The Weight Of Expectation

[Robbored]

27 minutes ago, spudski said:

We are playing attractive football and really effective going forward...however...it could look a lot worse if shots had gone in against us.

Slim margins. 

That kinda confirms my last post ‘lies, damn lies and statistics’ A team can have a 20% possession and yet win the game with one shot on target.

Its obvious to anyone that I’m no statistician and to me what stats shown usually confirms what the eye can see. 

[Mr Popodopolous]

18 minutes ago, Robbored said:

That kinda confirms my last post ‘lies, damn lies and statistics’ A team can have a 20% possession and yet win the game with one shot on target.

Its obvious to anyone that I’m no statistician and to me what stats shown usually confirms what the eye can see. 

Over time though, it's highly unlikely that a side will keep outperforming their baseline performance.

Doesn't even have to be xG as such, more like shots, chances created, possession, pressure absorbed- we were at times significantly outperforming our baseline performance for example when we went 2 months unbeaten early in 2019-20.

Forget the stats too. Put simply if a team are persistently getting better results than performances, either the performances need drastic improvement or the results will subside perhaps even fall off a cliff sooner or later.

Edited September 8, 2022 by Mr Popodopolous

[Hxj]

1 hour ago, James54De said:

Presume you’ve never done any statistical analysis in your life.

Plenty thanks, back in the good old days () before you could put the numbers into a table and a computer would work out the answer.

Personally I wouldn't count plotting expected goals for against expected goals against as statistical analysis, it's just two sets of data plotted together to create a pretty picture.

The secret of statistical analysis is understanding the data sets and the null hypothesis. But then again I suppose poissonly I'm normally a mean stats snob.

(I'll let my stats lecturer take the credit for the whole of the last sentence  )

[Hxj]

1 hour ago, Mr Popodopolous said:

Over time though, it's highly unlikely that a side will keep outperforming their baseline performance.

A team will always match its own performance.  It is perfectly possible for a team to consistently out perform someone else's opinion of the average performance of all teams.

Take Conway, he has scored 4 goals and has an xG of 2.42.  There are several possibilities, one is that this is just a small sample and in reality in this case 4 and 2.42 are simply equivalent answers, with no statistical difference.

Another possibility is that Conway is simply 50% better at scoring goals than the statistically average player.  Therefore he would expect to score at a rate 50% higher than the average, he will therefore expect to outperform the xG rating by a significant margin.  The same applies to teams.

Edited September 8, 2022 by Hxj

[chinapig]

1 hour ago, Robbored said:

That kinda confirms my last post ‘lies, damn lies and statistics’ A team can have a 20% possession and yet win the game with one shot on target.

Its obvious to anyone that I’m no statistician and to me what stats shown usually confirms what the eye can see. 

The problem with the eye test is that we are all subject to biases. Add to that the established unreliability of eye witness testimony and relying on what you see, or think you saw, alone is not ideal. I offer up all those who blame Vyner whenever we concede a goal in evidence.

Which is why clubs use analytics as an additional tool in their decision making. There is an interesting Tifo vodcast on YouTube today with a data scientist who has worked with Premier League clubs that you might find informative.

In any event it's not difficult to find out what football analytics actually does before rushing to judgement if you are genuinely interested.

[Mr Popodopolous]

39 minutes ago, Hxj said:

A team will always match its own performance.  It is perfectly possible for a team to consistently out perform someone else's opinion of the average performance of all teams.

Take Conway, he has scored 4 goals and has an xG of 2.42.  There are several possibilities, one is that this is just a small sample and in reality in this case 4 and 2.42 are simply equivalent answers, with no statistical difference.

Another possibility is that Conway is simply 50% better at scoring goals than the statistically average player.  Therefore he would expect to score at a rate 50% higher than the average, he will therefore expect to outperform the xG rating by a significant margin.  The same applies to teams.

On the Conway point, you might well be onto something there. Certainly possible for an individual player to outperform an average of shots/goals for example, all players have different characteristics after all.

Players will have dips too, will be interesting to see how he levels out over time if indeed he does- or if a barren spell drops his overall output down to the average even if his baseline still higher.

Team wise I dunno. XG may not even be the best metric for this, granted I'm using more basic methods (shots vs shots conceded- shorthand for attacking power vs pressure absorbed), possession (useful for control).

What I'm trying to say I guess is that if a side is frequently winning with less shots and somewhat less possession let's say, that won't last. There will be a correction of some kind, it's an outlier.

Slightly better example maybe if a team is roughly level for shots for and shots against that's kinda midtable numbers, although Luton got to the playoffs despite this, Reading also got a good period of 2020-21 outperformed their numbers, Burnley for years in the PL were probably better than numbers at both ends plus possession wise would indicate so there are cases for outliers over a sustained period.

Edited September 8, 2022 by Mr Popodopolous

[chinapig]

25 minutes ago, Hxj said:

A team will always match its own performance.  It is perfectly possible for a team to consistently out perform someone else's opinion of the average performance of all teams.

Take Conway, he has scored 4 goals and has an xG of 2.42.  There are several possibilities, one is that this is just a small sample and in reality in this case 4 and 2.42 are simply equivalent answers, with no statistical difference.

Another possibility is that Conway is simply 50% better at scoring goals than the statistically average player.  Therefore he would expect to score at a rate 50% higher than the average, he will therefore expect to outperform the xG rating by a significant margin.  The same applies to teams.

First I agree with your earlier point about statistical analysis. I doubt football clubs often use the Pearson (no relation!) product moment coefficient of correlation for instance. Or maybe they do.

But we're being pedantic about terminology. In a day to day sense analysis can be useful in revealing trends and comparisons that are not necessarily obvious to the eye or even show that what you think you observe is wrong. Our own @Davefevsis a great example of that.

As to your first sentence above it seems tautological. Do you mean results will always match performance or something else? And is it a matter of opinion or fact, or at least probability?

What you say about Conway is true of course but as you also suggest there is not enough data to reach a conclusion yet. I have a feeling he may prove to be exceptional compared to his peers but that's just unsupported guesswork.

[Robbored]

56 minutes ago, chinapig said:

The problem with the eye test is that we are all subject to biases. Add to that the established unreliability of eye witness testimony and relying on what you see, or think you saw, alone is not ideal. I offer up all those who blame Vyner whenever we concede a goal in evidence.

Which is why clubs use analytics as an additional tool in their decision making. There is an interesting Tifo vodcast on YouTube today with a data scientist who has worked with Premier League clubs that you might find informative.

In any event it's not difficult to find out what football analytics actually does before rushing to judgement if you are genuinely interested.

Two guys from the analytics team were speakers at Senior Reds about two years ago and we were all amazed to hear just in depth the various stats are. They include every possible stat you could think of including analysis of each individual players during the game from sprints, to ball carries, distance covered and even the pulse rate of each player. For obvious reasons the active analytical info is not permitted to be passed to manager/coaches during the match - that’s a significant  part of the fourth official duties.

I didn’t realise that every player has an app on their phone which they report on before traveling to training. How they slept, how tired they may feel or any other symptoms such a cough or thickly throat. That really surprised me.

Its not that I don’t appreciate the importance of match stats,  I obviously do but am not motivated enough to follow whatever site that they’re on.

To me that’s the job of Nige and the coaches who know how to apply the various stats information 

Edited September 8, 2022 by Robbored

[spudski]

2 hours ago, Robbored said:

That kinda confirms my last post ‘lies, damn lies and statistics’ A team can have a 20% possession and yet win the game with one shot on target.

Its obvious to anyone that I’m no statistician and to me what stats shown usually confirms what the eye can see. 

RR...your first paragraph contradicts your second paragraph

[Hxj]

18 minutes ago, chinapig said:

I doubt football clubs often use the Pearson (no relation!) product moment coefficient of correlation for instance.

Imagine doing that by hand for a large sample!

 

19 minutes ago, chinapig said:

In a day to day sense analysis can be useful in revealing trends and comparisons that are not necessarily obvious to the eye or even show that what you think you observe is wrong. Our own @Davefevsis a great example of that.

Agreed, I suppose my real issue is that if there is a unexpected result then the secret of the analysis is "why?", the answer isn't always "it will return to the expected value in due course."

 

22 minutes ago, chinapig said:

A team will always match its own performance. 

21 minutes ago, chinapig said:

Do you mean results will always match performance or something else?

No, far simpler, and back to my previous point  The performance a team/player achieves is a known.  As is the performance of the 'population' as a whole. 

Moving on a bit if the statistical analysis shows a difference, and I mean a statistically significant difference, not a numerical difference, then the reason why is the really important bit, not the difference itself.

[Hxj]

1 hour ago, Mr Popodopolous said:

Team wise I dunno. XG may not even be the best metric for this, granted I'm using more basic methods (shots vs shots conceded- shorthand for attacking power vs pressure absorbed), possession (useful for control).

If you find the secret you will become a very wealthy person!

[Red_Alligator]

To be fair, my expectations centre around the long wait, as opposed to it's mass. 

[Davefevs]

7 hours ago, ExiledAjax said:

And in our attack they are facing one of the most clinical attacks in the entire 92. Those same graphs, and other numbers, show that we generally need to only take 6 or so shots to score. Personally I take that to mean that should we be able to create chances we are one of the more likely teams to score.

It's not prevention of shots that has kept Preston's goals against figure down, the final (ludicrous) graph from E361 shows that they allow an above average number of shots (about 13 per game). we have been taking about 12 shots per game. 

Taking those numbers together, Preston allowing 13 shots per game, us taking 12 per game, and us needing about 6 or 7 shots to score...well that gives us a predicted goals of about 2. Funnily enough we've been scoring 2 goals per game as well. It's therefore demonstrable, on the famous paper upon which football is not played, that we have the tools and are even 'likely' to unlock this blessed Preston defence.

6 hours ago, Hxj said:

Is there is a more meaningless graph to plot out than 'Expected Goals For' v 'Expected Goals Against' I'd love to know what it is.

Do they actually know that football is decided on 'Actual Goals Scored' v 'Actual Goals Conceded'?

Whether to choose to look at statistics beyond Goals for and against, what you will find is that Bristol City both create good chances and finish a good proportion of them.  We don’t create 20 long range efforts that end up in the net once in a blue moon, we score our goals from good positions.

We take more of our shots inside the area (77%) versus outside (23%), which is the 4th best in the Champ.