2nd January 2020
With the FAWSL starting back after its winter break this weekend it felt a good time to check in with the numbers and see what’s been going on. I did initially want to present this in a better way, but I thought I’d just give a quick rundown of things that have stood out to me in the first half of the season or look interesting to follow in the second. Because of this and not wanting it to go on for too long, I’ve only really taken a quick look at the title race and a couple of the strugglers.
As always, thanks to StatsBomb for the free data.
While the men’s Premier League looks set to have a runaway winner, the WSL looks to be a good three-horse race. They do play each other fairly early in the new year, which may mean the race doesn’t extend right until the end of the season, but at the minute there still feels like a good amount of uncertainty regarding who will be lifting the trophy come May.
Arsenal went into the break in first spot. A feat that was made even sweeter as Chelsea dropped points away to Liverpool, meaning they can’t overtake the Gunners even if they win their game in hand. City are ahead of Chelsea with a game in hand but have let themselves down in the two games against their rivals.
From the first half of the season, it feels as though Chelsea will have the most reason to be frustrated with themselves. They have the best xGD (per game) in the division and took maximum points from their rivals only to draw against strugglers Brighton and Liverpool.
Chelsea’s strong numbers mostly stem from an incredible defensive record. Their 0.57 xG against per game is the best in the division and also skewed due to a massive outlier against West Ham – a game where they conceded 2.0 xG. Removing that game (which seems silly) their xG against has been just 0.4 per game, which still includes the games against Arsenal and Manchester City.
Chelsea’s strength to stop the opposition progressing the ball is a big reason why they concede so few chances. They regularly top the charts when it comes to pass completion for progressive passes by the opposition and see the lowest percentage of opposition possessions that start in the middle third reach the final third.
Chelsea have the best xG numbers in the division, are difficult to play against and have Sam Kerr joining in January. It all makes it feel hard to look past them, but there’s still hope out there for Arsenal and City fans.
Chelsea have had their two matches against their rivals on home soil. In the second half of the season, they have to travel to both Arsenal and City. After the strong results in the first half of the season, you can argue the ball is still in their court but their dropped points have levelled things up a bit, making it so (though still early) it’s not in Chelsea’s hands.
On the flip side of Chelsea, City will welcome both Arsenal and Chelsea to Manchester in 2020, possibly giving them a bit of an edge. The bad news for City is just how poor they were in these big games in the first half of the season. They lost both and offered little in attack. They only created 0.2 xG against Arsenal and 0.4 xG against Chelsea. It might be okay if they were tight games which City just happened to be on the receiving end of, but Arsenal created 1.7 xG against them and Chelsea 2.4.
The good news for City is how well they’ve dispatched the rest of the league. Looking only at games excluding the top three, City have the best xGD in the division. Their defence, conceding 0.58 xG per game, is bested only by Chelsea while their attacking xG is almost identical to Arsenal’s – who currently lead the way. These numbers shouldn’t come as a huge shock, they’ve won every game against the rest of the league without even conceding a goal, but they’re still hugely impressive.
Whether or not City can go on to win the title may well come down to the two home games they have against their rivals.
Champions Arsenal know that the rest of the season is in their hands, but will be hoping their small squad can cope when competing on four fronts. On the surface Arsenal’s first half of the season has been great, they’re top of the league and in the quarter-final of both the Champions League and Conti Cup.
Their xGD per game is behind Chelsea’s, with their xG against taking a bit of a jump from last year, but it may not be a huge worry for the Gunners.
Last season they had the best xG against numbers in the division with 0.57 per game, but this season has seen that leap to 0.71 per game. However, looking a bit into the numbers, it doesn’t look to be a huge concern. Conceding a penalty rebound from Yana Daniels late on against Bristol City skewed the numbers slightly. Removing that chance they’re conceding around 0.62 xG per game, which isn’t significantly more than last season. It is worth pointing out though, the penalty came through some sloppiness at the back, with Daniels intercepting a pass from the ‘keeper before being brought down.
Like with the xG against numbers, it’s not enough to worry about just yet but the last three games have all featured some sloppiness at the back leading to a good chance. The penalty against Bristol City late on was the first example, then a poor pass by Daniëlle van de Donk led to a Fara Williams 1v1 against Reading and a misjudgement by Leah Williamson led to Chloe Kelly’s goal for Everton in their last match. They were comfortably leading in two of these instances but it’s something they’ll be hoping doesn’t resurface in the New Year.
Like in the first half of the season, Arsenal’s crunch games are back to back in 2020. They face Chelsea at home on January 19th and City away on February 2nd. If Arsenal can come out of those two games still in first place, it may well be back to back titles for the North London club.
After expected goals, the next big thing in football analytics looks like it could be possession value models. There’s already plenty to read and watch about the different models. Some things that I’ve seen and would recommend are below:
Thanks to the socceraction package by @TomDecroos it’s easy to run through the notebooks and apply both VAEP and xT models on the free StatsBomb data. A post was published about the two models in the package here. I’ve used both but I was getting some funny values with the VAEP model, so I’ve opted to just use the xT model for the time being.
You’re better off reading the post by Karun (or watching the presentation from the StatsBomb conference) to better understand xT (I don’t want to butcher the definition of it here), but the basis of it seems to be looking at how actions increase a team’s chance of scoring. In the ‘Vizualisng xT’ section of the original xT post, you can see the xT value in different areas of the pitch, which is explained as the percentage of times a team will score in the next five actions when the ball is in that zone. So, applying it on a player level, we should be able to find players who move the ball into dangerous areas and increase their side’s chance of scoring.
Selfishly, my first port of call when getting player values was to see how they compared to the progression figures I’ve been doing since the summer. My progression method is simple and just looks at the distance players moved the possession towards goal, but I thought it’d be interesting to see how they differ.
(The dark plots happened because how I was originally going to post these included a dark background and it all blended in but I’m too lazy to change it for here)
For the most part, it seems my progression method overrates defensive players and underrates attacking players. This isn’t too surprising as my lists of most progressive players are almost always dominated by defensive players. It looks at distance gained and they have the most distance to play into. xT is better in that it looks at increasing a team’s chance of scoring rather than just moving the ball forward.
The standout on the graph is Manchester City’s Janine Beckie. Beckie started the season in an attacking position but, since the injury to Aoife Mannion, has been playing as a very attacking right-back in recent weeks. She’s supplied the width down the right for City and has been a strong attacking asset for them. She’s also been the most unlucky player in front of goal in the WSL this season too. She has a non-penalty xG of 3.7 – the 4th highest in the league – but is yet to score.
The plot below, showing her most valuable actions, shows how she’s offered width on the right and how most of her high-value actions are crosses.
This may also account for some of the differences with attackers – I took crosses out of my progression numbers a while back. It was probably silly, but when I did the method I was looking more for players who advance the ball with passes as opposed to crosses. The plot below shows a comparison of the values with crosses removed from the xT values.
The defensive bias of my progression numbers is still on display with the trio of Vivianne Miedema, Leah Galton and Remi Allen being the standouts for showing up much better with xT. Removing crosses sees Ji So-yun move into top spot for xT per 90 too, though it’s a lot closer with Kim Little and a few others not too far behind.
I’m going to leave this here for now, while there’s plenty more to talk about I just wanted to give a quick intro to xT in the WSL and I’ll use it in pieces from now on, rather than giving a big overview here. To sign off though, here are the top ten players for xT per 90 this season (more than 300 minutes played).
|Name||Team||xT per 90|
|Janine Beckie||Man City||0.54|
|Leah Galton||Man United||0.42|
|Tessa Wullaert||Man City||0.39|
Finishing in 4th place and being the only side to have a positive xGD outside the top three, Birmingham City had a strong season in 2018/19. Halfway through the 2019/20 season, however, and things aren’t looking so great. Some kind of drop was almost expected, Birmingham lost a huge number of players in the summer, but can they look to recover some of last season’s form in 2020?
Looking at their expected goals, it’s not positive for Birmingham. Their xGD per game of -0.71 is the 4th worst in the division. They’re not in the position they are because of some kind of bad luck or finishing slump. Removing fixtures against the top three makes for better viewing, but it’s still not great. Their xGD is -0.16, which is the 8th best in the division.
Defensively they’ve not been too bad. They have the 7th best xG against in the division, which includes games against the top three where they’ve lost to Manchester City and Chelsea with a combined score of 9-0. Removing games against the top three, they have the 5th best xG against in the division.
One area they suffer is xG from shots taken within 15 seconds of the opposition having possession. They’ve conceded the 2nd highest xG per game from these shots. However, looking closer at the chances, it’s mostly from the games against Manchester City and Chelsea. Against Manchester City there was an instance where they lost the ball in the City half, leading to Keira Walsh playing Janine Beckie in on goal and also the below counter-attack leading to a good chance for Lauren Hemp.
Then the same happened against Chelsea. In one instance Millie Bright grabbed herself a goal after breaking forward and in the second half Bright won the ball back in midfield and went on to set up a decent chance for Beth England.
In the games against City and Chelsea, it felt as though it was easy for the opposition to move through the midfield. With Birmingham wanting to get their full-backs wide and in advanced positions it can leave the middle a bit vacant. Despite this, their numbers for the completion rate of progressive passes is around average – while still including the games against City and Chelsea. Removing those two games sees the completion rate move to slightly better than average.
There may be some problems with their structure/pressing which was taken advantage of in the games against the top clubs, but the fact it isn’t manifesting itself in games against everyone else does it make it seem like it isn’t that worrisome for the Blues.
The main concern is in their attacking numbers. Only Bristol City have created less xG against sides not in the top three. The problem isn’t quantity, they take the 6th most shots per game, but the quality of their shots has been poor. They have the lowest xG per shot in the division and it’s no surprise when you see their shot map.
It’s not as though they’re taking long shots from just outside the area either. There’s plenty of shots that look to be a good 30+ yards out. Looking at shots that come from 30 or more yards out, Birmingham have the highest total despite playing the fewest games in the league. They’re taking 4.25 shots per game from 30+ yards out. It’s more than the top four combined.
I should probably dig deeper into looking at why they’re taking long shots. For instance, it may be a result of a lack of options in attack. It does feel almost pointless to point out a team takes bad shots without looking at why that’s the case or how they can improve. With that being said, there’s also been plenty of times when they’ve taken long shots with bodies forward and options available. It’s also not as though it’s one rogue player, all of Lucy Staniforth, Claudia Walker, Brianna Visalli and Chloe Arthur have more than one shot per game and none of them have an xG per shot above 0.05.
Despite the negatives when it comes to shot selection, there’s a lot to like about what this Birmingham side does with the ball. They look to play from the back and put together a lot of nice possessions. Birmingham average the 4th most passes per possession for possessions that begin in their defensive third and have the 4th most 10+ pass possessions per game in the division. While they don’t get a clear chance, the below clip is a nice move against Everton going from their ‘keeper to the opposition box.
A highlight of their season so far has been the form of 22-year-old Sarah Mayling. Last season I mentioned she had strong progressive passing numbers with good accuracy figures and even with the changed method she’s continued to impress. For players with more than 300 minutes played she has the 6th highest progression via passes per 90. More impressively, she’s leading the league when looking at progression per minute in possession, by quite some distance (as can be seen on the plot below).
A part of me is unsure if I did this the best way, I took the duration of passes (excluding crosses and set pieces) and divided their progression from passes by this figure. I constantly feel like I missed something obvious and I shouldn’t use this number, but her per 90 numbers are also strong. She’s the defender with the 3rd highest xT per 90 behind only Janine Beckie and Lisa Evans – who have both also played in more advanced positions. She has her club’s highest xT per 90 and even their 2nd highest xG + open play xA per 90 behind Lucy Staniforth.
Her strong xT and xG + xA numbers likely come thanks to a couple of good deliveries into the box, like the one below against Liverpool.
And a nice 1-2 before a low cross against Everton on the opening day.
Overall, Birmingham feel like quite an odd team in 2019/20. A lot of their numbers aren’t good but they don’t look like a bad side. They build well from the back, press high up the pitch with the 3rd highest number of pressures in the final third per game (@EveryTeam_Mark posted a cool graph looking at counterpressing in the WSL too), but a lot of that work is undone by poor shot selection in the final third. If they can improve in the final third there’s still time in the season for them to try and become more a part of the middle pack, as opposed to one of the strugglers.
It’d be wrong to talk about this season’s WSL without mentioning Vivianne Miedema. The 23-year-old Dutch forward looks set to top her record-breaking 2018/19 (though there are two more games this season). In 20 games last season she scored 22 goals and assisted another 10. This season, after playing 9 games, she’s scored 14 and assisted 7. She’s outperformed her xG numbers in both seasons but her underlying numbers are still hugely impressive. The plot below shows how no one has come close to her expected goal contribution so far this season.
Given her goal record and height, it’d be easy to assume she’s just a goal poacher hanging around the box and being a focal point for crosses, but it isn’t the case. She’s spoken about having a mindset more like a #10 than a #9 and the creativity numbers back that up. But to have those creativity numbers without any kind of drop off to her goalscoring is what makes it so incredible. Last season her xG was 0.76 per 90 compared to 0.77 per 90 this season, but her xA has climbed from 0.18 to 0.51 per 90.
The Bristol City game, where she assisted 4 goals, was when her creativity was most on show. She had one assist thanks to a right-footed cross from the right and two left-footed crosses from the left before a nice pass into the box to find Lisa Evans running in from the right.
I do sometimes worry that when I talk about the WSL I talk too much about Arsenal, but it’s because they’ve quickly become one of my favourite teams to watch and Miedema is a big part of that. It’s scary to think she’s still only 23-years-old and already seems such a complete player.
With Manchester City’s start to the season, many of their players have impressive numbers. After an injury in pre-season Ellen White has hit the ground running, Tessa Wullaert has added some xG contribution to her impressive progression numbers, teenager Lauren Hemp has impressed and Caroline Weir has been quite progressive in possession and hugely active when out of possession for the Citizens. Behind all of these, however, Keira Walsh seems to be somewhat quietly having a great season.
The 22-year-old had some strong progressive passing numbers last season, looking like the best midfielder in the league for that category, and this season has seen her continue that trend. The plot below shows her progression via passes per 90 and her xT from passes per 90 compared with midfielders and forwards.
Like Miedema, she’s also added more creativity this season. Walsh has great passing ability but wasn’t one to make the final ball in 2018/19. Her open play xA per 90 was 0.11 but this season it’s jumped up to 0.34 per 90, 2nd in the league behind only Vivianne Miedema.
While City have lined up a bit differently in recent weeks, with a 4-4-2 kind of shape featuring an attacking Janine Beckie at right-back, Jill Scott tucked in on the right and one of Tessa Wullaert or Georgia Stanway just behind Ellen White, Walsh’s role in the system is largely the same. Besides her assist for Weir in the game against Chelsea, where she pinged a nice ball into the box, most high-value chances she’s created have been a result of a great ball from deep sending a player in on goal.
After handing in a transfer request and attracting interest from Lyon in the summer, City will be glad Walsh decided to stay in Manchester as she seems to be cementing her place as the best deep-lying midfielder in the division.
There are still two sides in the WSL who are yet to win, bottom-placed Bristol City and 2nd from bottom Liverpool.
The real question seems to be whether or not Liverpool can improve to make sure there’s not an actual relegation battle. Bristol City have the worst xGD in the division but the big difference in the two sides comes when looking at the xT numbers.
For xT Difference there’s a bigger gap between Bristol City and next lowest Brighton than there is between Brighton and 4th placed Reading. @footballfactman mentioned on the opening weekend that if Bristol City aren’t careful they could become the new Yeovil, but Liverpool’s poor start to the season could give Bristol some hope they can stay up.
Liverpool have the 3rd worst xGD in the division and, like Birmingham, have struggled to create good goalscoring chances. @mostlyinane wrote about some of their struggles, particularly in creating good chances, on StatBomb. What makes it more frustrating for Liverpool is they possess some talented individuals and are capable of being better, with their performance in the Merseyside derby being a minor positive of the season so far.
They have an experienced striker in Courtney Sweetman-Kirk while the younger trio of Rinsola Babajide, Melissa Lawley and Niamh Charles all put forward strong numbers for carrying the ball. Watching Liverpool it can sometimes feel like they’re too individualistic when they attack. The three names just mentioned have the biggest average carry distance for all players in the league.
Looking at carries that come from counter-attacking situations Liverpool have the greatest average carry distance and worst xG per shot, suggesting their counters consist more of one player on a lone run taking a low quality shot more than a fast-paced team move to hurt a team in transition. This is a huge assumption but it’s looked true from a lot of what I’ve seen from Liverpool.
The pace and dribbling ability of players like Lawley, Charles and Babajide could help make Liverpool a dangerous counter-attacking side, but so far it hasn’t been the case.
Fortunately for Liverpool, Bristol City haven’t offered much this season. @EveryTeam_Mark tweeted that they have the highest proportion of passes from their ‘keeper and it’s not surprising based on what I’ve seen of them. It feels like they play out from the back, before passing around the defence for a few moments before sending a hopeful ball long.
It seems unlikely that they’ll be able to turn their form around, but one positive from the first half of the season has been 18-year-old Ebony Salmon. Excluding players from the top three for a moment, Salmon has the 5th best xG + xA per 90 and 11th best xT per 90.
It’s hard to see anything other than Bristol City finishing bottom, but if Liverpool can’t improve going forward it may not be so clear-cut.