The always excellent @TheGersReport wrote on the equally innovative Modern Fitba site (a great addition to the Scottish football online landscape) about which team can expect to see an increase or decrease in scoring this season. The article is here.
The basic principle is that over time a players’ goals and assists totals should revert to the Expected Goals (xG) and Expected Assists (xA) values. The article is more complex than that (read it!) but I want to take the basic principle above and look at what the last two treble seasons tells us about what to expect this season.
The good news is I agree with the aforementioned article!
Over the last two seasons, of you look at Celtic’s total goals and assists versus xG and xA:
Aaagh, lots of numbers! What does it all mean?
There is little doubt Celtic had an exceptional 2016/17 season. Especially domestically where they were unbeaten. I only have xG and xA data for SPFL matches. Celtic MASSIVELY over performed during the unbeaten season. 24 league goals from outside the box hints at this. That was great news for Celtic fans as an unbeaten domestic treble followed. Celtic scored 7 more goals that can reasonably have been expected and created a massive 19.85 more assists.
But last season the cold turkey of regression to the mean (bleurghh) followed with the icy certainty of statistical probability. In 17/18 Celtic underperformed across the season to the tune of 3.39 less goals and 3.8 less assists than could have been expected.
It does all suggest 16/17 was a bit of a one-off. But in our hearts, we probably know this. It will be quoted not far behind 1966/67 in Celtic lore forever.
For Celtic fans, to echo the Modern Fitba article, we should expect to see a wee upswing in goals and assists this season. But only a wee one.
In terms of which players under / over performed their Expected Goals and Assists, the following is in “overperformance” order. That is, those that scored more goals and created more assists than the model predicted are at the top, and those that were perhaps unlucky, at the bottom.
Basically, Edouard achieved around 4 more goals and assists than the Expected Goals model would suggest. Boyata benefitted from at least 2 assists where he passed the ball on half way and someone scored. Forrest had a good scoring streak that may be difficult to maintain. And Ntcham as we know, benefitted from some fortunate deflections for some of his goals.
Of the productive players, Rogic and Dembele in particular under performed compared to the chances they had or created and perhaps can be expected to improve. In fact, most of the regular squad under performed. This was a very obvious and sudden correction back to the mean compared to the previous season.
What does it all mean? Even one whole season is a small sample, so it could go either way! What is less likely is a rerun of 16/17, sadly. More likely Celtic will score a few more goals than last season given the same chances.
A modestly heartening thought.
(PS This season in 3 matches Celtic have scored 9 goals compared to the xG of 5.48 and created 9 assists compared to a xA of 5.282. Can that be maintained?)
Another good summary. Always interesting to see the depth behind the xG and xA measures, and also the variation when the team is doing exceptionally well.
Understanding the variation against the mean measures or expected measures of success is always going to be challenging, because even with the most accurate statistical history (as any betting fan will tell you) the numbers never follow the form when dealing with human beings, even more so when there is an interconnected team of them. Many factors could affect a team’s overall x And xG scores:
Changes to an established midfield pairing, success at an International tournament, receiving an award, selling or buying a key player, the length of grass on the pitch … ;0)
The qualities of the manager must also have a significant bearing on xG and xA overall, as a result of interaction with the players, coaching their belief as much as their skill, “creating a mentality” as it were.
This is a key factor in sporting success all over the world. The NFL to some people is a very loud, brash and over-exaggerated sport, and it’s interesting to note that although NFL players play in generally short bursts of extreme speed and contact, their overall persona of ultra-confident, ultra-physical athletes (even the 25 stone ones) is all about making them 100% mentally ready for every single play they make, in every game, all the time. It’s very extreme, but essential for success.
The difference with football is the key interaction occur between all team members through 90 minutes, with a constant evolving ebb and flow of possession and build up. So many variables can affect the chances created or goals scored during a game that it can get mind boggling (butterflies, tsunamis etc) but it’s always good to see statistical facts nonetheless.
It will be interesting to see how the xA and xG scores move throughout this year. If many of them improve then it must be the impact of the manager which brings the most benefit? IBWT ?