It may be the whole point of how traditional xG was designed. But it's not the whole point in assessing what actually happened in a game or a half or even an individual shot. And you yourself brought up that there are post-shot xG models. That must surely imply that at least some people think that the existing/traditional xG is not the whole point. You also suggest that you can just watch the game to know. Sure watching the actual game is at the top of my pyramid for how to understand what happened. But it's insufficient on it's own, for so many reasons. And everyone recognizes that stats, in particular backward looking ones, are vital... to quantify & compare certain categories & factors. For the vast majority of people, the eye test and short-term memory are just not well suited to that.
It seems that way. And it's one of the limitations of xG IMO. TBF, even old traditional stats like possession have some level of flux, and inconsistency between different modelers. But my gut says those differences are smaller amplitude than what seems to be happening with xG. Like a lot of things in the world, standardization can be a very good, useful achievement (e.g. EV charging stations?) but getting there can be a little hard.
All this conversation and you still don't get it ... I'm at a loss. xG is an aggregation statistic that measures the expected goals one would expect to see from any given shot. Aggregation means you take all the shots and add them up and see what spits out -- yes, that means new shots too. Does any single shot affect the outcome -- yes, but it's so minute that it's immeasurable -- it's probably more like .00000000000X -- it's one out of all the football shots taken since data was recorded in that particular model. The aggregation is specifically made to do the opposite of what you're trying to get out of it -- it's meant to take human calculation out of it because we are biased and make (stupid) decisions based on feelings. You're trying to beat xG for some better metric that takes feelings into account and that's literally the whole point of the stat -- to inform decisions based on data. Or, to put on my data science hat, ******** your feelings.
Haha, I could say the same about you (and others) You're talking about something different than I'm talking/asking about. You seem to be in a narrow traditional xG mindset. I AM NOT. And as I mentioned somewhere above, if there were a metric for what I'm thinking about it might not even have the expected "x" prefix. If you didn't catch that post, you might wanna reread it.
On a lighter note, I just heard James Richardson refer to this game as the "Charles Dickens Derby". First I'd heard that. Made me smile. As he often does.
Analytics made sport a bit boring to be honest it was nice to see different systems rather than every top side wanting to play a 433 with a false 9 and inverted fullbacks etc same as the nba bar outliers like Steph it’s all very boring
Just procrastinated down an xG rabbit hole for a half hour or so. This was the top hit when I googled "post-shot xG". Lo and behold it's by "our own" Scott Willis... https://www.cannonstats.com/p/advanced-stats-101 Scott, who we've heard on ArsenalVison pod, seems a nice honest guy, deep into stats, open to tweaks/improvements, perhaps not so much drinkin-the-koolaid. In the context of our little discussion above, a few things I found interesting: He really likes to use 70% xG & 30% G, kinda blending what I was calling origin and outcome data. But his examples of using this is for bigger data sets, e.g. a team's entire season. AND the outcome data is very narrow, only whether it's a goal, not how far it missed by, or if it hit the post. With those limitations, it wouldn't be so hot for comparing 2 halves of AFC-PSG, I suppose. He talks about PSxG (post-shot xG) in the context of analyzing a goalie's defensive performance, more than an attacker. Different than what I've been thinking about, but it does take into consideration the ball position at the goal line. He says it's a very noisy stat, needing big data sets to begin to overcome that. I wonder if his thinking or application of PSxG has evolved in the year since he wrote this. Or the stat itself. Or maybe he said more about it at the end, but it seems pay-walled so IDK.
The metric you’re looking for is the actual outcome data. You don’t need any fancy math for that. Just watch the games and enjoy them and be surprised by the goals that happen or don’t happen. It’s great!
This kind of thing Scott is doing I think is getting lost in bending your model towards the outcomes, which in my opinion waters down the output. It’s like FIFA trying to get their world ranking to match a snapshot of who is the best team right now, which they did for a long time*. If they get the formula right they can make it match the current situation, but that doesn’t mean the model will produce accurate results going forward. *Maybe they still do this? It’s been so long since I’ve given a shit about international football that I have no idea what’s going on there anymore.
it’s a generic statement and he is and isn’t but u get my point The game isn’t as exciting imo even if I enjoy us and how we set up it’s miles less run than 02/04 or 98
Except then what’s the point of any backward-looking stats like possession, duels won, etc, etc? ps. I know you know that’s how I mainly do games… I watch. Rare glances at any stats during play. And of course copious PBP posts gotta cut back on those!
this is my biggest pet peeve on his games for germany. it feels like everyone is accused of being a false 9 these days unless they play like a trad-9
Feels like a lot of stats are only really used to justify gambling which is a big issue with sports in Australia. And as a consequence a lot of pundits who don't have a lot going on from the shoulders up have a job because they can pick a stat or two and parrot about it and act like an expert.
I’m not going to be naive and suggest that none of us are sports gamblers, but I suspect that most of us use stats, developed by and for gamblers, to confirm what we’ve seen. To use a player referenced above, if you go to a Warriors game and see Steph Curry get hot, you automatically check the box score the next day to confirm how hot he was. With respect to this game, I saw a dominate defensive performance and PSG’s xG confirmed that dominance, but football being football, their shots that hit the woodwork might instead have hit the back of the net. However, that doesn’t change how dominate our first half press was against a very skilled team or how dominate our mid to low block defense was in the second half. NorthBank’s argument seems to be that we were lucky but xG doesn’t reflect how lucky we were. But they really weren’t high quality chances, the shot that hit the post was taken on the run, under pressure from a wide angle and the shot that was was bounced off the ground and onto the bar was taken on the stretch with the studs of his boot. As a glass half full guy I think we would have been very unlucky if either shot hit the back of the net. As a side note, with games every three to four days Arsenal can’t press all game, every game, so I suspect we will see more games like this, where they press in the first half to get the lead and then settle into a mid to low block in the second half to conserve energy.
The most artistic ways of scoring in basketball (the post up and the mid range) have been diminished by the efficiency of the 3 point shot. Steph's and Klay's special power ruining basketball is unpopular to say, but it's largely true. That being said, Steph's ability around the rim is greatly underrated.
The three point shot has changed the way basketball is played, but it hasn’t ruined it. The spacing created by three point shooters opens up more lanes for drives to the basket, something you allude to when talking about Steph’s ability around the rim and something I find infinitely more artistic than someone like Shaq, bullying his way to the rim.
I don't think it's stats that are wrecking the game so much as insane 'fitness' levels - there was data released recently about how much more players run compared to 10 years ago with all the defensive pressures - yet that era of players (think peps barca) were already drug monsters compared to the early 00s This is why you see such complex patterns of play. Nowdays a classic playmaker type guy like pires wouldn't survive against the defensive pressure which is why those players died out. it sucks IMO
I dunno, moving without the ball is a god damn art to me. It must be absolutely terrifying to defend now -- the pick and roll seems relatively straightforward compared to what guys have to contend with now -- darting around 10 million screens and trying figure out if you have to go high or low, and then the ball moves and then you STILL have to chase the guy around.
Good post. A few reax... You got pretty close to characterizing what I said and how I felt. (Note: I've actually now spent way more time writing about that than the time I spent actually watching/focusing on that half and the chances in it, and never watched it back, so I guess I could apply the few-grains-of-salt metaphor) But I wouldn't say that we were lucky overall, because even with 35% possession, overall we drove the outcome. As I said at FT it was a game of 2 halves, the 1st felt like dominance, and the 2nd felt like we could've conceded (obviously that shot onto the bar was in my head). Right after FT, when I heard of PSG's 0.26 xG, I wrote: I probably should watch 2T again, but without doing that I still feel about the same, i.e. a bit fortunate to not concede, regardless of xG saying that the shot onto the woodwork was statistically low probability at the moment it was struck. Lastly, your comment about this game plan becoming more common. I think Tim Stillman said the same thing on one of the pods (AV?) this week: We may see more 1T's with high intensity and (hopefully) a few goals, and then 2T's where we relax/recede more, with the dual goals of a) winning, and b) preserving the health & fitness of our most valuable players. If this turns out to be true, I could gladly accept it. Even if my typical instinct is to keep going for more goals through the whole match, with the mantra of a-good-offense-is-the-best-defense. How Mikel manages game plans, rotations and in-game adjustments this year is going to be fascinating, and pivotal to our title challenge.
Yeah the defensive press stifled the game that gave more creative players a bit more time and space sadly the ten is dead and the classic players all teams would play the most efficient if they could so you can blame them we have started to look more like a Russian Olympic team that is yamming steroids lately and it’s working
I just think based on my experiences in AFL in Australia and the gambling plague we have here, the obsession with stats to justify anything in matches coincides with multis for just about any stat and its too big a coincidence for my liking. You are right that a LOT of stats are worth having, I quite like seeing the xG after games to see how shots stack up because I think its interesting especially after, for example, the Leicester match where you could see from watching the match we had quality chances but the GK has gone beast mode and was saving everything. I also think for the most part a lot of the big football sites have been good with not listing every stat. I just checked the Arse/Lester match stats on Transfermarkt and they only list; Possession %, Total shots, shots on target, shots saved, corners, free kicks, fouls and offsides. You could probably also list cards and thats probably all you need to get a good view of how the game was for Arsenal v Leicester.