Another, somewhat related way to look at it, is that Arsenal are averaging 3.5 fewer shots per game this year than last season. They have scored 19 goals while averaging 12 shots per game, while City have needed 22 shots per game to score their 21 goals.
I don't believe it can, it's just looking at shots, not the players taking the shots. This does raise a possible flaw in xG. For their their entire careers Lacazette and PEA have finished 21% and 19% of their shots respectively. I'd be interested in seeing a comparison of their expected goals vs their goals scored in serveral of their most prolific seasons.
Both have been about even, iirc. 20% is typical output for an elite striker. What you guys are missing is: 1) Arsenal aren't creating enough chances, and 2) Arsenal are conceding about as many chances as they create. That's rather bad.
It's based on the league average striker. It is not a flaw, the whole point is to isolate the value of the shooting positions.
Yes - Arsenal are benefiting from having elite strikers. Exactly - the 25% shot conversion will cool off at some stage. And if there is a slump (e.g. some random variation turns up) and conversion were to drop to 10-15% for some time, then we will why these underlying numbers are not great.
True, but that aside, and stats aside, the most important thing is that we've garnered 18 pts out of 24 available. And that we're 2 pts off first place. I think we should all just be enjoying this position, taking "one match at a time" and just seeing how things work out against Leicester, Palace and then most interestingly Liverpool.
I am enjoying it! I am back to watching games, and having interesting discussions about actual elite strikers and holding midfielders. Ozil doesn't play but we can bring in a player like Mhki? JFC! The quality is so much higher. But having followed Kloppo at Liverpool, I am realistic that fixing the defence is a multi-year project.
Maybe, but not necessarily. XG is a model based on averages. For every average there is a distribution. It’s entirely possible (and expected) that some performers will ride certain bands of the distribution for a significant period (I.e. bc they are better/worse than average shooters). Hazard routinely has more goals than expected. Lacazette scored 25% more than expected his last season in France. Salah had 27% more last season. Messi often has seasons of 30% or more than expected. None of these XG model accounts give their error or distributions which is annoying. What is the likelihood that we have more goals than expected after 8 games? Is it 30% or 2%? XG also assumes every shot is an independent situation like a coin flip. They are not. Consider a striker gets the exact same chance three times in a game. The first two times he blasts over the bar. Think those first two shots will not affect the third? Also, people misuse xG and just blindly add the xG. Consider a stiker takes 0.5xg shot and it’s saved and a player gets off a 0.6xg shot from the rebound. Do we really expect 1.1 goals from that attack? Of course not, so why add up all the xG from each team and use that as some expected goal total for the game. All that said, our finishing is fantastic and I don’t expect streaks like the current one to continue all season, but purple patches are common among elite players and championships are won because of such patches.
Correct. This is also the reason some guys are working on/moving towards xG models with post-shot data (on-target or off target) included rather than the 'naive' xG models of the past. I'm a little bit of two minds about this. On the one hand, these models do pick up signal earlier (i.e. with smaller sample sizes) and correspond more closely with the results of individual games. On the other, I think it kind of loses a bit of the purpose of xG, which as Jitty said can be looked at as about isolating the value of shooting locations (and shot types from those locations).
As I said a page or two ago, I watched it on my phone from the hospital, so was a little distracted. We certainly didn't play all that well in the first half though.
Agree Especially models do different jobs. The point of xG in this case is not to prove how Arsenal won, or to prove that Auba is an elite striker - we already know that so we don't really want to weight his personal xG We want to have a benchmarked stat that can compare between the teams over time. On the other hand, I can see internally you might want to delve into post-shot data - e.g. to compare Arsenal against itself.
IIRC teams don't outperform xG in the long run. Many of them use multi-year data though - so the sample size for a team is actually quite big. e.g Klopps liverpool doesn't have 8 games of data - it has nearly 170 games so from that you can isolate significant trends - especially by using rolling averages Hot hand fallacy! I think I am yet to see any evidence that shooting streaks are a thing? A lot of people have looked at this to see if shots are independent.
But this is the problem with looking at xG data for a small sample size like 8 games. Currently, we are +8 goals to xG expected goals. But nearly half of that gap was based on yesterday's game. That is a a really weird outlier. Over the course of the season, our goals scored/conceded is likely to be fairly close to xG but in a small sample size, much less a single game, it isn't super helpful.
But that sample size is the point. Either xG For will have to increase significantly to reflect how good this team really is, OR conversion will revert because this team isn't good. One of these two things will definitely happen.
I wasn't familiar with post-shot xG, but I think it's less useful than xG in evaluating strikers. It more confirms that somebody hit a great shot or a poor shot, but that's the whole point of xG, which is that scoring from a given spot is dependent on a lot of factors. You could evaluate "finishing" by comparing xG pre and post shot to see who derives more goals from their shooting ability.
I think you boys should seriously think about starting a new thread dedicated to xG and other stats. p.s. I'm sure it must exist already somewhere in BigSoccer, but if you want to try to keep the discussion amongst Gooners, then start one in our forum.
You're not the goddamn topic police. You bitch about basketball talk and now you're bitching about PEOPLE TALKING ABOUT THE GAME IN THE GAME THREAD. Why don't you just stfu and deal with it for once in your life
Sorry if you're having a bad day, but that's not a very nice way to talk to people, now is it? p.s. I'm not the only one who's a bit sick of the xG obsession. And since this MD thread had devolved into a general discussion about xG, that took it off-topic for a page or two of posts, I frankly thought it was a reasonable request to make. And I did it in good spirits and good faith.