Lies, damn lies.....

[Boysownstuff]

Bringing this from comments in the Sunderland match thread so as not to divert that too far off topic.

I quoted a post which showed this tweet:


I wasn't having a go at the poster at all, but not only did I have no idea what it was meant to show but it seemed an example of something which is bugging me - and which I would like to hear others' views on.

My former career was in a public service which became horribly destabilised once academics and theorists who were not practitioners got hold of it and started to introduce targets based on certain statistics. Because the targets set were not always in tune with what needed to be accomplished in dynamic situations, yet the staff were very conscious of criticism if they stayed away from the targets. So they didn't do what the situation actually required and efficiency fell.

That is what I think is happening in football. The abundance of statistics and analysis now available - provided by academics, not footballers - has persuaded modern managers that things like possession, completed passes, take-ons and similar are what really matters. I have no doubt that Monday mornings are filled with one-on-one meetings with analysts who go through performances minutely and look at the numbers. Which creates an atmosphere of fear in the player, who will naturally want the feedback to be positive and ultimately to protect his position in the team by having good scores.

So when the choice, in a game, is to make a telling but difficult pass or to kick it back or square safely he avoids risking a negative number and takes the safe option. It makes Monday's meeting better but slows down the game and does nothing to progress the attack. Moreover, the player who has made the run which the risky pass might have found realises it is not going to happen and so thinks twice before making the run next time. Which makes telling balls even more unlikely in future as there is nobody to pass it to, there is a vicious circle which results in the dull, slow, sideways and back stuff we have got all too used to recently.

We all moan constantly about our lack of creativity, could it be that the geeks are driving it out of the game?

PS - I appreciate the usefulness of statistics in trying to predict outcomes, as in odds-setting, but my point is that they are much less reliable as a tool to try to inhibit or initiate behaviour such as the actual mode of play in fast-moving and unpredictable football matches.

 

[1882]

It's similar (but nothing at all like) the unlikely, but genuine statistic that you could stay unbeaten ALL season, yet still go down... by drawing every game, earning 38 points which would, in many a season, see you fall 2 points short of the 'Magic 40'!
Yet I wouldn't be referring to that crop of players as our Invincibles!

Stats are merely there to pour over smugly when you win, and sift through trying to find a tangible, mathematical formulaic reason why one player had a brain-fart in the 90th minute, and the previous 89 minutes of 'Pro-Opta-obsessed' possession coefficient rating and pass completion go down the drain!

 

[LLB Part Deux]

Well, the caption beneath the stat is correct - we were poor for 70 mins but then improved enough at the end that our win wasn't a complete robbery.

As for stats, I'm biased (lol stats joke) because my job revolves around statisitcal accuracy. The problem is not so much about stats but the improper use of them by people without that insight capability. Traditional methods of analysis simply don't apply to football. I've tried running regression models on player data in the past and it's just not workable in a 'general' sense because there are so many external factors that impact performance. Most statistical understanding comes from scientific models where nearly all variables are measurable and scalable.

Where stats do work is in understanding basic player concepts. It was entirely evident from a bit of analysis on Soldado that a very high % of his goals came from specific opportunity types that Spurs traditionally don't create, and that in his final season before coming to us he scored a higher % of goals from opportunities than he would do in most seasons. That suggested that he was a significant gamble because we were paying for form (outlier) and technical capability at odds with our tactical model (low model co-efficient) and therefore would not be successful unless he could adapt (risk). In short, we shouldn't have bought him at that price.

Fast forward to our recent new expensive signing and a very basic analysis of Son would show you that the majority of his goals come from opportunities where the opposition defence is stretched out of position either through movement or attacks on-the-break. Once again, this is counter to the attacking positions we usually face against compact defences. However, unlike Soldado, Son is more adept at own-chance creation which means he isn't as reliant on the 'team model' for effectiveness. As a result, Son is less of a gamble than Soldado in terms of goalscoring, but still represents a strange choice of signing based on the tactics we've used in the last 2-3 years. He also appears from reports and what I've seen to have a lowish pass-completion / cross effectiveness, which might not impact scoring but might represent more of a defensive risk.

 

[Millbanks]

So when the choice, in a game, is to make a telling but difficult pass or to kick it back or square safely he avoids risking a negative number and takes the safe option. It makes Monday's meeting better but slows down the game and does nothing to progress the attack. Moreover, the player who has made the run which the risky pass might have found realises it is not going to happen and so thinks twice before making the run next time. Which makes telling balls even more unlikely in future as there is nobody to pass it to, there is a vicious circle which results in the dull, slow, sideways and back stuff we have got all too used to recently.

Excellent paragraph.

 

[alexwooty]

Bringing this from comments in the Sunderland match thread so as not to divert that too far off topic.

I quoted a post which showed this tweet:


I wasn't having a go at the poster at all, but not only did I have no idea what it was meant to show but it seemed an example of something which is bugging me - and which I would like to hear others' views on.

My former career was in a public service which became horribly destabilised once academics and theorists who were not practitioners got hold of it and started to introduce targets based on certain statistics. Because the targets set were not always in tune with what needed to be accomplished in dynamic situations, yet the staff were very conscious of criticism if they stayed away from the targets. So they didn't do what the situation actually required and efficiency fell.

That is what I think is happening in football. The abundance of statistics and analysis now available - provided by academics, not footballers - has persuaded modern managers that things like possession, completed passes, take-ons and similar are what really matters. I have no doubt that Monday mornings are filled with one-on-one meetings with analysts who go through performances minutely and look at the numbers. Which creates an atmosphere of fear in the player, who will naturally want the feedback to be positive and ultimately to protect his position in the team by having good scores.

So when the choice, in a game, is to make a telling but difficult pass or to kick it back or square safely he avoids risking a negative number and takes the safe option. It makes Monday's meeting better but slows down the game and does nothing to progress the attack. Moreover, the player who has made the run which the risky pass might have found realises it is not going to happen and so thinks twice before making the run next time. Which makes telling balls even more unlikely in future as there is nobody to pass it to, there is a vicious circle which results in the dull, slow, sideways and back stuff we have got all too used to recently.

We all moan constantly about our lack of creativity, could it be that the geeks are driving it out of the game?

PS - I appreciate the usefulness of statistics in trying to predict outcomes, as in odds-setting, but my point is that they are much less reliable as a tool to try to inhibit or initiate behaviour such as the actual mode of play in fast-moving and unpredictable football matches.

There is nothing wrong per se with using statistics as measures of player performance - as long as the RIGHT statistics are used. To take your example of pass completion - when measured as a simple percentage, this is definitely not the right statistic that should be used, as it is far too simplistic, and as you point out it can lead to players favouring an easy but unproductive pass over a higher risk, higher reward pass. If I were a football analyst/statistician, I would seek to define something a bit deeper, such as forward passes of 5 yards or more completed in the final third of the pitch, or ratio of forward to sideways/backward passes in the final third.

Another example would be assists being used as a measure of creative players' performance. Assists do a terrible job of this: some assists are pin-point through balls and some are sideways passes, or corners. They shouldn't be treated equally. Not to mention that even when a wonderful through ball is played putting someone through on goal, that doesn't mean that they'll score - so even though the creative player has done his job perfectly, he might not be given credit for it.

The difficulty comes in actually getting the statistic right so that it is a true reflection of what you want your players to be achieving. This is a key difference between the statistics that Sky Sports give us and the statistics that top clubs will actually be using. You can be sure that Paul Mitchell's 'black box' won't be looking at pass completion % or assists, but at far more sophisticated metrics. Whereas Sky and MOTD are more than happy to quote these simple statistics, because they're easy to measure and easy for viewers to understand, even if they're actually fairly meaningless. I've no idea what Michael Caley's 'xG' statistic is, but whatever it is I would guess that it's been very carefully thought through.

Any modern business with any sense uses KPIs to measure and drive performance, and football is no different. But the right KPIs have to be used!

 

[Boysownstuff]

Interesting, thanks. I am 100% in favour of using stats to influence player acquisition and indeed for the manager and coaching staff to understand what works best. I think where it fails is in the effect it has on decisions made by individual players during a match. Without being offensive, I doubt that many if any pros are entirely comfortable with trying to understand the complexities of the analysis. What is needed is efficient interpretation of the analysis so it can be translated into instructions which are slightly more detailed than 'run around a bit' - but not too much. And the key might be that raw numbers are not used as a stick with which to beat a player after a game.

So while the coaching staff should know and understand what the numbers mean they insulate the players from them and talk about their performance in a more generalised and easy to understand way? KPIs are a management, not a tactical, tool. You ask them to do things you want for the sake of the overall performance, not just to make numbers better, so you shouldn't give them the numbers.

 

[BUYLEDLEYNEWKNEES]

I think Poch seems to blend other statistics and just general player talent, seems to be able to coach things into players rather than suck life out of them.

 

[MKYid]

I suppose stats go wrong when you look at the stats of a winning team and think that replicating those stats is the key to winning.

 

[Éperons]

The abundance of statistics and analysis now available - provided by academics, not footballers - has persuaded modern managers that things like possession, completed passes, take-ons and similar are what really matters.

I think this is a mischaracterisation.

First, the simple part: possession, at least, was proven decades ago not to correlate with wins (this comes up in Inverting the Triangle). A manager can have a tactic of playing a high-possession game, but, pace Coach Brendan, "winning" the possession game truly is winning absolutely nothing.

Next, wins is what matters. I think your conclusion from this (that players are wary of their own passing stats, for example) reveals the problem in this characterisation. No manager will ever avoid the sack because, though his team got relegated with 0 wins, they lead the league in possession, completed passes, and take-ons.

The problem is that wins (and, especially, goals) are difficult to predict. So a secondary set of data has emerged to try and help show why goals happen, both for and against. And to try and show what players are valuable in making goals happen, both for and against. That's what these statistics show.

Modrić is a perfect example of this. Scouts could tell that he was a vital player, but they had no way of quantifying it, since he had few assists, few goals, etc. But quantifying is "important", because Modrić has to have a price affixed to him. So they kept looking at the models, trying to see if there was something that he was doing that was, actually, special. A video of Comolli posted in the summer's transfer thread shows what they were able to determine.

xG is rather straightforward. IIRC, it considers where you take your shots from and calculates whether they are high-percentage shots or low-percentage shots. And then adds them up to guess about what your goal tally "should have" been. Hence it's already a better indicator of offensive (and defensive!) performance than just considering shots and shots on target, because with xG, not all shots/shots on target are created equal. I don't see anything wrong with that, or anything even mysterious about it. I know in basketball, one talks about "high-percentage shots" all of the time. But that doesn't stop players from taking low-percentage shots.

But it can focus tactics. The xG map shows that, though we were getting skinned often, we didn't give Sunderland a whole lot of looks at the net; Defoe was forced to take low-percentage shots (those turn and blast shots from outside the box that both thrilled us at Spurs when they—rarely—went in and infuriated us because, wait for it, they are low-percentage), and we kept him quiet. On the other hand, they were good at keeping us to shooting from outside the box for the most part, too.

TL;DR: this stat in particular shows you something you already know and feel intuitively from having watched the game (and probably played it) for however many years. It isn't the end of romantic football.

 

[mouse]

So when the choice, in a game, is to make a telling but difficult pass or to kick it back or square safely he avoids risking a negative number and takes the safe option. It makes Monday's meeting better but slows down the game and does nothing to progress the attack. Moreover, the player who has made the run which the risky pass might have found realises it is not going to happen and so thinks twice before making the run next time. Which makes telling balls even more unlikely in future as there is nobody to pass it to, there is a vicious circle which results in the dull, slow, sideways and back stuff we have got all too used to recently.

We all moan constantly about our lack of creativity, could it be that the geeks are driving it out of the game?

Agree entirely, but It's not just the geeks that effect this, its the pressure to perform in general.

A player tries a risky though ball or a bit of flair in a one on one situation and it doesn't come off - then they are a mug for losing possession.

They stay safe and pass the ball sideways then they are a mug for being too pedestrian.

It's lose / lose from their point of view and you can understand why direct criticism from fans can affect a players confidence so much.

 

[LLB Part Deux]

Taken purely statistically, we'd have been idiots not to buy Payet as his chance creation was over 20% higher than any other top league player on the planet last year.

Over 20% higher. Take that in as a stat - usually you'd expect the top players to be within 5% of each other at most, but he created 135 chances compared to the next best who was £55m DeBruyne (112) and ahead of Eden Hazard (101). They weren't bad chances either, as he was 4th on the list of assists (17). He scores goals (7) and takes set-pieces. On the face of it, he's one of the new Gods of football.

So, how did West Ham get him for £10.7m then? Why have France left him out of their last squad? Why did every major club overlook him? Well his pass completion isn't amazing (but reflects his efforts to always make chances rather than keep possession) and his tackling is average at best. Yet on their own, these factors don't detract from 'the most creative player in Europe' so I can only assume that the stats don't paint the whole picture.

But... what if he turns out to be really, really good for West Ham? In that instance have we ignored stats to our detriment? It'll be worth mulling over at the end of the season for sure.

*edit* worth noting that Trippier completed more crosses than any other player in Europe last season... for Burnley *edit*

 

[alexwooty]

Taken purely statistically, we'd have been idiots not to buy Payet as his chance creation was over 20% higher than any other top league player on the planet last year.

Over 20% higher. Take that in as a stat - usually you'd expect the top players to be within 5% of each other at most, but he created 135 chances compared to the next best who was £55m DeBruyne (112) and ahead of Eden Hazard (101). They weren't bad chances either, as he was 4th on the list of assists (17). He scores goals (7) and takes set-pieces. On the face of it, he's one of the new Gods of football.

So, how did West Ham get him for £10.7m then? Why have France left him out of their last squad? Why did every major club overlook him? Well his pass completion isn't amazing (but reflects his efforts to always make chances rather than keep possession) and his tackling is average at best. Yet on their own, these factors don't detract from 'the most creative player in Europe' so I can only assume that the stats don't paint the whole picture.

But... what if he turns out to be really, really good for West Ham? In that instance have we ignored stats to our detriment? It'll be worth mulling over at the end of the season for sure.

*edit* worth noting that Trippier completed more crosses than any other player in Europe last season... for Burnley *edit*

Yes I'm interested to see how he turns out for them...

But remember, absolute numbers of things like chances created, assists and crosses can be misleading when used to compare players, as they don't factor in how much each player played - it's much better to look at these things on a thing-per-90-minute or a number-of-minutes-per-thing basis!

 

[LLB Part Deux]

it's much better to look at these things on a thing-per-90-minute or a number-of-minutes-per-thing basis!

I like those stats when it suits me - like when slagging off Rooney's because his goals per game ratio is worse than nearly every other England striker of note, including Peter Crouch (every 99 mins compared to Rooney's 190 mins or thereabouts).

* Payet got his chances from roughly the same amount of minutes as DeBruyne and a reasonable amount less than Hazard.

 

[Juicy Sushi]

Bringing this from comments in the Sunderland match thread so as not to divert that too far off topic.

I quoted a post which showed this tweet:


I wasn't having a go at the poster at all, but not only did I have no idea what it was meant to show but it seemed an example of something which is bugging me - and which I would like to hear others' views on.

My former career was in a public service which became horribly destabilised once academics and theorists who were not practitioners got hold of it and started to introduce targets based on certain statistics. Because the targets set were not always in tune with what needed to be accomplished in dynamic situations, yet the staff were very conscious of criticism if they stayed away from the targets. So they didn't do what the situation actually required and efficiency fell.

That is what I think is happening in football. The abundance of statistics and analysis now available - provided by academics, not footballers - has persuaded modern managers that things like possession, completed passes, take-ons and similar are what really matters. I have no doubt that Monday mornings are filled with one-on-one meetings with analysts who go through performances minutely and look at the numbers. Which creates an atmosphere of fear in the player, who will naturally want the feedback to be positive and ultimately to protect his position in the team by having good scores.

So when the choice, in a game, is to make a telling but difficult pass or to kick it back or square safely he avoids risking a negative number and takes the safe option. It makes Monday's meeting better but slows down the game and does nothing to progress the attack. Moreover, the player who has made the run which the risky pass might have found realises it is not going to happen and so thinks twice before making the run next time. Which makes telling balls even more unlikely in future as there is nobody to pass it to, there is a vicious circle which results in the dull, slow, sideways and back stuff we have got all too used to recently.

We all moan constantly about our lack of creativity, could it be that the geeks are driving it out of the game?

PS - I appreciate the usefulness of statistics in trying to predict outcomes, as in odds-setting, but my point is that they are much less reliable as a tool to try to inhibit or initiate behaviour such as the actual mode of play in fast-moving and unpredictable football matches.

I think you are assuming something about how the analytics are used, and what they are describing.

If you give players targets in terms of numbers that they must achieve (say shots, or passes or whatever), then it will probably hurt their performance as they'll be trying to count things they do, rather than following the tactics.

But that isn't what analytics are used for, if applied correctly. They generally aren't discussed with players as individual targets. You use them to evaluate the process, not the result.

The post you're showing is a description of the shot quality in the Spurs-Sunderland game. It shows every shot taken by both teams (Mason's goal is the pink square) and shows the location and quality of the shot (bigger squares mean better quality chances).

It's not something of value if you tried to beat players over the head with it. What it does do, is show the quality of the process. Meaning, did our tactics work against Sunderland? Pochettino, Mitchell and the rest of the management can sit down and assess if they were successful in stopping Sunderland from shooting from decent locations, and if Spurs were able to create better chances for themselves. You could then match this up with other metrics, and video to look at specific moments, and video to see times when the tactics were successfully applied, and the team did the right thing, and times when the tactics failed, and then learn from that. You could also then take some of the video clips back to the players and show them things they'd been doing right, and things they'd done wrong, and work on how to fix that.

It's not about counting individual actions and then praising people based on that, since the individual actions don't matter. Pochettino stresses the collective, and it's about whether the players are working together to execute the plan or not. So, to use your example. in the Monday meeting, the safe, back pass would then be pointed out as a problem, since it was a moment the attack failed, rather than saluted as a "good" number.

 

[Allan Nielsens Groin]

How your average forum member read 87% of posts with any kind of reference to statistics:

Make that 89%.

 

[Juicy Sushi]

How your average forum member read 87% of posts with any kind of reference to statistics:

Make that 89%.

That's completely ok. Most fans won't be interested in this sort of stuff, and that's their right.

But some are, and the clubs themselves do use this sort of stuff. Wenger actually mentioned xG in a recent press conference, when explaining their start to the year, and Dortmund did last year when explaining the first half of their season to their fans.

Successful teams find value in this way of looking at the game, so I share this kind of stuff to show how they look at things behind closed doors. Because what they see, and what we as fans and the media see, are very different things.

 

[Allan Nielsens Groin]

That's completely ok. Most fans won't be interested in this sort of stuff, and that's their right.

But some are, and the clubs themselves do use this sort of stuff. Wenger actually mentioned xG in a recent press conference, when explaining their start to the year, and Dortmund did last year when explaining the first half of their season to their fans.

Successful teams find value in this way of looking at the game, so I share this kind of stuff to show how they look at things behind closed doors. Because what they see, and what we as fans and the media see, are very different things.

Thank you. Personally I love them though. And yes football in general and the EPL in particular have been very slow at picking it up. But it's part of the charm of football imo. On the stats front though I have a spreadsheet that predicts the entire league to almost 93% acuracy since I started running it in 92. The last 7% I just make up. Or is it the other way around - anyways...

 

[Juicy Sushi]

And, just to throw further annoyance on the fire:

 

[Cripps14]

And, just to throw further annoyance on the fire:

What the fuck does that even mean?

The problem I see with stats is the way they are used in football. Some people believe them to be the be all and end all of everything, whereas I like to add a decent dose of common sense too. Sadly some people seem to lack this vital ingredient.

 

[Juicy Sushi]

What the fuck does that even mean?

The problem I see with stats is the way they are used in football. Some people believe them to be the be all and end all of everything, whereas I like to add a decent dose of common sense too. Sadly some people seem to lack this vital ingredient.

It means that the difference between the quality of shots we're allowing, and the quality we're taking ourselves is the fourth best in the league, according to the model he has built.

If you brush things away claiming "common sense" when you don't even know what it means, then...