So if someone neutral would've watched last night's game, what result would that person have found to be fair based on the chances that were created for each team in that particular game?
xG under Ralf
- Thread starter Pogue Mahone
- Start date
Well its maybe semantics but they are made up mate, xG isn't an observable black and white event like a goal or an assist that can be interpreted differently by different people.
As you say yourself there are competing models run by different people and some that even use AI which is why we end up with different xG for the same team/player depending on the source.
The principle behind xG seems sound enough but I think there is still far too much subjectivity involved in the process. With the rate computing power is increasing and how efficient data analytics is becoming then no doubt as you say eventually one model will emerge with perhaps a very high level of accuracy but I don't think we're quite there yet.
To be honest I know I'm just being grumpy whenever I hear about xG. I think my biggest gripe with xG is the fear that it will become too accurate and subtract from the unpredictability of football.
It's literally a statistic that helps analyze the game. I don't understand how anyone can be upset at someone using the tools at their disposal to make a better analysis. Stats are used to make better decisions. When you're dealing with hundreds of millions, it's important to make the best and most informed decision possible. Stats can literally be used in every area of life to point out issues, gaps, strengths and weaknesses, just as they can in football.
Anyway, go yell to the sky about how having more information is annoying you
I agree, more sensationalism from the media misdirecting people who don't really understand the game.
United controlled that match for the vast majority of the half. Brentford created two real, separate chances (one of their 'chances' led to another 'chance'). Both of these chances were a result of largely unforced mistakes by us. One was a bad giveaway by Diego Dalot and one was our usual Achilles heel of almost conceding from our own corners.
Again, it comes down to the narrative. Keown cluelessly (or maliciously, depending how you look at it) banging on about Brentford being the better side, or on top. Absolute garbage. As I just said on another thread, the media know that sensationalising everything relating to United gets engagements.
There's no story in saying "United dominated but gave up a couple of half-chances through individual errors". There's a massive story in saying "United awful again under Rangnick", "United dominated by newly-promoted Brentford" etc...
Because in it's truest sense it's a made up statistic!It's literally a statistic that helps analyze the game. I don't understand how anyone can be upset at someone using the tools at their disposal to make a better analysis. Stats are used to make better decisions. When you're dealing with hundreds of millions, it's important to make the best and most informed decision possible. Stats can literally be used in every area of life to point out issues, gaps, strengths and weaknesses, just as they can in football.
Anyway, go yell to the sky about how having more information is annoying you
It's like me saying I thought that Brentford should have won 3-2. Great, I'm basing it off a number of things, is that now a statistic that can be used after every game?
If someone can point out to me why Greenwoods chance was 55% (or even 66%), and give me examples of similar chances that I'm assuming the model is based on where a striker misses 1 in 3, then I'm yet to be convinced
Might as well throw xA under here, as well, since it is connected.
Top 5 creative players under Ragnick (PL only):
- Fernandes - 1.64
- Fred - 1.57
- Rashford - 0.89
- Telles - 0.86
- Sancho - 0.63
I think this is pretty telling that our forwards are lacking creativity.
Top 5 creative players under Ragnick (PL only):
- Fernandes - 1.64
- Fred - 1.57
- Rashford - 0.89
- Telles - 0.86
- Sancho - 0.63
I think this is pretty telling that our forwards are lacking creativity.
Yes, good point. Have been thinking about this myself. It's not uncommon for teams to create multiple chances in one and the same attacking sequence.
You'd think that the more sophisticated models are able to take this into account, though. If not, it's an obvious weakness since this sort of thing (finish - block/parry - 2nd finish is extremely common) is too widespread for it to be statistically insignificant, or so I would think at least.
Every single football club uses stats. XG is a great stat when used properly and in conjunction with the other stats. No stat should really ever be used in isolation, although I'm not sure why that constantly needs to be repeated.
As much as people like to think the eye test is best, it really isn't. Memory is unreliable and each person's inherent biases and emotional state can affect how they perceive what's happening around them. I don't think there is any way the average person can have an accurate in depth opinion of how a squad performed over a season or even half a season based on memory alone.
Last edited:
The same people who say xG is useless will also shout 'we had 20 shots to their 4'. Every stat needs context and understanding. 60% possession is meaningless if it's just Maguire passing to Lindelof and back again. Stats are a tool to be used but only alongside an intuitive understanding of the game.
Which is exactly my point, over a long term the pre-shot xG is probably accurate when read as intended and generalised. That doesn't mean Brentford actually had an X% percentage of turning those into goals. That isn't how statistics work or get applied.
Now if the statement is Brentford had easier opportunities to score goals then we did then yes that would be true. However, their lack of quality meant this wasn't realised.
It's also true that we can't afford to offer up opportunities like that against better teams or we'll suffer.
The idea that Brentford were unlucky or would have usually won that game based on xG however is nonsense and only someone that doesn't understand statistics would make such an argument. Anyone making such a statement needs an xG model made up of purely Brentford player data.
Does xG take into account whether the striker is right or left footed? If Rashford had the exact same chance on his left as he scored on his right yesterday it would be far less likely to end up in the back of the net. If not, that’s a big/obvious flaw in xG if it’s intended to be consistent.
They weren't unlucky to a sense, they fecked up the finishing on their chances/de Gea who is a united player performed really well. Lucky or unlucky can be said from a lot of different perspectives though. We were lucky that they fecked up their chances, as they were chances that De Gea shouldn't have had a chance to make a save. You can point to either one, luck or just below average finishing ability, but the chance is still a good chance.
I do think it's fair to say they normally would have won that game based on the chances, though that's my opinion. Football is a high variance game, but considering the chances they had, looking at those in a vacuum and assuming nothing else was affected, you would expect more goals on average from their chances than from ours I'd say. There's not enough data on Brentford to be able to read into that too much, and year on year xG over/underperformance isn't something that is consistent or sticky anyway. There's nothing to say a hot finishing team will keep it up next year or next game, or that the opposite will occur.
So I don't think it's all that relevant to say the value of these chances was less because they fell to a Brentford player vs a Man United player (as I said, the actual finishing difference between Lewandowski/Ronaldo/Callum Wilson/Chris Wood is essentially negligible). There are of course intangibles like how those chances are dispersed over time, do the great players show a higher rate of clinical finishing in important situations and miss easy chances when it's not as important while lesser players don't have that boost during important moments? Hard to say, as I doubt there's any data on that, but as an overall body of work, xG essentially says that when you're in the position to take the shot, over time any average premier league attacker won't necessarily be any worse at scoring them than even the best. Some players are just xG over/under performers, but it doesn't generally mean they are good or bad players. What can be sure, is that good scorers accumulate high rates of xG/90 over time. I doubt there is much correlation to good scorers over time with significant overperformance compared to xG itself.
Last edited:
Hmm not sure on that. I don't think it says anything on if the player is on his strong/weak foot, as xG generally is just trying to quantify the quality of the chance, not the ability of the player striking the ball. It takes into account the location of the shooter, body part of the shot, type of pass incoming to the player, type of attack (rebound vs slow attack on a set defence vs counter attack vs a shot after a dribble, pressure on the shooter, bodies in front of the ball, etc). How that player chooses to strike it, or the ability of that player and whether they are 2 footed or not I don't think comes into play. It's still a chance worth "0.x" xG, that was either executed well or executed poorly, and then the next layer where you'd look at strong vs weak foot would be to explain why that player scored or missed.
I'm sure there is something to be found in what type of finishes are certain players good at, and the ability of teams to get those players into those types of situations/ability of teams to prevent players from those situations but if they allow chances, allowing them to players who are less likely to take them/in situations they are less comfortable than average in, but that's a whole other world of research that Im sure some clubs try to plan for internally but won't be seen elsewhere
The value in xG is seeing over time how good the team is at creating chances/keeping them out... How the players execute them isn't exactly related to team play, that's just down to the individual to execute the shot. Or for an individual you can see how often they find themselves in good chances, as the highest scorers are regularly the players with the highest xG, and not the players who overperform xG the most.
Their chances were from two players who have never scored a premier league goal.
Sorenson has 2 goals in 3 seasons and Jensen has 3 in 3 seasons. All in the championship.
So no they wouldn't normally put those away. You can't replace player records with xG like that.
It's a great chance irrespective of the player taking the chance though. The chance isn't lessened because the player is worse. That's just down to the player missing a chance that the average player from their database would score at whatever percentage of time. Expecting a player to score it... Well I dunno, there's a lot that goes into that, but it's also hard to draw conclusions from a player finishing ability if they pretty much just never get in that position in the first place. Non existent sample size, so you start with the average as a base marker. Besides, if there's not much historical difference between the actual xG overperformance over time between someone like Robert Lewandowski and someone like Josh King or even Tom Cleverley with his also small sample size... I'm not sure it's fair to say you wouldn't expect a player to score what is an easy chance. I get your point that these players never score so it lessens the quality of the chance... But IMO, the quality of a chance is essentially just measured by xG, it has nothing to do with the player quality at the end of a chance, or the execution of the chance. It's a mixture of team play to create that chance, and the players movement to find himself in that chance. How a player carries out the taking of that chance is not the quality of the chance in the first place, but just the execution of the chance. Maybe that's just how people interpret the terminology differently though.
Do they take only the ball position at the moment of the shot for the xG calculation? If that's the case, than it's a major flaw of xG.
Our chances might have looked smaller because the average number of shots from similar positions result in less goals, but when you factor the position of the defenders and keeper, the weight and height of the ball and the body position of the striker, they were definitely much bigger chances, than xG calculated.
A shot from exactly the same position could be an almost 100% chance or almost 0% chance depending on how many defenders you have on your back, where the keeper is and the speed and height of the ball and weather you are well positioned for a shot or unbalanced and barely reaching for it.
Our chances might have looked smaller because the average number of shots from similar positions result in less goals, but when you factor the position of the defenders and keeper, the weight and height of the ball and the body position of the striker, they were definitely much bigger chances, than xG calculated.
A shot from exactly the same position could be an almost 100% chance or almost 0% chance depending on how many defenders you have on your back, where the keeper is and the speed and height of the ball and weather you are well positioned for a shot or unbalanced and barely reaching for it.
Most of them but especially the better xG models that most people pay attention to take into account the positioning of defenders and goalkeepers, the pressure a player is under, the type of pass they received, the height of the ball and the type of shot attempted (header/volley/ground shot/first time shot/shot after a touch etc).
https://fbref.com/en/expected-goals...mply, xG (or expected,events leading up to it.
Donny is an afterthought, Matic has no legs and Pogba is simply too erratic to be relied upon deep. Technical ability isn’t just about being on the ball and being able to pick a pass. It’s also about positioning and understanding which the best midfielders use to control matches every bit as much as the former. None of our midfielders have that in their locker.
Think it's too early to read into xG, xA right now. Many confounding factors
- Focus on control / pressing
- Trying different formations, different players in roles, trying to squeeze all our attackers in at the same time.
- Players themselves saying (read: leaking) that they want more instructions on what to do in attack.
- 10 sec rule, style change etc.
Some (most? all?) providers adjust for that. That's why a team's xG at the end of a match will often be different to the individual xG of all its players combined.
Say those three individual Brentford attempts in the same move had an xG of 0.60, 0.65 and 0.70. Combined that would be an xG of 1.95, which obviously doesn't make sense as you can't actually score more than one goal from that entire attack and a single goal obviously wasn't certain from that attack anyway.
So instead you flip it and look at the the probability that the defending team doesn't concede from any of those individual chances. So 0.40 × 0.35 × 0.30 = 0.042 or 4.2% chance the team doesn't concede from any of the three chances. And you then use that to get the opponent's xG for the entire attack. So a 4.2% chance of not conceding from the entire attack is a 95.8% chance of the opponent scoring from the attack, or an xG of 0.958.
As usual, some models take it into account, some don't. If you are interested in potentially the best model right now:
https://statsbomb.com/2020/03/a-beginners-guide-to-analyzing-players-using-stats/
https://fbref.com/en/expected-goals-model-explained/
I am sure, there is even more stuff to be found on the page, more recent. They seem to do a good job trying to be transparent.
But it's logical that xg should increase for each half chance in the same sequence. Shooting at goal 5 x 0.2 should have a higher xg (likelihood of getting a goal) than shooting 1 x 0.2 irrespective of sequence. Maybe there is a weighing factor there but surely both situations can't have the same xg.
Last edited:
Yup. If the Greenwood chance falls to Sterling, 0.65 or whatever is very generous.
Ah ok. That's good to know. Do we know which model Sky use? Is it Opta's own?
Tweet
— Twitter API (@user) date
xG matching up with the eye test today.
The final chance is doing a lot of lifting. 0-0 was very fair until then, but a late chance is equally valid.
Absolutely, and Ronnie's header earlier. Apart from that, neither team had a clear chance.
Not really. We were far better than them. Dominated possession and had 18 shots at goal to their 6. You could definitely argue we didn’t do enough in the final third to deserve all 3 points but we were comfortably the better team.
Must he around 0.99 from aRashfords tap in. Besides that one and Fred shot All the rest must have come from 0.05 xG chances
Fred’s shot well saved. Varane’s header just over. The left footed chances messed up by Bruno and Martial. We had a few reasonable chances.
See below.
0.27 against a quality team like West Ham is pretty impressive defensively.
Obviously the last chance probably account for around 0.97 of our own xG.
Obviously the last chance probably account for around 0.97 of our own xG.
We didn't create a plethora of chances, but apart from the winner, we also had the Ronaldo sitter in the first half.
So a 0-0 was fair after all?
We were the slightly better side, yes, and today was definitely more about the result than the performance against a tough rival.
Fred had the best chance but shot too high and close to the keeper instead of going across goal. Well clear into goal too.
If we're relying on Fred's finishing to win the game we're doing it wrong. He hit the target which is something I think only he and Rashford could say!
Key take away today is the defensive improvement. Been a while since De Gea had that little to do. Hopefully it's the start of a pattern rather than a one-off.