You are. We were hardly sparkling in the first half and we did struggle to create any real chances, but we largely controlled the half. Brentford did have the better chances via counter attacks and set pieces, but they were isolated incidents and not a genuine reflection of the pattern of play.
xG under Ralf
- Thread starter Pogue Mahone
- Start date
We have Donny, Matic and to a lesser extent Pogba who are all technical and they simply aren’t a good fit. I just hope we get someone strong enough to hold onto the ball under pressure in the defensive 1/3 and who can also tackle and push up. I really like Kessie on a free and Tchouameni from Monaco because both can go box to box like city and Liverpools players. What I’ve realised this season is we really need harder working forwards, for the league especially. Average defenders have our lads worked out far too often.
A classy but hard working 9 in his prime or approaching it for the wide lads to come in and play off and cross to and possibly a harder working direct winger too (maybe Pellistri or Amad next year) Then we can press with the best of them. It’s not control or creativity in the middle we are missing. It’s really punishing teams that just want to sit back and counter us more than anything. Actually making all of our attacking play and possession count for something.
Last edited:
Well thats a strange attitude. One could argue that some folks tend to feel the need to belittle something, they don't understand. How somebody can say it is madeup is beyond me to be honest. It certainly isn't the best thing since sliced bread and some people give too much weight on it in isolation, but that isn't the fault of the stat, but the person misusing it.
Agree with every word. That often makes it so frantic and difficult to discuss United these days. The players certainly have international class but sometimes their determination is way too little. You have to blend workrate and technique - inside an individual player and (probably even more importantly) within a squad. Happy to see this issue finally seeing the sunlight it deserves.
Not exactly how it works. For example... Ronaldo over the past 8 seasons has pretty much just as much actual goals as he has xG. Same with Lewandowski. It's just a case that on the day, their players missed their chances/didn't execute while ours did. They had much more chances than we did, and on another day could've won 4-1 or something. Or 3-2 like the xG there states. Football is a game of fine margins. Run the same game with the same quality chances back multiple times, with only the execution of the chances being changed and you'll probably end in a Brentford win more often than not, but on the day we won.
Using xG as a future predictive stat based off a small sample is useless, but you can still use it to help look at the game and come to the conclusion that yes, de Gea bailed us out/we got lucky that their attackers fecked up the finishes. They are both equally true. They had chances that De Gea shouldn't have had a chance in, but they gave him a chance and he produced a save out of them.
Absolute drivel. It's more most don't realise that a low scoring game like football will see wide swings of variance as most don't properly understand what it represents. They just assume seeing Brentford 2.73 - 1.98 Man Utd is nonsense because we won. xG represents what will happen over the long run if the chances in the game were recreated, hence in the long run it will generally be an accurate measure of what happens. It's like in poker if you get it in as a 2:1 favourite, the odds aren't all of a sudden wrong because you lost, you were just unlucky, over the long run the odds will play out.
Someone surely has to relook into the xG calculation algorithms. Its a wrong indicator and presenting a false picture on performance.
Its mainly driven by attempts and shots that most probably leads to goal. In today’s match, except 2 or 3 brentford attempts nothing were threatening the keeper or will most likely end in goal still their xG was high where we scored 3 beautiful goals and few half chances with one hit the post but ours was xG less than 2.
Its mainly driven by attempts and shots that most probably leads to goal. In today’s match, except 2 or 3 brentford attempts nothing were threatening the keeper or will most likely end in goal still their xG was high where we scored 3 beautiful goals and few half chances with one hit the post but ours was xG less than 2.
This xG stuff is complete bullshit. It ignores the quality of crosses and passes to the final player... Football is not about stats and will never be, people should stop trying to americanize the game. Just watch it and make your conclusions based on what you watched.
People who focus on these kinda stats have never stepped foot on the pitch..
People who focus on these kinda stats have never stepped foot on the pitch..
Spot on. And I don't think you'll get a retort of substance, because there really doesn't seem to be one.
it’s not supposed to indicate an accurate picture of performance
it’s supposed to indicate what it indicates, which is the average number of times a player scores from the positions of each shot taken in the game
Is it an average player or a Brentford player? Because that would seem pretty relevant
it's calculated from shots in the OPTA database, the sample size is in the 100s of thousands
Fun fact that I always bring up - over the past 8 seasons, both Lewandowski and Cristiano Ronaldo have pretty much just as much goals as they have xG. 223 goals from 220 xG for Ronaldo (101%), 226 goals from 223.9 xG for Lewandowski (101%). Or you take a relegation fodder striker like Chris Wood... 50 goals from 52.9 xG, 94.5%. Callum Wilson is pretty much 100% matching. The list goes on.
So basically, random Brentford player vs Ronaldo or Lewandowski, you're getting a tiny variance at best when you are in the position of taking the chance. The difference is getting the chances. But in this game they had that anyway.
Thats surprising. I wouldn't have expected that.
I thought they did well to create the chances but their goal attempts were quite weak. Ultimately routine saves for De Gea in current form. But if what your saying is right then then maybe its bias on my part dismissing the chances.
xG doesn’t account for who is in goal either. It’s a great indicator/measure, but it’s not worth anything if used in a single game. It’s a big data metric.
This Ramsey fella seems to be a complete fraud. We are worse at set pieces than we have ever been and its not like we are fielding a bunch on techincal midgets either
Expected goals is a metric that is designed to be used only over the course of long term, ie, a season...NOT INDIVIDUAL GAMES!!!
And yet we know for a fact that it's a better predictor of future outcome than something like actual goals scored. But maybe you guys know something top clubs' data science and analytics teams don't.
It depends on the model. xG Philosophy uses a basic model so isn't that great. FBref use Statsbomb which is a lot more sophisticated and accounts for defensive pressure and various other factors.
Their xG for the past two games has been
xG1.0 - 2.01xGA vs Villa
xG2.3 - 2.2xGA vs Brentford
These are the worst xGA numbers since Ralf took over. Worse than Ole's side was usually producing at the worst moments in his decline. We've been defensively poor in both games and given up a number of big chances. We especially got away with it against Villa. DDG has bailed us out but we've also benefited from bad finishing from our opponents. It's perhaps a warning of a false dawn for those that get too emotionally invested in the bottom line. When we play better opponents if we perform as badly as this in the defensive phase we could get serverely punished.
https://fbref.com/en/squads/19538871/Manchester-United-Stats
The below article highlights some of the advances in the Statsomb model.
https://statsbomb.com/2020/07/statsbomb-release-expected-goals-with-shot-impact-height/
As a neutral you seem to have really lost the art of scoring a basic team goal through thirds of the pitch consistantly against reasonable opposition.
Start of season you scored some nice team goals v Leeds but that was obviously illusion. Think Everton at home that Martial curled in was another example.
I've watched you for 180 minutes this week and really bar the goals we gifted you the main other chances you created were us parking the defence 40 yards up and you just ran into the vacant space e.g. Cavani chance in the cup and also second half when Greenwood hit it straight and Rashford didn't bother with rebound. Tonight Greenwood hit one just wide of the post in a similar move.
For all being better first 30 minutes don't recall too many shots compared to our good spell for most of second half.
Seems a real lack of co-ordination in final third considering some of the names you have up there. Midfield is poor but ball can still get up to the forwards but they can't seem to play effective 1-2s or passing triangles at all compared to Man. City doing it with ease every game.
Last few weeks you've created very little from open play v Wolves, Norwich and Newcastle and a few things here and there in the two games this week and none of these four opponents are even top 6 so Rangnick not solving much in his brief tenure bar launching top level career of Elanga.
A Tielemans type would certainly help for final third so that's a player you need to be seriously targeting in near future.
Long post just to say no idea how that relates to xG anyway as couldn't care less about that being used as football stats metric.
Fair assessment and one I actually share, we don't create a lot from open play, unfortunately we are shite from standards as well.
I said this in another thread but I'm going to repeat myself here. The team is very imbalanced and we essentially lack two types of players. Attacking players who have the instinct and desire to get involved in build up play. Too many of them, Ronaldo, Greenwood, Rashford, Cavani and Bruno all want to immediately get into advanced positions. Now I know we bought Sancho to somewhat fix that, but he hasn't really started well and he can't do it on his own. If you rely on one player to do a certain task it's easy for our opponents to just mark that player out of play.
Now there is another par of the problem we also have too few people who can carry the ball forward or play a progressive long pass. Which leads to these stupidly huge gaps between defense, defensive midfield and our forwards, you see in our play pretty much all the time.
So yeah someone like Tielemans in midfield would be great and we need a quality right FB, as well as a quality DM that can sit behind two pressing and progressive 8s. Of course those two 8s are another topic, Bruno has the instinct of a 9 1/2 it feels like, he really loves to play high up the pitch but we would need at least two players who get involved in pressing, winning the ball back and then fast advance the ball.
I feel a good team needs to have like max 2 striker types, maybe can get away with 3, meaning players who immediately look to get forward and hope for a good pass but the rest all need to get involved in building up play or else we are going to remain blunt in attack from open play.
That's not how xG works. They take loads of chances scored from similar positions on the field in similar circumstances and calculate the average of this chance getting scored. Since there are surely some players out there who even muck up chances like this or the GK simply being a monster and saving it there is never going to be a 1.00 xG for any chance I would assume.
Very interesting article by statsbomb I like the idea of integrating the z-axis into calculating the xG, seems reasonable to assume a ball high in the air got a lower chance of bing scored than one on the ground and if the statistical analysis backs this up it makes sense to integrate this into xG.
So is every source of xG stats identical?
It's exactly how it works. They feed in data from thousand of shots across a the full range of players. They don't balance the stat towards the best players only.
The fact better players are close to their xG is a demonstration that even they aren't consistently brilliant across a long season.
You can't use xG to say a shit player should have scored that because xG doesn't account for their quality.
I'd believe that might be the case. But then pulling random scorelines out of a hat might statistically be a better predictor as well, who knows.
Paul the octopus wasn't bad at predicting score lines either.
There's so much money in football these days that teams can't afford not to track xG along with hundreds of other types of data on the off chance it might give you an edge. Wouldn't be surprised if they had someone on the staff following Horoscopes either to be honest.
I made this point in the match day thread but Brentford had a lot of 2nd and 3rd attempts in the same sequence of play. So they may have had 3x 0.2xg adding upto 0.6xg in that incident. They can in fact only score one of those chances though.
I support xG. They are good stats to used as barometer to see quality of certain aspect. Higher xG indicates positive quality chances. If you have high quality xG but not enough goals then you need better finisher, if you have low quality xG but much more goals than your xG then it's not a good thing for long term or next matches. Thus, the team's creativity must improved (whether it changes the system, changes the midfield, or add more creative players).
The same can be applied on xGA or even more obvious because if the opponent have more clinical finisher than Brentford, we would likely lose to 2.73 xGA. We need to improve this as we keep conceded 1 or more xGA. Need a more consistent in controlling the game.
The same can be applied on xGA or even more obvious because if the opponent have more clinical finisher than Brentford, we would likely lose to 2.73 xGA. We need to improve this as we keep conceded 1 or more xGA. Need a more consistent in controlling the game.
Yup, their goal attempts were shit at the end of the day but de Gea still made very good saves to keep them out. But they were great chances that their forwards shouldn't have given de Gea a chance with in the first place. That's part of the variance in football, but it's still helpful to show that we did concede way too many big chances that nobody should be comfortable with.
That's an opinion that is very contradictory to the statistical models used yesterday. Brentford's xG (according to FotMob who I think use Opta) had Brentford's xG at 2.58. That's calculated at the moment of the shot.
Their post-shot xG, aka xGOT (xG on-target), was even higher though: 3.17. Essentially, Brentford made good use of their chances, but De Gea is credited with having saved 2.17 goals yesterday with his shot stopping. He had a very good game.
Players are close to their xG over time precisely because over time there isn't a massive variance in how clinical these players are. Some massively overperform xG regularly, but that doesn't really tie in with the best players really.
I feel like your last paragraph contradicts your first 2. You say that xG takes everyone into account and then great players over time are pretty much dead level with their xG, just as some mediocre forwards are right there with their xG. So from that perspective right there, the forward quality doesn't matter that much. A great chance is a great chance. It's not like the chances fell to their centerbacks or players who just don't score goals (and even still, they would be great chances still). Their chances fell to attackers, and they had over 2 xG in the game. You'd expect them to have scored 2 goals from those chances, whether that was Callum Wilson up top or Robert Lewandowski. Their career statistics and their own variance from xG shows this. This doesn't always happen, and a mixture of bad finishing on the day and great goalkeeping from de Gea on the day kept them to 1. But they still definitely had the chances that 2 goals would've been what you'd expect from their chances.
I can't take any seriousness from "Expected Goals" if you're telling me that Greenwoods goal was 0.55 last night.
You're telling me that on average 45% of players miss that shot from there with no goalkeeper.
Absolute rubbish.
You're telling me that on average 45% of players miss that shot from there with no goalkeeper.
Absolute rubbish.
Let's ignore models for a second. Which of thier chances would you expect De Gea or any top goalkeeper not to save?
The two big saves De Gea made were with his feet and he didn't really have to adjust them, it was poor finishing/good positioning. From the finishes you would expect him to save them both but both were good chances another player might have put into the corners.
Yeah this. Greenwood's goal was a 0.99 at minimum. And there was Bruno's chance late on that I felt was probably about a 0.55 (I see more of those goals scored than any of the chances Brentford had).
That said I'd agree with the xg statistics at half time, we were on 0.2 to Brentfords 2.something and that would have been fair.
You have to separate the chance before the shot was taken, and the chance after the shot was taken. XG is pre shot. It's everything about the situation, the positions of all players and goalkeeper, the height of the ball when the player is striker it, etc, just before the shot is taken. How the player executes the chance, is not what xG takes into account at all. That's at least my point on why their chances should have led to more success for them.
On the actual shots, I think the first big save de Gea made with his foot where Jensen went low and to the near post was a fantastic save. Most goalkeepers cheat the other way and can't react like that with their feet, while it's a de gea specialty to save with his feet. Even still, he covered the angles perfectly and gave himself a chance. The other one where their player was 1 v 1 with de gea was also a great foot save, though made easier for de Gea than the first. Then there were others they just missed the target with.
Understat has Bruno's at 0.41 which is fair imo. That's still a huge chance. People overestimate how frequently a chance is actually scored. They have Greenwoods at 0.64.
Also when you have videos like this one, pretty much no chance is going to be 1.00, precisely because they do get missed from time to time. Greenwoods was an easy tap in but let's not pretend like even easier tap ins haven't been missed, where the player is almost on the goal line and gets it over the net somehow. It happens quite a bit. For Greenwoods chance, he's sprinting and going at it first time mid sprint with a player on his back with a short reaction time to the pass, it could easily happen that he hits it wide or over. Still a massive chance, and I'd say 0.64 doesn't do it justice anyway, but yeah. Probably something likes 0.85 would do it justice IMO.
It's not binary. Shots aren't saveable or unsaveable. I think on average a goalkeeper concedes at least two or more goals based on the shots Brentford put on target yesterday.
Edit: rephrasing
Worth bearing in mind that feet saves are a speciality of De Gea's, something he utilises more than most goalkeepers and something we've become used to after a decade of him in goals. For average goalkeepers those sort of saves can be more difficult, because they often tend to try and save them with their hands or misjudge when they can/can't make the save with their feet.
No it isn't. There are some more simple ones (that are probably annotated via some sort of AI) which is the reason, why they available so fast after a match, and some more sophisticated ones. I guess, within the next 2-3 years, one model will emerge superior and will push all other models away and, with that, get rid of the discussion about differences between models.
I am not sure, if it is the best way to interpret the results of xG as scorelines. These are just sums. And they are the absolute opposite of made up: if you watch a game, you notice if somebody had a good chance or not. But maybe, because you only watch the matches of your team, your view is biased due to only knowing your teams players. There the statistical approach comes in and averages out every variance you have.
And even if I agree, that some models are too simple to make any kind of meaningful statement, the good thing is, the flaws are always consistent and therefor apply to everybody. That means that even if, for example, the understat model isn't the most perfect model, as it is applied to each time in the same way, comparing the sums over multiple matches and teams tells you a pretty good deal about how you stand in terms of chance creation and prevention.
There is no magic in that, stats are a manifestation of things you can see with your eyes. But you are not going to sit in front of your tv, stopping time while we are in possession or the number of passes we make in the final third etc.
I doubt it.
0.64 is considered high for just one chance since penalty is like 0.76. Man city have xG of 2.37 average per game this season, while Liverpool have xG of 2.61 average per game this season. Ours under Rangnick so far is 1.53 average per game. Considering that we haven’t play against the top teams, obviously you expect higher than 1.53. But it’s still in progress so hopefully we can improve it.