, as in some cases there was a significant gap between one permanent manager leaving and another being appointed.

Club-by-club graphics

I don’t have data for the National League or below, so Barrow, Harrogate, Leyton Orient and Salford all have less information on their charts than the rest unfortunately.

Carlisle are a good team to focus on as an example. They were enjoying a great run of results in the middle of 2018-19 when John Sheridan left to take over at Chesterfield. However they were running incredibly hot in front of goal – scoring about a goal per game more than the underlying data suggested – and this proved to be unsustainable. It wasn’t until their next managerial change the following season that performances started to improve: Chris Beech has presided over steady improvements at both ends of the pitch and results have followed.