Sardar Azmoun @ Zenit St. Petersburg | 2020-21

[nisfejahan]

Sorry but these statistical models are a mainly bit of entertainment industry more than anything else. In general, Football performance of any one individual CANNOT be accurately modeled and any such statistical data is at best a historical average and at worst a bit better than astrology.

The reason is fairly simple: Performance of any individual, is part of what is called "complex correlated system". Google it if you are interested, there is a ton of scientific literature.

Long story short, you can say that so and so is a good forward because he scores lots of goals. You can also say that there is a strong correlation between how much a forward runs and the number of goals he scores. BUT YOU CANNOT SAY that if the average Joe just keeps running like a headless chicken, he is going to score goals, because a lot of other shit (Complex) has to happen at the same time (Correlated) before the forward can score a goal.

I have used this example before (its my favorite!!). There is a strong correlation between number of ice creams sold at the beach and the number of people drowning at the beach. a) It does not mean that ice creams cause drowning and b) If you ban ice cream sales, it will have no effect of the number of drowning victims.

These XG and similar models should be taken with a huge bag of salt since they are an attempt at modelling a "Complex Correlated System". If they had enough validity, BHA would not be languishing at 17th place. (Hear the mic drop...BOOM)

NEXT UP.....I will explain the difference between Theory and Practice.

[Taz]

You know how much I like to debate this, you tease

1) I don't understand why you refer to it as entertainment industry, it's simple data analysis.

2) you're quite correct that it's a simplfied view of a complex system. But the theory behind it is solid. They've used an incredible amount of historic data to accurately model the probability of each chance and then it's simple a task of summing those for each player/team and comparing it to their actual output. If your argument is that it's not possible to assign an accurate probability to the chances fairplay but it's basically what we all do in watching a game, we decide how good the chance Azmoun just had was and then depending on if he scored it or not, draw our conclusions. The advantage a mathematical model has is it's not skewed by bias like filthy humans and has a much superior memory.

3) The conclusions to be drawn from the stat are again something the filthy humans are in charge of so it is prone to abuse. But there is no real arguing with the fact that if a player consistently (data set being significant e.g. 5+ years) overperforms on his xG, then he is quite clearly a gifted finisher. Simple conclusions like that are easily and accurately identified using xG.

4) there can naturally be outliers to the model, Brighton may be one such example. Think of tossing a coin 10 times. The models suggests it should be 5 heads and 5 tails as the outcome. But you could well get 9 heads. That is your Brighton. That outcome does not mean the model that predicts 5 heads and 5 tails is in any way innacurate. And if you were to toss the coin another 90 times. You would see the outcome move closer to the model's predicted outcome of 50 heads and 50 tails. Similarly, if Potter wasn't about to get hired by a bigger cluband was allowed to manage Brighton for a few consecutive seasons, you would see them either reduce their xG in line with their true ability or they would maintain their outstanding xG and their success would reflect their xG in the league table.

[nisfejahan]

You know how much I like to debate this, you tease

1) I don't understand why you refer to it as entertainment industry, it's simple data analysis.........

Yes, I agree with everything you have said. You simply stated what statistical analysis theory is. Which brings me to the difference between Theory and Practice. As usual let me demonstrate with an example:

Boy asks his father what is the difference between theory and practice. His father says go ask your grandma if she would sleep the neighbor for $1million dollars. The boy asks her and the grandma says yes. The boy goes back and reports the findings to the father. The father then says, so son, in Theory we could be millionaires, but in practice we have a slut living in the guest room.

Yes, in Theory, historical information can be used as a guide. But that requires a) a knowledge of significant factors that went influenced the information, and b) knowing the causal relationship between the factors and the underlying information. Correlation models try to get a stab at those causal relationships. But in practice, complex correlated systems, when there are numerous involved factors and there are binary and tertiary and higher order interaction, and these factors change over time, then finding the causal relationships becomes harder and harder and using the models to predict outcomes becomes more and more dicey. T

The xG model is simply one of these models that is trying to predict a very complex systems with not enough information, so it becomes a bit of entertainment. The historical information simply is not that applicable. The game has changed even in the last five years, dominant strategies have changed, and other teams have adapted to these changes by making changes on their own. BHA is not simply an outlier in the xG model. The difference between the xG points, 55 and actual points, 34, (a surplus of +21) is the failure of the model because it does not have an accurate accounting of many pertinent factors. Last year's champion, Liverpool had an xG of 74 and actual of 99 (a deficit of 25!!). It is not that Liverpool was an outlier. Any soccer fan who watched them knew they are going to dominate. Because the human mind intuitively can take those factors into account, where a simpler model cannot.

And in conclusion, this is my post from last August:

But rest assured, Leeds is not a direct threat to BHA, as BHA will fight relegation, and Leeds in all probability will finish mid table, well above the relegation zone where BHA permanently occupies.

What did I know back then that xG missed?!!!!!

But back to Sardar. If you want to continue, open up a separate thread on xG and will be happy to oblige.

[Taz]

But xG doesn't attempt to predict anything, it's a model to analyse events that have occured. Liverpool overperforming on xG last season has been shown up this season. Leeds were widely accepted to be a safe mid-table side at the start of the season by the betting markets as they use xG models to assess sides. It's all legit.

[KasraKhan]

For once, Taz's xG stats actually benefit Iranian football

[Taz]

[nisfejahan]

But xG doesn't attempt to predict anything, it's a model to analyse events that have occured. Liverpool overperforming on xG last season has been shown up this season. Leeds were widely accepted to be a safe mid-table side at the start of the season by the betting markets as they use xG models to assess sides. It's all legit.

Sorry but this is so full of contradictions, it borders on nonsense. You explain away xGs failures last season, but claim that Liverpool's xG last season is predicting their results this season. But Leeds preseason xG is accurate, although this is their season in EPL and last year's season they were in a completely different league. And then you make the claim that xG is not predictive, yet you tout its (failed) predictive power.

Bottom line is this.....Human behavior of 24 people on a team, 11 of them on the field at a time, their mood, injuries, a coaching staff of 10s of people, their mood and biases, is just too complicated to model. In theory you can, but in practice it fails. But it is good entertainment as these posts clearly show. There is no need to argue just look at the results.

Its like when you tout Potter as a coach and he keeps failing. BELIEVE THE RESULTS!!!! That's what matters!!

p.s. last post about this topic in this thread from me!

[Medzdidz]

Whether we agree or disagree on how useful these stats are, their are more and more clubs who are using them to make signings and help complete their line up.

[sam20]

Zenit Vs Lokomotive in a few min!! - If Zenit win, they clinch the title!

[Keshwardoost]

Zenit Vs Lokomotive in a few min!! - If Zenit win, they clinch the title!

Fantasic! Another title for Azmoun! This might one be his last games at Zenith if he is sold during the summer transfer window right?

#
In a few hours;
Breaking news! Leaked photos and videos of Azmoun and Dzyuba title celebration!

Hope this time, atleast Sardar is the giver, and not the taker..


#
Link: http://www.hesgoal.com/news/79052/Ze...iv_Moscow.html

[inarsenewetrust]

awesome chip by dzyuba after azmoun forces his way thru box and has shot saved

[Medzdidz]

Zenit 1-0 - lovely finish from Dzyuba and Sardar has a big part in the goal. He nearly played a lovely through ball for Malcolm who can’t get it under his control , ball bounces back to Sardar due to good hustle and goalie makes a save from his shot.
Ball then comes off the defender to Dzyuba who places it over the goalie.

[Keshwardoost]

Siiiick goal by Dzyuba after first giving an awsome NHL hockey tackle to win the ball.. WOW!

[MehdiKing]

any links to the game guys?

[Keshwardoost]

http://www.hesgoal.com/news/79052/Ze...iv_Moscow.html