New Projected Table suggests Rangers may find it difficult to finish in the top three

6117524-_Read-Only_.jpg

written by:  @TheGersReport

Background:

There are so many prediction models out there, in which data is used to project what a league table should look like.  The objective with this attempt at a model for the Scottish Premiership is to leverage stats that historically have had the greatest impact on the league table, while also factoring in what a team has already accomplished in a season.

The inspirations come from way more complicated models than my own, created for the NHL by Dom Luszczyszyn and by Micah Blake McCurdy.

I will go into much more detail about the origins of my model at the end of the post, but it's rooted in the fact that goal differential has historically had a 94% correlation to the rate in which teams earn points.  It's obvious - score more goals than you allow leads to winning more matches.  Duh...

I went back to see what extent goal differential impacted the league table going back to every Scottish Premiership season from 2009-10 to 2017-18.  That includes over 5,000 goals & more than 5,600 combined points earned.

The results looked this:

 For a live version of this - go  here

For a live version of this - go here

Again, that's a 94% correlation between a team's goal differential at the end of the season & the rate of available points that they earned.  

The model I've come up with combines a team's current form, with their Expected Goals data to project how many points a club will have at the end of the season.  Basically, the model banks the points a team has already earned & uses xG differentials to determine how many points they are likely to earn moving forward.  For example, Hearts already has 12 points...that's going nowhere.  Looking at their xG data, the model determines how many points they can expect to earn in the remaining 34 matches.

These projections will then evolve based on:  how many points a team is earning & how sustainable those results are based on the chances they are creating (& allowing).  I plan on publishing these projections weekly, both on Twitter & here on the website.

Here are the current projections.  One table lists how many points a team could expect to earn if they sustain their current form of actual goal differential, along with the projected table that combines their current form & is adjusted for the rest of the season based on xG differentials.

Screen Shot 2018-09-07 at 5.49.21 PM.png

It needs to be acknowledged that these projections will be very fluid for the first month or two of the season.  I'd hesitate to make any truly conclusive projections until each team has played each other at least once (so yeah...don't go run out & place bets on Hearts winning the league).

Now, how can a club use this kind of information?  Let's say you’re Rangers...clearly, their place in this table is the biggest talking point.  Obviously, playing so many minutes with ten men & having played Celtic skews the data.  That needs acknowledging.  But...it would also be foolish to live in a bubble of denial that this team has some issues it needs sorting out.

What's dictating their woeful projections?  If you apply the same model (again explained later at the end of the post) exclusively to the two factors that make up goal differential, goals scored & goals allowed, you begin to get entry points for further analysis.

Since the 2009-10 season, there is a 81% correlation between goals scored & points earned & a 73% correlation between goals allowed & points earned.  Basically, the amount of goals scored has a bigger impact on how many points a team accumulates, but that 73% correlation rate for goals allowed is still very significant.  The apparent difference is that it's harder to win without scoring goals.

Below, you'll find Rangers projected points when you isolate each statistical category.

Screen Shot 2018-09-07 at 6.21.26 PM.png

There is one clear outlier here & it's Rangers projected points when you only factor in Expected Goals.  In their four league matches, Rangers have really struggled to create quality chances.  Again, I know...ten men, playing Celtic...but some of the stats are quite surprising. 

Rangers currently have the seventh highest amount of xG this season - only Livingston, Dundee, St. Mirren, & Hamilton have lower Expected Goals (not the sort of company you'd expect Rangers to be chillin' with).

In an OptaPro blog post from 2017, Sam Gregory wrote:

Expected Goals is typically a more consistent measure of performance than actual goals. Whereas goals are relatively rare events that come and go in stretches, a team or player’s xG output tends to fluctuate much less from match-to-match. Clearly the goals that are actually scored are the ones that win points, but xG gives us more context for evaluating team performance.
— Sam Gregory, January 2017
GettyImages-1026099358.jpg

Ovie Ejaria currently averages 0.01 Expected Assists per 90

photo courtesy of Mark Runnacles (Getty)

Rangers, quite frankly, have been one of the worst attacking teams in the league.  While they have been opportunistic (they have the 2nd best Conversion Rate in the league, despite having the 8th best Expected Conversion Rate), the sheer volume of attacks has been pretty mute this season.

As the team analyzes their performances, they need to examine why:

Rangers...

  • Have the 4th lowest shot totals in the league
  • Have the second fewest pass attempts into the penalty box in the league (only Hamilton have less)
  • Are tied with Celtic for the fewest counter-attacks in the league. Celtic has a total of 71 shots this season, Rangers have 35. Celtic's lack of counter attacks make sense, Rangers less so.

Additionally, of the Rangers possessions that start in their own half, only 46% lead to possession in the attacking half.  Hamilton & Rangers are the only teams who fail to successfully transition into the attacking half at least 50% of the time.  

To add another layer of gloom, Rangers have the second fewest ball recoveries in the opponents half in league.

Now, do I really think Rangers are going to finish fourth bottom in the league?  Of course not.  Steven Gerrard has clearly come in with a priority of fixing Rangers defensive play.  That's still a work in progress.  Some inspired efforts in Europe has shaped the narrative, but the fact that Rangers have the sixth best Expected Goals against in the league suggests there is still a need for continued improvement (seven teams have allowed less actual goals). 

I'd expect those numbers to get better, & I assume that the club is very aware of their issues in transitional play....so that should improve as well.  

Do I think Rangers will be guaranteed a spot in the top three by the end of the season?  That I'm less sure of.  

image.jpg

Even though Hearts won't win the league, it's safe to say that they are better than most of us expected.  Hibs aren't necessarily clicking yet in attack, but their defence is better than they get credit for, while Kilmarnock may have the best defence in the league & the addition of Greg Stewart will make them better in attack.  Meanwhile, Motherwell who were big underachievers last season look to be a team on the rise (& that's with one of their most dangerous players still sidelined by injury).  And, while I think Aberdeen are kinda bad...they always find a way so it's hard to count them out (actually it's not that hard...they won't finish in the top three).

The point is, Rangers will have a lot of competition this season in that fight for second & their current form suggests they have a lot of work to do to get better.  It will be very interesting to see if how much they improve over the next set of games as their schedule gets a little easier & they likely will get to play games with eleven players. 

Some notes...

  • This post was written with the aid of StrataData, which is property of Stratagem Technologies. StrataData powers the StrataBet Sports Trading Platform, in addition to StrataBet Premium RecommendationsAdditional stats are courtesy of InStat Football  
  • The process in which I created this projection model will be explained after these notes.
  • Yes, it's only been four games but you'd be surprised how much you can learn this soon in a season, as @11tegen11 highlighted in his blog post from 2015
  • In my season preview, the numbers suggested that Rangers would score 10.72 LESS non-penalty goals this season; they are currently on pace to score 22.5 less goals.  
  • I've been a Rangers supporter all my life, as have multiple generations in my family.  I acknowledge that most of my coverage of the team hasn't fit the current narrative out there about the team.  However, if the underlying numbers suggest there should be more concern about the team's play, I'm going to shout it from the rooftops with the hope that shit gets noticed & shit gets better.
  • I plan to post these projections regularly - with weekly posts on Twitter & more articles like this one where I pick out one team & highlight how these kinds of projections can become entry points for teams to learn about what's working & what's not.
  • This was written under the influence of Sudakistan & Radio Moscow.
nerds-talent-show.gif

How the model works:

Once I stumbled upon the 94% correlation rate between goal differential & the rate of available points earned...I knew I had to keep this projection model simple.  When you're looking at such a high correlation between two variables....don't fuck with it.  

Luckily, Tableau did the hard work for me.  It provides a formula showing you how to determine the rate of points a team should earn based on their goal differential.

Screen Shot 2018-09-08 at 8.29.14 AM.png

For the projected table simply based on current form...it was easy.  Just plug in the goal differential (prorated to a 38 match season) & there you go.  However, like Sam Gregory suggested - if you are looking for a more consistent way to measure a team's future performance - it's not their current actual output...it's by applying Expected Goals.

For the xG adjusted projections - I bank the points a team already has earned.  Then I apply their xG differential (again prorated for an entire season) to the formula above to determine what percentage of points they would be expected to earn in the remaining matches in the season.  Then, combine the points already earned with the points they are expected to earn moving forward.