Welcome to Expected Villa (xV), a column about the stats and metrics behind Villa’s recent results. This week, we talk about Villa’s paradoxical start to the season, the best (results-wise) since the 1960s.
A note about this column going forward
I think this season, I’d like to tie xV a little less toward individual match analysis and trend this column more toward bigger-picture analysis. That means xV will be more of a (somewhat irregular) weekly thing rather than a regular post-match issue.
This is mostly because of a combination of three factors:
a) I’m an American and weeknight matches are sometimes hard
Kick-off for these matches is at 2:45 or 3:00 in the afternoon here, which means the first half is still during the work day and the second half is mostly played during my drive home. This isn’t really conducive to close study of Villa’s matches, and I don’t think it’s necessarily responsible to act like I paid more attention than I did.
b) Life gets in the way sometimes
So a little related to a), but I generally try to keep myself busy, and sometimes that means I might not be around to write this column as often as I’d like to. Other times, non-work things may pull me away from being able to follow Villa closely (hello, Notre Dame football weekends). If I pivot this column a little bit more toward something that lets me have a more irregular schedule, it makes it easier to align it with my availability.
c) Analyzing individual match xG in particular may not be the most reliable or appropriate analysis
So, there’s a handful of things that make individual match xG sometimes a bad fit for analysis. I still think there’s some use for the metrics as long as you use them alongside strong analysis of the match, game states, etc., but they shouldn’t be proclaimed as gospel that one team played better than the other.
People smarter than me have written about this before (example 1 and example 2), but there are a few reasons for this:
- Raw xG totals sometimes don’t tell the whole story. Example: A team that takes two 0.4 xG chances will have a better chance to win than an opponent that takes eight 0.1 xG chances, even though both teams had 0.8 total xG. Yay probability!
- Single-match xG doesn’t have much predictive power moving forward.
- For any expected goals model, there is a huge amount of variance included in any figure, simply because the model can’t know everything that’s going on. Say a teammate’s whipped in a cross and the striker is going to try to fire a chance at net. How good is the cross? Is the defense closing the striker down or is he unmarked? Is the striker hitting the ball with his dominant or secondary foot? Where’s the goalkeeper positioned? All these items will affect the actual chance of scoring from the chance, but the model probably doesn’t know this. It may assign a 30% probability when everything’s picture perfect and the striker will actually score 60% of the time. Alternatively, maybe the striker had to stretch to meet the cross and will only score 5% of the time from the position. This is the natural variance in any one xG and a big part of why xG is most relevant over prolonged periods (where the effect of this “noise” is mitigated).
- The majority of xG models are shot-based, but this may not always be most appropriate. Which leads me to…
I love having the chance to talk about Conor Hourihane’s goal against Yeovil Town
This goal and Birkir Bjarnason’s to win the Wigan Athletic share a lot of the same qualities, but Hourihane’s goal is best suited to this point. Most any goal-based xG model is going to assign greater than 0.9 xG to Hourihane’s goal because it’s the perfect chance — a tap-in that’s slid across the goal-mouth from inside the area. There’s pretty much no better opportunity in football as far as xG is concerned.
Yet it almost didn’t happen (or at least happen that way) because Jonathan Kodjia’s ball across the area was deflected by a defender.
Is a shot-based xG model really best here? Probably not, because it doesn’t take into consideration the various outcomes that could’ve happened if the defender played the ball differently. Perhaps he’s able to put it behind for a corner, which mitigates the threat — in this case, a goal-based xG model wouldn’t register a chance. Perhaps he puts it into his own net, which again, most goal-based xG models wouldn’t register a chance here either.
This is why I’m intrigued by the concept of an action-based xG model. Effectively, it would say that Kodjia played the ball across the front of the goal. What’s the probability that a goal results from that action, not Hourihane’s shot? We think of the probability at the end point pretty often (e.g. what’s the chance this is scored?), where the probability that the end point is even created is often interesting in itself.
What should we make of Villa’s start?
When I was in high school, we discussed that famous F. Scott Fitzgerald quote in class: “The test of a first-rate intelligence is the ability to hold two opposed ideas in mind at the same time and still retain the ability to function.”
So, in light of that, in one idea: Aston Villa have won three consecutive matches to start a season for the first time in greater than 50 years despite not playing their best football. This is good.
The other idea: Aston Villa have struggled to beat two likely bottom-half Championship sides and a League Two team, putting in largely uninspiring performances each time out.
Like xG, I think it’ll take time to see which one is true. I think Steve Bruce and the players know their performances are going to have to get better, but it would be nice to see that happen sooner rather than later. Ipswich Town away shouldn’t be an issue at the weekend (seriously, they’re awful), but Brentford at home mid-week will be an issue if performances don’t improve.
Broadly, I think we can say that it’s nice that Villa are getting results without their best football, all the while acknowledging that the football will need to improve as the season goes on to meet the goal of promotion. I really hope this isn’t a controversial view.
What’s Villa doing well so far?
Well, the plus side is that Villa have only conceded one “normal” goal in three matches this season, which is neat — Hull City’s goal in the season opener was an incredibly fluky result for the Claret and Blues, while Mile Jedinak’s back-pass that led to Wigan Athletic’s opener Saturday wasn’t the result of a poor scheme, rather a simple error.
Most tangibly, from a statistical point of view, 52 percent of Villa’s opponents’ shots have come from outside the penalty area, the third-best mark in the league so far. This is something that Villa did at their best last season, and it’s something I hope they continue to do well this year.
They’ve also posted one of the better attacking threats so far this year, putting 11 shots on target across two league matches. They won’t continue to score at a clip that high, and they won’t continue to get goals from a slew of defensive-minded players like they have, but there’s some reason to believe that, despite poor performances, Villa have one of the better attacks in the division.
Speaking of the attack, I thought Kodjia played really well on Saturday and worked some good chances for himself. We know he can finish, and now I think it’s all about getting him back in the groove. Hopefully that happens sooner rather than later.
What’s Villa doing poorly so far?
Well, a lot of things, really. I’m not sure anyone (maybe aside from John McGinn) is quite playing to potential yet, which I suppose is positive. But performances have largely looked lethargic and Villa have left it relatively late and really late to win what should’ve been straightforward matches.
Steve Bruce’s selection has also been a little disappointing. I thought the team was overly defensive for Hull, then he played Jedinak at centre back for some reason against Wigan when he had other options. I don’t like overreacting to mistakes, but I don’t think Jedinak is a centre back (particularly when Tommy Elphick and Axel Tuanzebe were both available for selection). He’s a fine option to have in case of a couple injuries to guys ahead of him, but don’t overthink this. Just play guys in more comfortable positions, Steve.
Villa aren’t in a hole, which is nice
It’s really hard to win promotion when you’re starting from, like, 10 points behind in September. Villa aren’t doing that this season it looks, which is really helpful.
John McGinn looks really good
You should be able to get an assist on an own goal, because damnit, he deserved one on Villa’s second. The Scotsman looks like he’s got a great ball on him, which is going to be really helpful.
I also think Villa finally have the personnel to play to Scott Hogan’s strengths once the forward returns from injury.
Let’s look ahead
Villa have four matches left before the international break, and they should be targeting a minimum of eight points from the run. A trip to Ipswich Town and a visit from Reading on successive Saturdays really should produce wins — these were two of the three worst teams in the division last season if you’re an xG fan (which, if you read this, you probably are). A home match with Brentford sits between the two, and a visit to Sheffield United finishes off the run. I think I’d settle for a draw right now against Brentford, and Villa should be in with a good chance to win the other three matches. Silly things happen in football, though, so getting two wins out of those three wouldn’t be bad.
I am cautiously optimistic right now about Villa. They’ve still done a lot of nice things despite not having their best football, but it begs the question of whether or not the “best football” will show up.
Let’s get after it again Saturday, friends. UTV!