Early season performance data visualised

[beatific blade]

Courtesy of Ben Mayhew, otherwise known as Experimental 361

These scatter graphs show the blades to be fully deserving of their title as one of the early front runners.

Look at Preston though





----------------------------------------






----------------------------------------






----------------------------------------

 

[Ned the Donkey]

Courtesy of Ben Mayhew, otherwise known as Experimental 361

These scatter graphs show the blades to be fully deserving of their title as one of the early front runners.

Look at Preston though


View attachment 143495


----------------------------------------


View attachment 143496



----------------------------------------



View attachment 143497


----------------------------------------


View attachment 143498

Re Preston (for instance): if your stats look like that then perhaps it's time to re-evaluate how you're collecting and presenting data?

Maybe they are that much of an outlier. Or maybe you're measuring the wrong things in the wrong way. Both are possible.

Look at the pretty colours.
What about the source? Is anything meaningful being visualised here?
The colours are really pretty.

 

[blade86]

Re Preston (for instance): if your stats look like that then perhaps it's time to re-evaluate how you're collecting and presenting data?

Maybe they are that much of an outlier. Or maybe you're measuring the wrong things in the wrong way. Both are possible.

Look at the pretty colours.
What about the source? Is anything meaningful being visualised here?
The colours are really pretty.

Preston really are an outlier - 8 games played, scored 2, conceded 1. Sounds dull to me

 

[Berkshire Blade]

Re Preston (for instance): if your stats look like that then perhaps it's time to re-evaluate how you're collecting and presenting data?

Maybe they are that much of an outlier. Or maybe you're measuring the wrong things in the wrong way. Both are possible.

Look at the pretty colours.
What about the source? Is anything meaningful being visualised here?
The colours are really pretty.

Outlier in terms of goals in their games + early in season.

 

[GrunkyPeep]

Re Preston (for instance): if your stats look like that then perhaps it's time to re-evaluate how you're collecting and presenting data?

Maybe they are that much of an outlier. Or maybe you're measuring the wrong things in the wrong way. Both are possible.

Look at the pretty colours.
What about the source? Is anything meaningful being visualised here?
The colours are really pretty.

How would you better present the stats? Looking at them on the league table, you can clearly see their defence and attack is massively different to the rest of the teams in their effectiveness.

The graph backs that up too. Their attack is very poor, their defence is very strong. Attack is way worse than the rest of the league and defence is way better than the rest of the league. Both the numerical and visual data show this.

Any graph that didn't show Preston out on their own would be misrepresenting the statistics.

 

[blades95]

FiveThirtyEight's model has us and Norwich level on points for the end of the season:

 

[beatific blade]

Preston's start to the season is a massive anomaly given that they've had five 0-0 draws.

It's a fact that the ratio of shots-to-goals will be ridiculous, and the graph represents that.

As the season goes on, the graph will look more normal in terms of the variance and Preston will be pulled back into the pack as they regress towards the mean.

 

[SwissBlade]

I looked at the league table and concluded that of the games played so far, we’re the best team in the division based on league results

 

[Ned the Donkey]

Outlier in terms of goals in their games + early in season.

This is a fair point. Small sample size can mean exaggerated "effects". But if stats are being presented honestly this should be acknowledged.

 

[beatific blade]

I looked at the league table and concluded that of the games played so far, we’re the best team in the division based on league results

Hey Swiss, did you look at the league table when Reading were top and conclude they were the best team in the league? Their performance data didn't (and still doesn't) stack up compared to the results, which suggests they won't be up there for much longer. The table does indeed lie sometimes.

Our data matches up. Better chance of us staying up there if we keep up the same level of performances.

But you know this

 

[Frank the tank]

I get the same feeling I had when covid derailed our rollercoaster
This is it boys and girls we will come back from this enforced break and fade away into nothingness like the weed covered streets of an abandoned dystopian city

 

[SwissBlade]

Hey Swiss, did you look at the league table when Reading were top and conclude they were the best team in the league? Their performance data didn't (and still doesn't) stack up compared to the results, which suggests they won't be up there for much longer. The table does indeed lie sometimes.

Our data matches up. Better chance of us staying up there if we keep up the same level of performances.

But you know this

Performance data doesn’t get you promoted points do

 

[Ned the Donkey]

How would you better present the stats? Looking at them on the league table, you can clearly see their defence and attack is massively different to the rest of the teams in their effectiveness.

The graph backs that up too. Their attack is very poor, their defence is very strong. Attack is way worse than the rest of the league and defence is way better than the rest of the league. Both the numerical and visual data show this.

Any graph that didn't show Preston out on their own would be misrepresenting the statistics.

This entirely misses the point. I mean totally and completely.

The whole point is that if your stats look like this then you should perhaps question the validity of the source data, not pick nicer colours.

Shots faced per goal conceded. All teams apart from two are grouped between 5 and 13. Rotherham are on 17. Preston are on 105.

The mean is about 10 or 11. The standard deviation is about 3.2, say.

So Preston are around 30 standard deviations from the mean. If I was getting results like that I'd look at my model

For a completely preposterous comparison which has all sorts of problems but might get a point across: given a population mean height of about 178.2cm and a standard deviation of 6.2 (both of which the internet tells me are true, so they must be) a British male 30 standard deviations from the mean would be...mixing units...12 feet tall.

If I've got a model predicting any 12 foot tall people I'm looking at my model, not going out in search of 12 foot tall people.

Again, in the end, none of this may apply, Preston may regress to the mean and there will be no 12 foot tall people.

Till then I'm not going to just gawp at the pretty colours, I'm going to question the data.

 

[BladesOnToast]

FiveThirtyEight's model has us and Norwich level on points for the end of the season:

View attachment 143500

Surely a modelled projection for the end of season table is only valid if done by a "supercomputer"? Any regular computer simply isn't up to the task.

 

[Panini]

Preston really are an outlier - 8 games played, scored 2, conceded 1. Sounds dull to me

Absolutely. These visuals are really telling.

It's a very specific stat but the previous record low total (scored + conceded) for the first 8 games of any English season ever is 7 goals. Preston are on 3.

 

[GrunkyPeep]

This entirely misses the point. I mean totally and completely.

The whole point is that if your stats look like this then you should perhaps question the validity of the source data, not pick nicer colours.

Shots faced per goal conceded. All teams apart from two are grouped between 5 and 13. Rotherham are on 17. Preston are on 105.

The mean is about 10 or 11. The standard deviation is about 3.2, say.

So Preston are around 30 standard deviations from the mean. If I was getting results like that I'd look at my model

For a completely preposterous comparison which has all sorts of problems but might get a point across: given a population mean height of about 178.2cm and a standard deviation of 6.2 (both of which the internet tells me are true, so they must be) a British male 30 standard deviations from the mean would be...mixing units...12 feet tall.

If I've got a model predicting any 12 foot tall people I'm looking at my model, not going out in search of 12 foot tall people.

Again, in the end, none of this may apply, Preston may regress to the mean and there will be no 12 foot tall people.

Till then I'm not going to just gawp at the pretty colours, I'm going to question the data.

Your comparison is preposterous because you are comparing a simple data source and a calculation. Your calculation is always going to look hugely 'different' because it scales differently.

Had you looked at the goals scored or goals conceded you'd see that Preston were inside 3 standard deviations for both. Outlier, sure, but not massively misleading like your 30 SDs.

 

[dc_blade]

If before the Blackburn game you said "I've been back and watched all their games so far and I think they're in a false position: they've hardly created anything and won games with fluky goals and long range shots" no one would really bat an eyelid.

But when you say exactly the same thing but in terms derived from data analysis ("they've scored 4 goals on an xG of just 0.5") then people start grumbling.

 

[Gimli]

This entirely misses the point. I mean totally and completely.

The whole point is that if your stats look like this then you should perhaps question the validity of the source data, not pick nicer colours.

Shots faced per goal conceded. All teams apart from two are grouped between 5 and 13. Rotherham are on 17. Preston are on 105.

The mean is about 10 or 11. The standard deviation is about 3.2, say.

So Preston are around 30 standard deviations from the mean. If I was getting results like that I'd look at my model

For a completely preposterous comparison which has all sorts of problems but might get a point across: given a population mean height of about 178.2cm and a standard deviation of 6.2 (both of which the internet tells me are true, so they must be) a British male 30 standard deviations from the mean would be...mixing units...12 feet tall.

If I've got a model predicting any 12 foot tall people I'm looking at my model, not going out in search of 12 foot tall people.

Again, in the end, none of this may apply, Preston may regress to the mean and there will be no 12 foot tall people.

Till then I'm not going to just gawp at the pretty colours, I'm going to question the data.

But this isn’t a model ‘predicting’ anything. It’s telling us what has happened. PNE figures might appear to make the graph look ridiculous, but that’s because their start to the season has been ridiculous - 8 games played, 1 goal conceded and 2 scored, clearly that’s going to skew the data somewhat.

By picking up on the outlier to discredit the whole graph misses the point - the rest of the teams fit pretty well into the model and also PNE will regress towards the norm at some point.

 

[SUFC Guy]

Surely a modelled projection for the end of season table is only valid if done by a "supercomputer"? Any regular computer simply isn't up to the task.

Or Timothy the Octopus