Season Review: When did the Projected League Table get it right for your team?

fEg3yAZ.jpg

written by:  @TheGersReport

I first introduced my variation of a weekly Projected League Table model back in the first week of September.  It was only based on a few matches worth of data, but I was eager to roll out what I planned on being a weekly feature of the blog. 

Keep in mind, there was a disclaimer that the projections would fluctuate as more matches were played.

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).
— Modern Fitba (September 8, 2018)

Spoiler:  Hearts didn’t win the league & Rangers didn’t finish ninth.  We all knew that those initial projections wouldn’t last. 

However, even my very first version of the Projected League Table after three matches got some of the predictions eerily/nearly correct. 

The table on the right has the projections after just three matches.

The table on the right has the projections after just three matches.

Celtic was tabbed to end the season with 94 points (a week later the projection dipped to 89…they ended with 87), it had Kilmarnock with 66 points (one point off their final tally) & Livingston with the same 44 points they finished the season with (despite the fact that their current form at the time was on pace for 67 points).

It also got a bunch wrong…because you know…it was based on three matches.  But, now that we have some hindsight…I figured that it would be fun to see when in the calendar the projections got it right. 

But before we do, I want to highlight the fact that using this kind of model goes beyond just coming up with a magic number for how many points a team will earn.  It can measure whether a team’s performance on the pitch is matching up with the results.  Look at the three teams mentioned above.  The real form of both Celtic & Kilmarnock mirrored the projected results.  Based on the actual points earned in those first few weeks Celtic was on pace for 86 points & Killie was on pace for 67. 

Livingston was on pace for the same point total of 67, but the underlying numbers driving the Projected League Table suggested that 44 points was a more realistic point total (& the actual amount they ended up with). 

courtesy of Kilmarnock Football Club

courtesy of Kilmarnock Football Club

On September 28th, I published the updated projections that had Celtic leading the way for the first time. 

Everyone knew that Hearts wasn’t going to win the league…but here’s the context of where the Scottish football world was back at the end of September. 

Celtic was coming off a 2-1 loss to Kilmarnock & now trailed Hearts by five points after only six matches. 

With 10 points from six games, Brendan Rodgers’ side are enduring their worst start to a league season for 20 years
— Jamie Lyall, BBC Scotland (September 2018)

Hearts were on pace for a 101 point season (haha…wow!) & Celtic was on pace for only 63!

Simultaneously, I published my weekly projections that said Celtic would finish top & that projection never changed for the rest of the season.

Of course, everyone & their mother knew that Celtic would ultimately turn things around & win the league. But, while publishing these projections each week I was able to hone in on what was causing teams to not see their actual results match up with the performance on the pitch.

The same model I applied to Expected Goals could also be applicable to other numbers like the ones below from the end of September. Here were Celtic’s projected points when you isolate goals for & allowed, along with Expected Goals & Expected Goals against so far this season.

Screen Shot 2019-05-19 at 8.26.21 PM.png

The clear driving force behind Celtic’s poor form at the time simply came down to the fact they were not finishing the chances they were generating (& they were creating pretty good chances). At the time I wrote…

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%)...Celtic isn’t going to score on only 5.6% of their shots for the rest of the season...The league average Conversion Rate is 10.3%, which is consistent with the normal rate over the course of a season.
— Modern Fitba (September 2018)

For the record, Celtic ended up scoring on 10.9% of their shots this season.

Also, from a Hearts perspective.  They were coming off a rather innocuous 0-0 draw with Livingston.  Nobody really thought Hearts were going to win the league…but again, this is also the moment when the projections flagged the beginning of Hearts decline in the table…SEVEN WEEKS before Celtic actually took over the top spot in the league table. 

giphy.gif

Ok…..after all that - let’s take a look at when the Projected League Table got it right for your team.

Note: I took a two month hiatus doing some consulting for a club in England so….there is a real gap in the “weekly” part of this whole process.

I already mentioned how early projections got it right for Celtic, Killie & Livingston.  The estimated points fluctuated as each team’s form & results progressed throughout the season.  However, in October the projections had Celtic finishing with 86 points, in November it was 87 & a few months later (remember I was off the grid from November to February) in March the projection was 88.  Celtic finished with 87 points this season.

In October, the Projected League Table had Kilmarnock finishing with 64-65 points & then the projections returned to 66 points in April (they finished with 67).  For much of the season, the projections were closer to 61 points & it wasn’t until April 6th that they were tabbed to finish with 66 points after a 2-0 victory over St. Johnstone off of an own goal & a penalty.

It’s interesting to take note of the team currently projected to be in third place. In “real life” Kilmarnock sit seventh in the league table with seven points.
— Modern Fitba (September 21, 2018)

For Livingston, the projections stuck on 44 points for the first few weeks of the season & then was in the 50s for the next few months. Once I returned to action in March…they were in the 40s. At the end of the calendar year, their actual form of 30 points in 21 matches projected out to 54 points. Whatever happened in the first couple of months of 2019 seems to have triggered Livi’s eventual decline to finishing with 44 points.

courtesy of Dundee Football Club

courtesy of Dundee Football Club

This next tier of clubs didn’t see their actual point totals projected until the second month of the season.

After five matches, St. Johnstone had earned eight points which translates to a pace of earning 61 points by the end of the season. They finished 2018-19 with 52. Also, after five matches the Projected League Table had St. Johnstone with…52 points.

After five matches, Motherwell was on pace for a relegation-ready 30 points. At the same time, the Projected League Table said there really was no concern for dropping down to the Championship, given they were tabbed to finish with 52 points. Six months later, the projections said they would finish with 50 points. They ended up with 51.

Screen Shot 2019-05-19 at 10.44.19 PM.png

Oh…Dundee.  I had such high hopes for you…but instead you signed Kenny Miller until the year 2020.  The numbers were never on Dundee’s side after that.  By the end of September, the projections had Dundee finishing with 23 points & they “showed us” by finishing with 21.

courtesy of St. Mirren Football Club

courtesy of St. Mirren Football Club

From the outset, the Projected League Table was anchored by the fact that St. Mirren & Hamilton would join Dundee in the bottom three.  The order changed from week-to-week, usually with Hamilton at the very bottom.  It wasn’t until late October that the projections got Hamilton’s point total right.  They actually earned a 1-1 draw with Kilmarnock to get their seventh point in ten matches.  That translates to a 27 point pace for the season.  The projections were slightly more favorable suggesting Hamilton could finish with 31 points (they ended up with 33).  At the same time, the projections had St. Mirren ending the season with 30 points, two points less then their eventual total.

198483173.jpg.gallery.jpg

So far…the model has been pretty impressive. But one of the clubs it took the longest to figure out was my beloved Rangers Football Club. When I left for my little hiatus in the winter, Steven Gerrard’s men were projected to finish with 65 points…nowhere near the final total of 78. It should also be noted that their actual form at the time of averaging 2.0 points per game translates to 76 points.

When I returned & published these projections again in February they had Rangers with 76 points. The underlying numbers finally caught up to the reality (Rangers still averaged 2.0 points per game at that point as well).

Earlier I showed you how you could use this projection model to parse out expected points based on actual goals for & against & Expected Goals & Against. Going into a very important summer, I wondered what could be learned when applying this approach to Rangers.

Screen Shot 2019-05-19 at 11.28.16 PM.png

There is a real sense that the club can/will compete for the league title next season. Finding a replacement for Alfredo Morelos will be a top priority, but the numbers also suggest that improving the defence should also be on the top of the club’s shopping list. Rangers defensive play was markedly improved this season, but their xG against numbers highlight a need to be even better. The signing of Glen Kamara helps but you wonder if finding an upgrade to Nikola Katić as a partner to Connor Goldson should be considered?

SNS-18363215-Hibernian-v-Aberdeen.jpg

It also wasn’t until my return in February that the projections got close for Aberdeen & Hibs. After 24 matches, Aberdeen was tabbed for 68 points (they ended up with 67) & Hibs were set to earn 52 points (they finished with 54).

In the case of Aberdeen the projections constantly undersold where Aberdeen eventually finished. That makes me wonder if Derek McInnes made some deliberate tactical shifts in the winter that changed the trajectory of the club this season. For Hibs, the numbers oversold them & suggested that they should have done much better this season. Their projected point totals were consistently in the 60s for much of the fall & at the end of October their actual form was also on pace to finish with 68 points. The downfall of Neil Lennon at the club really derailed Hibs’ season & internally there probably is a lot of hope for a bounce back season next year.

Then there’s Hearts…

When you look back at their season…you really have to ask yourself — “What the actual fuck?”

Below you can see where the projections had them week-by-week in correlation to their actual point totals. Basically for the latter, I simply took their points per game at that point & multiplied it out by 38 (again remember there are a couple of gaps in the calendar).

Screen Shot 2019-05-20 at 10.02.51 AM.png

It’s safe to say that Hearts management knew all along that they did not have an 80-90 point team…despite the incredible run of results they had at the outset of the season. However, you can see that from match-days 7-10…Hearts actual form took a real dip but then spiked back up to it’s final peak of the season. They were coming off a 3-0 victory over Dundee on October 23rd, were averaging 2.5 points per match, & had a six point lead over Celtic.

At the same time, there was a 26 point difference between the Projected League Table & the extrapolated point totals from their current points per game ratio. This was the moment in which some serious red flags should have been going off inside the offices of Hearts management.

Compare Hearts form in the Projected League Table to their actual form from weeks 7-10. Their actual form had Hearts spiking up again, but in reality the performances on the pitch were showing a team in steady decline. Notice the trajectory of that decline was basically steady for the rest of the season.

courtesy of SNS

courtesy of SNS

Ultimately, I hope that I have shown how these Projected League Tables can function as an entry point to real analysis internally at clubs. The league table tells us one thing, the performance based projections tell us another. Sometimes the projections are eerily accurate & that’s pretty cool. However, their greatest use is being able to trigger moments in the season in which changes in tactics or squad management need to occur.

Some notes…

  • Stats are courtesy of data provided by Ortec Sports.

  • Sorry for the gap in the timeline on these.

  • The story behind this model can be found at the end of this introductory post.

  • Given that the model banked points that clubs earned, the accuracy was expected to build as the season went on & for the most part it did. That’s why there is real discrepancies with some of the initial projections.

  • This was written under the influence of Hamilton Leithauser, the Rolling Stones, & Yak.