Allsvenskan 2016 summary pt. 1

With the season ending more than a week ago, I finally have enough time to sit down and write a summary. I’ll split it in parts, with the first two looking at team and player performance respectively, and hopefully I’ll get around to writing a third part in which I look at how my predictions have done, and the model’s performance on the betting markets.

These kind of updates likely won’t return for next season when I’ll be taking on a new job compiling odds on Swedish football. I’m hoping to continue writing in some form though.

But enough about that, let’s get to it:

00

As most expected, Malmö bounced back from last season’s 5th place to reclaim the title from Norrköping. Luckily they did so without the need of the extra win awarded to them after the abandonded game against Göteborg where the home fans threw pyrotechnics against the Malmö players – and Tobias Sana responded with a spear throw.

At the other end of the table, Gefle were finally relegated after several years of clinging to their place in the top flight. Falkenberg’s extremely poor season saw them relegated as well, while Helsingborg will have to face third placed Superettan side Halmstad in a two-leg relegation play-off.

Take a look at Djurgården’s row of results by the way, only one draw!

01

Malmö were the best side in terms of shots taken and conceded as well, while bottom duo Gefle and Falkenberg really struggled together with Sundsvall, whose good start to the season saw them able to avoid the relegation battle despite only picking up two wins after the summer break. Örebro was an outlier throughout the season, usually producing some high-shooting games.

02

Göteborg were the most efficient attacking side during the season, but their low shot volume saw them unable to compete with the real top sides. AIK and Malmö relied on pure shot volume instead, probably a result of their ability to dominate games. Falkenberg on the other hand really struggled with both volume and effectiveness, usually needing nearly 11 shots to score.

03

At the other end of the pitch we see partly why Malmö were the superior side this season, and why AIK finally overtook Norrköping in second place: they both enjoyed some very efficient defending, clearly outperforming their opponents. Falkenberg struggled here as well, conceding a goal about every 5th shot, while Örebro actually did well efficiency-wise despite conceding a lot of shots.

04

Champions Malmö were the best side in terms of both xG and xG conceded, while Falkenberg’s poor defence was a big factor in their relegation. Djurgården, Hammarby and Östersund did better defensively than their final positions in the table might suggest, unable to break into the top mostly because of their weaker attacking output.

05

Ranking the teams by Expected Goals Difference did well to explain both ends of the table, getting the top 3 and bottom 2 correct. As mentioned, Sundsvall’s ‘lucky’ results at the start of the season saw them avoid the relegation battle, while Östersund, Hammarby and Djurgården formed an underperfoming trio just below the top sides.

06

Simulating every game based on the shots taken and their xG values, we can give the teams ‘Expected Points’. This is very close to the xGD rankings above but we can see some differences, like Falkenberg ‘earning’ almost as many Expected Points as Gefle, which they were no way near in reality.

Let’s see how the teams actually did compared to their Expected Points:

07

What we see is that typically the winning sides overperform against Expected Points, while the losing sides underperform. This is to be expected as you’ll very rarely (or never) dominate a game enough for your expected points to match the three actual points awarded for a win. The same goes for losing, since you’ll pretty much always ‘earn’ more than zero Expected Points.

08

There’s always exceptions to the rule though, and this season Gefle stands out as having picked up pretty much exactly the points expected from them which I would say is rare for a losing side, while Falkenberg look to have been very unlucky to pick up so few points.

09

Looking at the Expected Points distributions for the teams, we really see just how ‘unlucky’ Falkenberg have been. As mentioned above, losing sides will very often underperform against Expected Points but Falkenberg really stands out with a 100% chance of picking up at least the 10 points they ended up with, implied by the 10,000 seasons I ran through my simulation.

Djurgården

As I’ve done in a few updates, I’ll end this one looking at Djurgården. As a lifelong supporter I’ve become used to the ups and downs but despite that and the underlying good numbers I was still a bit concerned this season.

10

Luckily though, Mark Dempsey, the right man at the right time, stepped in and turned things around much like his former mentor Per-Mathias Høgmo did in 2013. Just like I’ve seen him do in Norway, Dempsey focused on a very direct attack which worked well to improve shot numbers and level out Djurgården’s dropping xGD, while at the same time crucially also getting some real results.

Defence continued to struggle though, and no real tactics but ‘get the ball up to the big boys up front’ was clearly visible – a decent game plan to get them out of the hole they’ve dug themselves into but not something to build on for the future so the club’s decision to not give Dempsey a new contract looks reasonable.

That’s it for now, in a few days I’ll be looking closer at individual player performance.

Advertisements
Allsvenskan 2016 summary pt. 1

Allsvenskan 2016 – The Endgame

Before I continue with another Allsvenskan 2016 update – the last before the season ends – I have some news regarding the blog.

As some of you may know, I’ve been working part time for StrataBet this season, mostly writing game previews for the Norwegian Tippeligaen. As I soon take on a new, full-time job elsewhere I likely won’t have the time to write as much as I want. Also, with my new job focusing on Allsvenskan and Swedish football in general, I may be reluctant to give away too much information to the general public, so the future of this blog is very uncertain.

I’m hoping to continue writing in some form though, and what I do write will likely be closely linked to StrataBet as they’ve given me access to their great dataset.

Allsvenskan 2016 – The Endgame

Ok, so let’s get on with another update. With only 3 rounds left – the next starts tonight – we can see how much of the drama has gone out of the league table since last time. Malmö have retaken the top spot and thanks to Norrköping’s recent poor form the gap down to the title contenders is now 4 points. Sure, both Norrköping and AIK can still theoretically win the title, but I would be very surprised if Malmö let this slip out of their hands, despite the disappointing defeat to Östersund. They do have some disturbing injury problems though…

01

Göteborg have a chance to break into the top-3 and gain a European spot for next season, but this looks even more unlikely with 7 points up to AIK. At the other end of the table the bottom-3 have looked locked in for a long time. Helsingborg still have a chance to overtake Sundsvall, but again I’d be very surprised if this happens. In mid-table we see how Elfsborg, Kalmar and Hammarby have climbed a few spots at the expense of Örebro, Häcken and Östersund.

02

Counting up shots we see how Djurgården surprisingly is the best defensive side when it comes to denying the opposition chances to shoot. We also see how Gefle continue to be very bad and that Örebro still is the main outlier with A LOT of shots both taken and conceded.

03

Looking at effectiveness up front we see few changes since last time. Elfsborg have been slightly more effective with their shooting though, partly explaining their climb in the table. On the other end of the scale, Helsingborg have had a real problem scoring on their chances lately, with ZERO goals since the last update.

04

Looking at defensive effectiveness we see why Djurgården’s ability to deny the opposition chances hasn’t seen them climb into the upper half of the table: They still concede a lot of goals on the chances they do allow. Only bottom-of-the-table Falkenberg are worse. With Malmö and Norrköping’s effectiveness declining since last time, AIK now stands out as the far superior defensive side.

05

Not much have changed in terms of chance quality either – but what is interesting here is that Djurgården is the best defensive side when it come to xG as well. So if they concede very few chances, and very little xG – why are they conceding all those goals? My guess is – I don’t have time to look it up – that the few chances they do concede are of higher quality. Djurgården have also had a lot of problems with goalkeepers this season. Having used 4 keepers so far, only star signing Andreas Isaksson has looked stable enough but he has picked up an injury and will be out for the remainder of the season.

I don’t know much about evaluating goalkeepers but have been thinking about doing a blog post about it for some time now, hopefully I’ll get to it in the near future.

06

Looking at Expected Goals Difference, we see how Djurgården’s lack of defensive effectiveness has robbed them of a nice upper half finish. My model currently ranks them as 5th in the league, close to Hammarby in 4th – far above their current 11th place.

We also see how AIK have overtaken Norrköping in 2nd place, and with the reigning champions in poor form and just 3 points above AIK, this is where most of the drama left in the season lies. At the bottom of the table, Helsingborg are actually ranked far better than Sundsvall above them, but the 7 point gap will likely be too much for Henrik Larsson’s men with only 3 games remaining.

07

The model has always liked Malmö and they actually have the chance to secure the title tonight, if Norrköping lose away to Elfsborg while Malmö win away to Falkenberg – a not too unlikely outcome. In the race for 2nd place, AIK now have the upper hand much thanks to Norrköping’s recent poor results. Göteborg seems to have all but locked in the 4th place and the same goes for the bottom 3.

To continue my slight focus on Djurgården in this post, they’re interestingly projected to take about 6 points from their 3 remaining games: Helsingborg away, Häcken at home and Sundsvall away. Given their very disappointing season, and as a cynical Djurgården supporter, I doubt this.

Allsvenskan 2016 – The Endgame

Allsvenskan round 23 update

It’s been over six weeks since my last Allsvenskan update but now I finally have time to get to it. Six rounds have been played since last time and a lot has happened. Let’s take a look at the league table:

00

Compared to last time, we can immediately see that reigning champions Norrköping have climbed up above Malmö to claim the top spot, which is very impressing given the players who have left the club, and the mid-season managerial change.

At the other end of the table, Djurgården have (luckily for me) picked up pace under new manager Dempsey and moved up from 14th to 11th, while Helsingborg and Sundsvall have struggled – only picking up 2 points each.

Let’s have a closer look on how the teams have performed:01

Despite giving up the first place in the table to Norrköping, Malmö have distanced themselves from the rest in terms of shot dominance. Not much else has changed, Örebro are still involved in some very open games while Gefle struggle to create chances.

02

Örebro and Elfsborg have moved into the ‘constant threat’ quadrant thanks to some effective scoring, while Hammarby have done the opposite. Kalmar have improved their effectiveness, but at the same time seen a drop in shots taken per game.

03

Here we see how AIK’s and Norrköping’s improvements come mainly from their defensive work; both sides have been better at keeping shots from going in since the last update. Kalmar’s defensive effectiveness has improved as well.

04

05

Expected goals for and against look much like they did last time but AIK’s defensive improvements have seen them close in on the top 2 sides, as they’ve increased their xGD by nearly 0.20 per game.

How about a prediction then?06

Malmö’s defeat to Djurgården has really opened up the title race, but my model still fancy them. Norrköping have improved though, and we could be in for a very interesting finish to the season. AIK have improved as well, and have seemingly all but locked in a top-3 spot. In the other end of the table Falkenberg have plummeted from around 22 expected points to less than 16, with the model giving them no chance of reaching the relegation play-off spot occupied by Helsingborg.

Djurgården under Mark Dempsey

As mentioned earlier, as a Djurgården supporter I’m very happy with how the form has improved under new manager Dempsey. In the last update I showed the long-term trends leading up to Olsson’s sacking, and now that Dempsey’s been in charge for 7 games we can see how he’s managed to turn things around:

07

While shots conceded actually declined during Olsson’s last season, so did shots taken. What we see under Dempsey’s rule is clear: everything have improved! Djurgården now concede less and take more shots but more importantly, both actual goal difference and xG difference has improved, leading to more points and a climb in the league table.

Though a bit of hindsight, through my work with Norwegian football I was optimistic about Dempsey coming in as I knew he would provide the energy needed for a turnaround. Let’s hope Djurgården can continue to pick up points to climb further.

Passing spiders

Another thing I mentioned in the last update was how Opta data is now available for Allsvenskan, and I showed some passing maps heavily inspired by 11tegen11 and David Sumpter. I’ve since then played around with the script to create passing map animations, which received a lot of positive feedback on twitter and have now been dubbed ‘passing spiders’, often a quite fitting name.

I don’t know enough about tactics to determine if these animations holds some analytical value, but they are fun to look at and could possibly be used to provide an interesting narrative of individual games combined with other types of analysis. I got a lot of good advice on improvements on the animation and will implement some of it in the future.

That’s it for now!

Allsvenskan round 23 update

Allsvenskan round 17 update & Opta data!

Ok, it’s time for another Allsvenskan update. With 5 rounds having been played since last time we should be able to see some changes.

00

Looking at the table we see how Djurgården, Örebro and Sundsvall all have dropped a few places while Östersund and Häcken are the big winners. As a result of Djurgården’s poor performance, Pelle Olsson has been sacked and replaced by Mark Dempsey.

01

As I noted last time, the league seems to have settled when it comes to shots. Indeed, no team has changed quadrant since the last update.

02

Looking at attacking effectiveness we see how Gefle have become more clinical in front of goal while Djurgården have become more ineffective, partly explaining their struggles.

03

Only Elfsborg have changed defensive quadrant since last time, dropping from ‘competent but busy’ to ‘pushovers’. Hammarby and Sundsvall have also dropped a bit while Malmö have improved their defence.

04

05

Malmö and Norrköping are still at the top of the xG-table, but the big surprise is Östersund’s rise to third place – mostly due to an improvement in their attacking output. Despite their recent struggles, Djurgården still sits in sixth place. Helsingborg have climbed up to 13th, leaving Gefle at the bottom.

06

07

08

Looking at Expected Points, we can see just how bad Djurgården have performed recently. They are ranked sixth by xPoints but sit at the bottom of the xPoints Performance table, about 9 points below expectation. In my simulations, they reached at least their current total of 18 points about 98% of the time – indicating a massive underperformance.

09

My game simulation model still consider Malmö heavy favourites for the title. I certainly agree but 93% is much too high considering they’re only 1 point ahead of Norrköping at the moment.

Djurgården managerial change, and Opta data

As a Djurgården supporter I’ve welcomed Pelle Olsson’s sacking, as DIF have been very poor under him this season. Shot dominance has decreased since last season, but more importantly, the actual goal difference and Expected Goals Difference have plummeted. The club’s situation look alarmingly similar to when Per-Mathias Høgmo came in to save us from relegation in 2013. Hopefully Høgmo’s former assistant coach can repeat that feat this autumn.

pelle_01

More news is that Opta data is now available for Allsvenskan. I probably won’t have time to dig too deep into it at the moment, but I’ve written a script for plotting passing networks, heavily influenced by 11tegen11 and David Sumpter.

Using these plots, we can compare Pelle Olsson’s last game using a 4-4-2 formation (Opta has it down as a weird 4-2-2-2 though) against Dempsey’s first game in charge where he used the same formation.1517

Sure, it’s only one game – but we can see some distinct differences here as Dempsey used a midfield diamond with Kevin Walker pushing up while Alexander Faltsetas dropped down deeper. Olsson has always favoured two holding central midfielders. Also, Dempsey has gone for a more straight forward approach to attacking, with more direct passes up towards the strikers, while Olsson used more crossing.

Allsvenskan round 17 update & Opta data!

Norrköping vs. Djurgården

Though I had planned not to share any shot maps before I had discussed the concept of Expected Goals and my model properly first, after watching last night’s game between Norrköping and Djurgården I just couldn’t help myself. Now some of you probably have seen this kind of plot before so I will leave out the explanation for later posts.

Peking-DIF_2015Here’s last night’s high-scoring game. Being a lifelong Djurgården supporter this was not a pleasant game to watch, and to add to the pain I actually had a bet on the under here. Sigh.

While I don’t believe much in year to year trends such as Team A vs. Team B always produces a lot of goals in today’s modern football where players and managers change teams frequently, watching the game I got a vague feeling of déjà vu and just had to look up this fixture from the last few years.

Peking-DIF_2014Peking-DIF_2013Looking at the games from seasons 2014 and 2013, it seemed there is some truth to the myth. But while it may look like this particular fixture usually end up a high scoring affair, in the other seasons in my database (2011 and 2012), the games ended 2-1 and 1-1 respectively. Furthermore, Djurgården actually only had two players starting in all three games: Kenneth Høie and Emil Bergström. The same goes for Norrköping with only David Mitov Nilsson and Andreas Johansson starting all three games.

With so few players playing all three games and the games therefore being played under completely different preconditions, I think we can safely put this high-scoring trend down to pure coincidence. I still feel like a fool for betting the under though.

Norrköping vs. Djurgården