Since expected goals only measure where the shot was taken from, they tend to undervalue counterattacking shots, which are typically taken with fewer defenders in the vicinity able to block or obstruct the shot. Also, expected goals are a very interesting stat in that they may claim to measure team success minus finishing success. A counter stat would be (quality) finishes: Spurs 1-0 City. The result, expected goals indicate, was flukey. But was it? For all their 550 million pound team, Man City conspicuously lacked a quality finisher, while the best finisher in the top five leagues by their expected goals to actual goals ratio was a fellow named Heung-Min Son. It's a bit absurd to exclude finishing success from an evaluation of team success when it's the skill people pay the most for and delight most in seeing. The result and expected goals stat make a powerful argument for City paying not just 130 or 150 million pounds for Kane, but 200 million. If you've already spent 550 million, and are willing to spend 120 million more, what's 80 million on top of that, if it makes you favorites for the league? Especially when the strategy of unsettling Kane, then getting him on the cheap, has already backfired, considering City have just dropped three points that may prove crucial to their league hopes, while Spurs don't have to care how long negotiations go on for, since they won't win anything big this year no matter what happens.
Boring stattery incoming (for those who wish to spare themselves ) ... Went thru this some time ago, This "xG" stat is both a misnomer and (in data analytics parlance) "non-actionable" . 1. Every shooting opportunity OP (position of shooter, position of other players, distance from goal etc) has a probability P(OP) of resulting in a goal. The mean P(OP) , MOP(OP) , could be derived from giving every outfield player 100 attempts to score from OP. 2. The basic measure of attacking prowess for any team over a game, would be to list all the opportunities that arise, and get a mean team P(OP) TOP per game. The higher the TOP, the more likely your team scores goals. 3. An opportunity OP falls to player PL. From 1, by definition you wish PL to be a player for whom P(OP) >= MOP(OP) . - Reality 1. The problem is that the set OP is infinite, the number of times any OP is observed would be statistically low (so getting MOP(OP) across a league would be difficult) . 2. The "xG" stuff AFAIK is not even public domain, so nobody can freely do methodology on it (even against theoretical methods) .
I completely agree with this. The point I was trying to make was that however well we play, the opposition can still win if they take the chances they will inevitably take. That's why I'm not a result merchant. Too many things are outside the control of the manager and the team for individual results to matter.
Very basic probability/statistics. And more of an attempt to stop the rise of "xG" in any statto discourse about football.
Not a very good presentation (I assume tis a graph - but no axis dimensions) , but I will look into the measure further (and give my opinions on it when I understand it) .
I have a relative on my wifes side who is a Trekkie and a gooner , he watches Star Trek constantly and I never fail to piss him off when I ask " is this the trouble with Tribbles episode " childish I know but if it winds a gooner up it's worth it
