Projected League Table: There’s a new “surprise” team at the top of the table

 courtesy of Kenny Ramsay (The Sun, Glasgow)

courtesy of Kenny Ramsay (The Sun, Glasgow)

written by: @TheGersReport

This is part of an ongoing series in which I will regularly share a projection model that makes data based predictions of the final Scottish Premiership table.  It is rooted in projecting goal differentials for the season & the fact there is a direct correlation between goal differential & points earned.

The model banks points already earned by each club & applies Expected Goal differentials to project out the rest of the season.  For more, check out this introductory post.

As more matches are played:

  • Teams will have banked more (or less) actual points & that will be reflected in the projections

  • The Expected Goals Differentials will become more solidified

This means the projections will build accuracy as each week of the season passes.


So, I guess this is becoming a weekly column. 

In the initial post, I highlighted how these kinds of projections can be used to pinpoint areas that need improving for a team.  In that first table, Rangers were projected really, really low (9th).  Playing Celtic & finishing multiple matches with 10 men led to a tepid attack & poor underlying numbers.  I mean, they were on par with the likes of Hamilton at the time. 

This clearly wasn’t going to last, but the projections identified some glaring needs (an improvement in transition play & the need to get more balls into the box).  The most recent numbers support the fact that they have made marked improvements in those areas (it helps playing with eleven men).

Last week, the projections suggested that Kilmarnock were playing well enough to finish third in the league table.  At the time, they were actually seventh (!) in the real league table.  This projection was most influenced by the fact that Killie had some of the best defensive numbers in the league.  Then they went out & pulled off a 2-1 victory over a Celtic side who are learning that the competition has been kicked up a notch this season.

Screen Shot 2018-09-26 at 10.00.58 PM.png

Let’s see what the Projected League Table looks like now that six matches have been played.  It is still early in the season & these should not be taken as final, locked in predictions.  Rather, these projections can help teams highlight areas of strength & aspects of play that need improving & can also really help track the real form of a team.

Screen Shot 2018-09-27 at 5.51.46 PM.png

There’s been a lot written about what’s gone wrong with Celtic so far this season & (full transparency) I had planned to use this post as an opportunity to peel back the statistical layers of what’s missing this season for the team most of us had predestined to win another league title.

I was looking forward to it…

8b3b5654-568d-4283-92bc-f22b0f918934.gif

Then…the numbers came in & well….the projections reacted to a Celtic loss by boosting them into first place???

shia.gif

I mean…what the actual fuck???? 

It makes no sense…they lost!  And…how the hell did Kilmarnock DROP in the table?

First, remember this kind of model is attempting to combine actual results (banking points earned) with how a team has really performed on the pitch as measured by Expected Goals.  That xG differential is projected for the rest of the season…fed into the model & out comes the results.

When it comes to Kilmarnock, their defence has been very good, but the team is still struggling to create high volumes of quality chances.  Against Celtic, their xG was only 0.47.  They were lifted by a wonder strike by Chris Burke on a shot he will fail to score on 24 out of 25 times.  That’s not sustainable.  Given that their xG totals are middle of the pack, it’s very clear that they need to find ways to generate more in attack if they want to be a true threat to the top three (something I think they can do).

Now…onto Celtic.  What the hell is happening here?

Let’s start by looking at Celtic’s projected points when you isolate goals for & allowed, along with Expected Goals & Expected Goals against so far this season.

Screen Shot 2018-09-27 at 7.43.44 PM.png

Ding…ding…ding. Hello, outlier!

Before we address that, while some observers place the blame on Celtic’s defence, clearly that’s not the real problem. 

Now, the defence is not perfect. 

A big part of the reason they still are good at limiting chances is simply from the fact they still dominate in possession.  Celtic have the league’s highest possession rate at 68%, the next best team is Hibs with 56%. 

It’s easy to defend, when you always have the ball. 

But the facts also are that Celtic are one of five teams who have only allowed two goals from open play &  they have the lowest Expected Goals Against in the league which suggests their defensive record may be pretty sustainable.

The real outlier here is that Celtic simply isn’t finishing the chances they are creating. Look at the difference in projected points when looking at only actual goals compared to when you look at Expected Goals. That’s extreme….

Quick fact:  Celtic have the same amount of open play goals as Hamilton this season.  Celtic’s xG for the season so far is 12.18; Hamilton’s is 3.67.

giphy.gif

How is that possible? Well, Celtic has a Conversion Rate of 5.6% this season.  That’s bad…that’s almost Dundee bad (3.9%). 

Meanwhile, Rangers have a Conversion Rate of 19.4%.  So, if you’re wondering why Rangers still rank behind Celtic in the projections….it’s because Celtic isn’t going to score on only 5.6% of their shots for the rest of the season….and Rangers aren’t going to score on 19% of there’s.

The league average Conversion Rate is 10.3%, which is consistent with the normal rate over the course of a season.

Celtic’s Expected Conversion Rate (xCR) is 13.5%, while Rangers’ is 10.3% which backs up the claim that both teams should expect a change in their scoring rates. 

Furthermore, if you apply Projected Goals to both teams, Celtic would average 1.79 goals per match, while Rangers would average 1.26.  Over the course of a full season, that’s a difference of 20 goals. 

For Rangers supporters, the hope is that internally the team is very aware of all this & that they are looking for ways to improve the volume of quality chances they create to soften the blow of inevitable regression.

For Celtic, the team needs to be analyzing WHY their finishing has been so horrible.

 Meanwhile, someone is averaging 0.86 goals per 90 in France (photo courtesy of Reuters)

Meanwhile, someone is averaging 0.86 goals per 90 in France (photo courtesy of Reuters)

While Moussa Dembele gets off to a pretty damn impressive start in France, it’s the loss of another Celtic superstar which, to me, has had the biggest impact on the club’s attack.

Even though most observers tabbed last season as a real disappointment for Stuart Armstrong, the underlying numbers suggest he was an invaluable part of Celtic’s success….despite seeing his goals dry up.

Armstrong became the case study of what regression looks like in football.  Two seasons ago, he averaged 0.62 goals per 90 riding a Shooting Percentage (Sh%) of 47%.  Last season? The goals dried up, as the reality hit that anyone scoring more then 40% of their shots on target…probably isn’t going to do that again in their next season.  For the record, in 2017-18 Armstrong averaged 0.18 goals per 90 off of a Shooting Percentage of 14%.

 LOOK BEYOND THE GOALS!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

LOOK BEYOND THE GOALS!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

So yeah, the narrative was that Armstrong had a down year & again the narrative doesn’t match with the actual facts. 

The midfielder’s impact on the creation of shots for Celtic was extreme & when he wasn’t on the pitch…the results were quite noticeable.

Now, I don’t normally put much weight in stats like, “the team lost four straight matches when <random player> was in the starting eleven”, but this case is different. 

Why? 

You’ll soon see that there was no other player in the league that drove shot creation like Armstrong did last season.

But first…

Screen Shot 2018-09-26 at 8.53.43 PM.png

You can see that in the matches that Armstrong played at least 60 minutes in last season, Celtic scored nearly a full goal more than in the matches he missed completely. 

I’d argue that Celtic’s woeful goal output this season, beyond Conversion Rates, is also largely influenced by the fact that Celtic never replaced Armstrong after he was sold to Southampton this summer.

If you’re curious, here are the Projected Points for Celtic for a season - solely looking at those three different goals scored averages:

Now, in order to prove that this shit isn’t just a random occurrence, it’s important to follow up these kind of numbers with supporting facts. I’ve been stressing how uniquely dominant Armstrong was in the build-up to chances & here are the numbers to show that dominance.

Screen Shot 2018-09-26 at 9.28.22 PM.png

Notice that in nearly every playmaking statistical category:

  1. Stuart Armstrong was the best in the league...usually by a lot.
  2. For the most part, the next best guy is already in the Celtic squad.
Screen Shot 2018-09-26 at 11.15.34 PM.png

Celtic’s stubborn resistance to giving Scott Allan a chance deserves a whole other blog post, but there clearly is a playmaking void in Celtic’s midfield & the fact that the team thinks Youssouf Mulumbu is the answer is laughable.

Mulumbu is a very good midfielder….don’t get me wrong…but he averaged 0.10 Expected Assists last season, Allan averaged 0.47. Mulumbu had 0.25 Scoring Chance Key Passes per 90, Allan averaged FOUR TIMES that amount with 1.00 per 90.

 courtesy of SNS

courtesy of SNS

Beyond the Armstrong Effect, there’s a few more issues Celtic is dealing with. 

The club invested a lot of money in bringing back Odsonne Edouard & with that price tag comes the expectation that he repeats last season’s goal scoring form…which was really, really good. 

Edouard averaged a league high 0.76 goals per 90, scoring two less non-penalty goals than Kyle Lafferty…in 1,663 less minutes.  Edouard’s Goals Above Average rate of +0.44 was the third best rate in the past five years of Scottish Premiership football (Leigh Griffiths owns the top two spots).

So far, this season his goal scoring rate is 0.36 per 90, which is a Goals Above Average of +0.03 (ironically, the same rate as Lafferty last season).  A big part of that is the fact that Edouard’s not getting many chances (three shots so far this season….see, there’s the Armstrong Effect) & the bigger concern is that he is due for a general decline in scoring this season.  His Projected Goals per 90 last season was 0.51, which is still very, very good —but not 0.76 goals per 90 good.

But you know what? If he averages 0.51 goals per 90 this season, Celtic will be pretty happy.  But he needs service (only 9% of Edouard’s shots were unassisted last season…for context, 18% of Dembele’s shots were self-made).

5e5a47f75a1e70a26c617767bf6d056a.jpg

That’s where James Forrest steps in…right?  I mean, he was third in the league in Key Passes per 90 last season.

The problem is…he’s not stepping in.  He currently ranks 21st in Key Passes per 90.  21st!

Additionally, when you look at Forrest’s MF Passing Rating, which measures how good a player is at creating high volumes of chances that are consistently better than the league average chance, he currently has a -0.57 rating…which means he keeps setting up chances that score at a lower rate than the league average. 

For the record, the player with the lowest MF Passing Rating?  Leigh Griffiths with -2.09.  He leads the league in Key Passes, but most of the shots he creates are low percentage chances.

Ultimately, the numbers suggest that Celtic is primed to start climbing the league table but there are still some issues they will have to correct.  Losing Armstrong, while seeing two of their best playmakers (not named Scott Allan) falling into a trend of setting up low percentage shots isn’t going to lead to a quick fix. 

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 Recommendations.  Additional stats are courtesy of InStat Football.

  • I’ve come under fire from Rangers supporters in recent weeks because the numbers don’t match the current narrative.  I agree the team looks so much better than anything we’ve seen in nearly ten years.  But, the fact is - when a team scores on 20% of their shots…they are going to look like world-beaters.  The reality is that scoring rate isn’t sustainable.  The hope is that the team is very aware of that & will ride those scoring rates as far as they can, while honing in on tactical adjustments that will help the team increase its volume of quality chances to counter the inevitable regression.  It’s the job of team analysts to have those hard conversations with management.  When things are going so well, they need to be highlighting where the team may be vulnerable moving forward. 

  • A few notes on stats:  Shot Accuracy is the rate of shots that are on target.  Shooting Percentage is the rate of shots on target that are goals.  For a tutorial on the Shot Creation Passing numbers mentioned in that Stuart Armstrong/Scott Allan infographic - check out this write-up.

  • This was written under the influence of Sudakistan, John & Yoko’s Double Fantasy, the Jesus & Mary Chain & ZZ Top.