Birdy, ok.
After reading this, I can see one flaw already.
Our game is designed around Messi.
And for example, if Barca creates 5 chances.
And if RM or Inter create the same number of chances, we will have the same XG, right?
But our results will be better.
Due to luck?
No.
Because we have Messi and Suarez shooting.
Inter has Lautaro who can't score.
RM has Vinicius who gets in 5 chances but can't score.
So, our XG would be higher with Ansu or Braithwaite, but our results will be higher with Suarez.
Some players get into more chances but are bad finishers.
Some don't get into too many chances, but they score a lot.
About a low XG og EV's Barca: our team is not as bad as XG shows.
We just don't shoot too much.
Instead of shooting, we pass to Messi.
So, we have low amount of chances, but we seem to win more than we should, right?
When I posted stats, people often told me: stats can't tell the whole story.
I have spent only 2 minutes thinking about XG and you can already find flaws, as explained above.
1)
@BBZ8800
There is some truth to what you say, but there is also a strong flip side.
It's true that world-class attackers for example tend to finish their chances better, and xG reflects the probability given all different levels of players, thus reflects the probability of an 'average' finisher.
So, Messi for example will make the 0.1 chance into a goal not 1/10 times which is the indicated probability, but maybe 5 out of 10 or something like that.
And i agree that the next leve lfo evolution of xG stats will be to 'adapt' the probabilities to the track records of specific players, and that would give a more accurate picture.
Now, notwithstanding the above, the argument that 'we have Messi so we adapt our chances to his qualities' is largely a rationalization of the pathology of the team, and I will tell you why:
Yes, Messi can finish better, but giving Messi for example a game where he shoots 10 times with 0.01 every time is praying for a miracle to happen.
In a system that would utilize Messi's qualities best, you would give Messi the chance maybe to shoot less, let's say 5 times during a game, and every time the probability of 0.1-0.4, which means bringing Messi to positions of finishing where he can be really clinical given his qualities.
Messi-dependence is a pathology, not a viable tactical strategy.
Of course working the in and outs of how can a tactical plan bring players to positions where they can take a chance with higher probability requires work, both from the coach, and from the players. Work, which has not been done at least the last 4-5 years.
Likewise, in defense the same can be said about Ter Stegen.
Yes, Ter Stegen saves better than probably all keepers in the world at the moment. chances of let's say 0.3-0.4 other keepers will concede 3 out of 10 times, and Ter Stegen might concede only 1 or 0.5 times.
But the 'we rely on Ter Stegen' line is not a defensive plan at all. It is rather the indication that the team is vulnerable defensively and needs a keeper to save its ass every game.
Do the necessary tactical work, so you are not exposed to good chances of the opposition...
2)
Overall
I see a lot of discussion about 'flawed statistics'
There are no flawed statistics. Stats reflect something.
The question is : i) what they reflect, ii) how you interpret it
Until some years ago, we would see the same stats displayed as in the mid-70s: shots, shots on target, possession, cards etc.
Now someone who thinks stats are exhausted to that is likely to assume that stats are useless or they don't tell the whole (or any) truth.
Of course, if a shot from 50 yards is counted as a number of equal value to a shot on 1v1 with the keeper, the total number of shots is at best an incomplete, and worst a misleading metric.
Room for interpretation is vast, given that set of stats.
The revolution in football stats the last 5-7 years has completely changed the rule of the game.
First of all the new stats, as more sophisticated, are more indicative of effectiveness.
The xG for example is the best metric available to measure how 'good a chance' is. And when it becomes relative to the individual, it will become the most accurate metric to measure how 'good' a chance is.
Example about how the new stats reflect more accurately the game:
Atletico-Bayern CL semi-finals 2016. Both teams demonstrate similar stats in the 'old set' of stats between 1st and 2nd leg: similar possession in both 1st and 2nd leg, similar shots, similar shots on target etc. However it was obvious to anyone watching the games that 1st leg had nothing to do with 2nd leg.
A look into the new stats, completely sets the record straight: 1st leg Atletico won on xG, and 2nd leg Bayern utterly battered Atletico on xG.
And of course there are deeper alleys you can walk through, why Bayern battered them on the 2nd leg, what changed tactically etc...
The increased accuracy of the new stats, incites also another change: it leaves less room for interpretation.
With the old set of stats, there was a lot of room and rightfully so. Now, there is still room, but less.
For example if the xG scoreline of a game is 0.4-2.2 for the opposition, you cannot argue convincingly anymore that your team played 'well', which was more difficult to stipulate in the past.
Liverpool's success story the last years under Klopp is also a testimony to that new 'scientific' approach to football. They are light years ahead of all other clubs in Europe in recruitment because they have specialized departments devoted to that type of 'statistical analysis' when identifying potential targets.