Football fans and data

[CheddarReds]

Data usage has increased hugely in recent years and football's choice to adopt data was predictable. We're aware football uses data to analyse team's, individuals, even coaches to an extent and it can provide information on key passes, fitness, strength as well as newer statistics such as expected goals (xG), expected threat (xT) and possession value. 

Personally I like to read about whatever stats are available which might make me see something differently, whether it's a player's performance or a team's string of results. I think having a data guy working alongside pundits would give a good insight and cause further debate too. Having worked in a performance analysis capacity I'm biased though!

There's been a bit of debate how much fans want to hear about these statistics. So are these type of stats something you are actually interested in hearing about? Would you actually want to read about a team's xG and xT relative to the result or is it something you have no interest in at all? Or would you be interested in the use of data for a particular niche that currently isn't available?

Interested to hear some thoughts.

 

[ExiledAjax]

So long as the limitations of the data are explained and considered as part of the analysis then I'm all for it. 

They also have to be used correctly. For example I wouldn't want a pundit to sit there at half time and use xG to support their argument that team A should be beating team B in the game they are covering. But if they used long term xG to explain why it was surprising that team A were high up the table going into the match, then that's fine.

It's about context, clarity, nuance and communication. The time restraints of football coverage maybe don't leave much room for that kind of stuff, but I'd be interested to see it being tried.

[Lrrr]

I'd like to see more of heatmaps for shot and shot assist locations, see where teams are trying to get their chances away from and the positions on the pitch they get into to do so. IE Man City well known for cut backs to create chances, we tried to upgrade shot xG to get into better value positions to shoot from so you'd expect a heat map for us to be around the penalty spot etc. Before a game you could show the heatmap of each side for each to show the viewer what to look out for in the game. 

[Davefevs]

Good OP and responses so far.

Its all about context.  What we are starting to see is the debunking of oldskool fans views of things like:

• possession stats
• shots / shots on target
• etc

being good proxies for who should’ve won a match.  It bugs the hell out of me that shots on target is still used so much.  I totally get that if you don’t hit the target you don’t score, but if we are using it is black and white as this example, then we might as well not bother, literally just use goals scored and conceded.  It also bugs me that people take xG out of context too.

The key to data use is understanding what is a 1 and what is a zero.  Passing is a classic.  Nagy passes to Martin.  Nagy gets 1 pass made, Martin gets 1 pass received.  The fact that it’s only just in Martin’s sphere of control, has been pinged at him at 100mph, at hip height and he mis-controls it means Nagy gets a 1 for a successful pass, yet Martin gets a 1 for balls lost.  Your eyes tell you it was a crap pass from Nagy, the data tells you different.  That’s why data isn’t the be-all and end-all....and I love my data.

 

[ExiledAjax]

3 hours ago, Lrrr said:

I'd like to see more of heatmaps for shot and shot assist locations, see where teams are trying to get their chances away from and the positions on the pitch they get into to do so. IE Man City well known for cut backs to create chances, we tried to upgrade shot xG to get into better value positions to shoot from so you'd expect a heat map for us to be around the penalty spot etc. Before a game you could show the heatmap of each side for each to show the viewer what to look out for in the game. 

Yes this is good.

If the data is useful, can be presented in an appealing way, and contributes something more than just numbers - ie raw shots on target the Fevs mentions - then it's useful.

Heatmaps are useful, and personally I'd like to see what you describe as well as, possibly, pass heatmaps. Something to show where and how an attacking team is probing at a defensive line, and whether than is creating good chances to score or whether those probes are being successfully rebutted or neutralised by the defending team.

2 hours ago, Davefevs said:

The key to data use is understanding what is a 1 and what is a zero.  Passing is a classic.  Nagy passes to Martin.  Nagy gets 1 pass made, Martin gets 1 pass received.  The fact that it’s only just in Martin’s sphere of control, has been pinged at him at 100mph, at hip height and he mis-controls it means Nagy gets a 1 for a successful pass, yet Martin gets a 1 for balls lost.  Your eyes tell you it was a crap pass from Nagy, the data tells you different.  That’s why data isn’t the be-all and end-all....and I love my data.

 

A CLEAR example of why we desperately need to create an xPasses metric. I am only partly joking.

My personal bugbear is the whole "Key" passes and "Key" contributions stuff. It's a proper bodge attempt at distinguishing between clear assists and generic passes that don't seem to do much. I get that you want to note something like Semenyo's through ball that lead to Luton's own goal as being more than "just" a pass, but generally just calling that a "Key" pass is nebulous, means nothing, and suffers from the same issue you raise - that it's entirely in the control of the activating player not the receiving player. Passing data as a whole is incredibly simplified at the moment, yes it's a complex part of the game but it is ripe for reassessment. Packing is an interesting attempt at this but seems to be struggling to get a foothold in the analysis world - at least publicly.

[Lrrr]

2 minutes ago, ExiledAjax said:

Packing is an interesting attempt at this but seems to be struggling to get a foothold in the analysis world - at least publicly.

Due to being a reactionary analysis tool rather than predictive/informative, you can't use packing to help you with how you want to play against someone.

[IAmNick]

I love looking at the stats people show, and using them to supplement what I see.

In addition to what others have said above I think they're a really important tool for understanding and inspecting natural biases we all form about teams, players, formations etc.

It's easy to think, for example "Christ, Vyner's long balls never find their man" or "O'Dowda always bloody passes backwards" - and maybe those are true, but I find it useful to then look at the stats and see whether that's really the case, or I just have a confirmation bias and only notice the instances when something happens to reinforce a view I already held.

I find that when people use stats to try and disprove a one off event and imply that means the stats themselves are useless, or say they have no place at all it's usually because they're not using them in the right way.

[Robbored]

3 hours ago, CheddarReds said:

Data usage has increased hugely in recent years and football's choice to adopt data was predictable. We're aware football uses data to analyse team's, individuals, even coaches to an extent and it can provide information on key passes, fitness, strength as well as newer statistics such as expected goals (xG), expected threat (xT) and possession value. 

Personally I like to read about whatever stats are available which might make me see something differently, whether it's a player's performance or a team's string of results. I think having a data guy working alongside pundits would give a good insight and cause further debate too. Having worked in a performance analysis capacity I'm biased though!

There's been a bit of debate how much fans want to hear about these statistics. So are these type of stats something you are actually interested in hearing about? Would you actually want to read about a team's xG and xT relative to the result or is it something you have no interest in at all? Or would you be interested in the use of data for a particular niche that currently isn't available?

Interested to hear some thoughts.

 

Undoubtedly there would be some posters who would find this type of information extremely fascinating  but I’m not one of them.............

However, having heard the data guys speak at Senior Reds I did actually find the amount of statistical analysis involved extremely interesting. Up then I’d had no idea just how much was involved.

To me all the data collected is an integral part of the coaching staffs remit and plays a significant role in each player’s individual strengths and weaknesses that may need addressing.

Using data analysis is the modern way but there is a risk of placing too much emphasis on it and I think that LJ was/is a victim of that.

[Pearcy]

I absolutely love football stats to the point I can study them for hours. However, even I find the stats for XG and XT pointless. I ignore them completey.

[Northern Red]

4 hours ago, Davefevs said:

Good OP and responses so far.

Its all about context.  What we are starting to see is the debunking of oldskool fans views of things like:

• possession stats
• shots / shots on target
• etc

being good proxies for who should’ve won a match.  It bugs the hell out of me that shots on target is still used so much.  I totally get that if you don’t hit the target you don’t score, but if we are using it is black and white as this example, then we might as well not bother, literally just use goals scored and conceded.  It also bugs me that people take xG out of context too.

The key to data use is understanding what is a 1 and what is a zero.  Passing is a classic.  Nagy passes to Martin.  Nagy gets 1 pass made, Martin gets 1 pass received.  The fact that it’s only just in Martin’s sphere of control, has been pinged at him at 100mph, at hip height and he mis-controls it means Nagy gets a 1 for a successful pass, yet Martin gets a 1 for balls lost.  Your eyes tell you it was a crap pass from Nagy, the data tells you different.  That’s why data isn’t the be-all and end-all....and I love my data.

 

Agreed. What would everyone consider to be a better chance - a shot that's fired in and has the keeper beaten all ends up but misses the target by an inch, or a tame shot from 20 yards that's straight down the middle of the goal that the keeper could bend down and pick up without moving? Yet someone just looking at the stats of a match would conclude that a team having 6 of the former is less deserving of winning the game than one with 6 of the latter.

[Davefevs]

2 hours ago, ExiledAjax said:

Yes this is good.

If the data is useful, can be presented in an appealing way, and contributes something more than just numbers - ie raw shots on target the Fevs mentions - then it's useful.

Heatmaps are useful, and personally I'd like to see what you describe as well as, possibly, pass heatmaps. Something to show where and how an attacking team is probing at a defensive line, and whether than is creating good chances to score or whether those probes are being successfully rebutted or neutralised by the defending team.

A CLEAR example of why we desperately need to create an xPasses metric. I am only partly joking.

My personal bugbear is the whole "Key" passes and "Key" contributions stuff. It's a proper bodge attempt at distinguishing between clear assists and generic passes that don't seem to do much. I get that you want to note something like Semenyo's through ball that lead to Luton's own goal as being more than "just" a pass, but generally just calling that a "Key" pass is nebulous, means nothing, and suffers from the same issue you raise - that it's entirely in the control of the activating player not the receiving player. Passing data as a whole is incredibly simplified at the moment, yes it's a complex part of the game but it is ripe for reassessment. Packing is an interesting attempt at this but seems to be struggling to get a foothold in the analysis world - at least publicly.

The good / expensive “paid for” data includes a model that scores every pass.  I haven’t quite got my head around it because it seems to use +/- values as to whether the pass went forward or backwards.  They key pass in a move could be backwards, so I need to study it more.

2 hours ago, IAmNick said:

I love looking at the stats people show, and using them to supplement what I see.

In addition to what others have said above I think they're a really important tool for understanding and inspecting natural biases we all form about teams, players, formations etc.

It's easy to think, for example "Christ, Vyner's long balls never find their man" or "O'Dowda always bloody passes backwards" - and maybe those are true, but I find it useful to then look at the stats and see whether that's really the case, or I just have a confirmation bias and only notice the instances when something happens to reinforce a view I already held.

I find that when people use stats to try and disprove a one off event and imply that means the stats themselves are useless, or say they have no place at all it's usually because they're not using them in the right way.

That’s what I do in the main.  Nagy made 48 successful passes out of 53 on Sunday.  I bet most of us, me included, thought he made more than 5 incomplete passes?  And that’s where if you then go in and watch his videos of his passes you see a successful pass that Martin loses possession off of....when in reality it was a crap pass.

For me the process is eyes (tv), data, video.  But the data and video aren’t always available straight after a game, so you make posts on OTIB / Twitter or comments on Forever Bristol pod from what you’ve seen initially.  Others see things differently. And even after data and video review you can still have a difference of opinion.

We all have our biases, it’s difficult not to, but I would say you can train yourself to lessen the extent of that bias. It’s why I try to do my previews and reviews with pictures.

[ExiledAjax]

11 minutes ago, Davefevs said:

The good / expensive “paid for” data includes a model that scores every pass.  I haven’t quite got my head around it because it seems to use +/- values as to whether the pass went forward or backwards.  They key pass in a move could be backwards, so I need to study it more.

That’s what I do in the main.  Nagy made 48 successful passes out of 53 on Sunday.  I bet most of us, me included, thought he made more than 5 incomplete passes?  And that’s where if you then go in and watch his videos of his passes you see a successful pass that Martin loses possession off of....when in reality it was a crap pass.

For me the process is eyes (tv), data, video.  But the data and video aren’t always available straight after a game, so you make posts on OTIB / Twitter or comments on Forever Bristol pod from what you’ve seen initially.  Others see things differently. And even after data and video review you can still have a difference of opinion.

We all have our biases, it’s difficult not to, but I would say you can train yourself to lessen the extent of that bias. It’s why I try to do my previews and reviews with pictures.

Agreed, weird that they've automatically called a backwards pass as a -ve. I'd have thought you could come up with some other way of representing backwards, forward, and just for Nakhi Wells' sake, sideways as well.

Interesting what you say about the review process. I think that goes to @CheddarReds OP. Essentially there's a time issue with having a 'data guy' sat next to the commentators. Doing that sort of analysis live during a football game is tricky. It's very different to something like TMS and the Cricket - in that sport there's tons of dead time to fill, and a 'data guy' can happily spend 30 minutes crunching and reviewing numbers that will still be relevant when he reports them because there will have only been 6 overs bowled in the intervening time. Do that in football and you get a comms team discussing in the 35th minute a shot that was taken in the 5th. That would be crazy.

You'd then only be able to talk about stats that come to you quickly - and thus we get discussions around very simple stats like SoT and possession %. Anything deeper just takes too much time, brain power, and consideration to give any value.

[1960maaan]

5 hours ago, Lrrr said:

I'd like to see more of heatmaps for shot and shot assist locations,

Not sure if you know of this one, I like to have a look every now and then. I'd like more info but I'm not willing to pay so that limits it.  
https://www.whoscored.com/Teams/182/Show/England-Bristol-City

I like the data, or most of it, it's just that some of it could come with explanations. They are constantly being updated and improved graphically , but I'm never 100& sure what I'm looking at. I've only just got my head around XG 

 

[Hxj]

I love data and data manipulation.

Two things you always need to bear in mind:

1. Data is the plural of a datum

2. To paraphrase Einstein - Not everything that counts can be counted and not everything that can be counted counts.

3. xG is rubbish, see above.

[Abraham Romanovich]

I'm probably in a minority of one but this emphasis on stats does my head in, they are a useful tool in assessing players and team style but they are only a tool and should provide another just another string to a bow not the only one. I feel there is a real danger of some managers not being able to see the wood for the trees in assessing individuals

Take for example Jimmy Greaves if he was 16 years old today would he be coached to track back learn patterns of play etc or just allowed to play his game get into the penalty area and score goals.

Apologies if I sound like a relic from a forgotten age, but nothing stays the same and in 10 years time I feel that the current fad of analysis will be viewed in a rather more negative light in terms of player development.

[Robin101]

Data is interesting when it adds an insight that we wouldn’t have seen otherwise. 

A bit like writing an article. If you add a photo, make sure that photo is contributing to the writing and not just there to look pretty.

Being bombarded with data for the sake of it is boring. Too much data - even if you think it’s relevant - means most people just switch off. Obviously it has its place in football, but does the average armchair fan want to see it in the half time analysis? Nah, show us the goal again.

[Davefevs]

Florence Nightingale:

• lady with the lamp or
• data statistician 

This article is really interesting.

https://thisisstatistics.org/florence-nightingale-the-lady-with-the-data/

This is little different to the player radar charts I’ve produced in the past.

[cider-manc]

4 minutes ago, Abraham Romanovich said:

I'm probably in a minority of one but this emphasis on stats does my head in, they are a useful tool in assessing players and team style but they are only a tool and should provide another just another string to a bow not the only one. I feel there is a real danger of some managers not being able to see the wood for the trees in assessing individuals

Take for example Jimmy Greaves if he was 16 years old today would he be coached to track back learn patterns of play etc or just allowed to play his game get into the penalty area and score goals.

Apologies if I sound like a relic from a forgotten age, but nothing stays the same and in 10 years time I feel that the current fad of analysis will be viewed in a rather more negative light in terms of player development.

I also can't stand the over emphasis. And I blame skysports news for needing a headline ever hour.

Statistics will tell whatever story you want to tell. Especially when used in isolation - which is what fans tend to do.

For instance - Its alright saying that a player made the most tackles in a game but what if they are only making all these tackles as they were out of position in the first place? It only takes them to miss one tackle and you could be in trouble.

Whilst data can be very useful for understanding the game (i see some great stuff on here and twitter) and i understand why clubs use it- the constant bombardment of stand alone stats from news sources and fans is annoying beyond belief. 

I dont know... maybe its because Im a data analyst by trade and therefore hate seeing my work life flow into my leisure time!

 

[Davefevs]

15 minutes ago, cider-manc said:

I also can't stand the over emphasis. And I blame skysports news for needing a headline ever hour.

Statistics will tell whatever story you want to tell. Especially when used in isolation - which is what fans tend to do.

For instance - Its alright saying that a player made the most tackles in a game but what if they are only making all these tackles as they were out of position in the first place? It only takes them to miss one tackle and you could be in trouble.

Whilst data can be very useful for understanding the game (i see some great stuff on here and twitter) and i understand why clubs use it- the constant bombardment of stand alone stats from news sources and fans is annoying beyond belief. 

I dont know... maybe its because Im a data analyst by trade and therefore hate seeing my work life flow into my leisure time!

 

Last season Matthew Pennington was being lauded for the most interceptions per 90.  Mainly because of Hull’s style of play and the amount of defending they did, not because of how good a defender he is.  He’s a bombscare....although Cotts has just signed him....could be his Adam El-Abd signing.

[Erithacus]

I too feel that statistics in football have become overbearing and an end to itself in many instances. Whilst there are some useful data sets that can influence a teams' performance I think there is an excessive dependance on them.  Wood for the trees?

I can see that stats are easier to collate in, say, cricket or American football due to the stop/start nature and the discrete punctuation that it affords.

However, Association Football is a much more lively and free-flowing game that it cannot be so easily defined. It's one of the great things about it: the very nature demands that players are constantly thinking, processing, reacting and creating all on the move. The best moments are those that inspire and amaze at the merest subtlety and emotions can swing around in a heartbeat. That's why I think that it will be impossible to reduce the game down to simple number-crunching.