Sporting Intelligence Unit?

[Comrade 86]

9 minutes ago, WestKentRam said:

I'm by no means an expert on this but am happy to write about what I've learned from reading this book so far (started it yesterday and only 1/4 of the way through) as well as a few others on analytics in football. You might regret you asked. Brace yourself for an essay... 😁

I do wonder about the influence of David Clowes on the new appointments, in that perhaps he has used statistical analysis already, thinking back to the pivotal moment in last season of Stevenage away. He kept calm and didn't sack PW despite the discontent among the fan base. Maybe this was more based on stats than just sticking with his man for sentimental reasons? By then we had already 'lost' five points to Bolton and Portsmouth due to refereeing decisions so the table position at the time did lie (a theme in the book), and there were dubious decisions in the Stevenage game where we also conceded two goals directly from free kicks that didn't indicate an overall malaise.

This links into two other examples in the book with regard to managers.

When Klopp left Dortmund to join Liverpool amateur analysts at the time had shown that his last 'poor' season at Dortmund was actually due to bad luck with regard to results of matches versus his team's stats. If this was realised at the time then who knows he might never have joined Liverpool.

With Alan Pardew at Newcastle in 11/12 his team finished 5th in the PL but based on the stats such as goal difference, distribution of goals and net shots on target, they should have finished lower. He was given an 8 year deal on the basis of the 'lucky' season, thinking he'd cracked it, but next season they finished 16th and he didn't last much longer. However the stats for both seasons were actually remarkably similar so he got a higher placing than expected in 'lucky' season 1 but lower in 'unlucky' season 2. If analytics had been used then more rational decisions concerning his contract might have been made.

Another interesting aspect of this is outcome bias, ie the perceived wisdom after the event to make the narrative fit the outcome rather than being a cool analysis of the facts. So season 1 Pardew is praised for attractive play and courage to play youngsters, in season 2 criticised for having a too gung-ho style of play and lack of experience in the team. 

As an aside and potential use of analytics, the current situation of Harry Kane and his playing for England in the Euros could be given. Different opinions from the pundits are he's too old, doesn't press and plays too deep, versus well he scored the goal against Denmark so needs to play. I would think that rather than making gut decisions on his selection then data could be analysed to make one that is more informed, considering variables such as, how many goals against Denmark 'should' England be scoring, does HK in the team affect this by actually reducing the total number of goals scored even allowing for him netting one. Analysis of different games in tournaments, friendlies, high and lower ranked teams, could be made to give some sort of an answer to this.

This sort of information does exist, in that the author mentions Matthew Benham at Brentford and Tony Bloom at Brighton, who both made fortunes from particular forms of gambling on football, enough to enable them to each buy their supported-since-childhood clubs and during their ownership get them promoted to the PL. They didn't do this by making lucky punts on games but adopted large scale analysis using teams of workers, knowing they only had to calculate the outcome of matches more than 50% of the time to make money. A factor included in this was the likely outcome if a particular player was in or out of the team, similar to the HK situation, so hard stats can be used rather than just guesswork.

The book discusses player ratings made by watching a game versus via stats, linked into the cognitive biases involved in scouting and the risks of this. Also, I like mention of the biases of narrative, such as Pep being a genius with regard to tactics and personnel, whereas he took over winning teams that have continued to win post him leaving, and has only coached the most or second most expensive teams in each league. 

There's mention of football being influenced by luck and how good and bad luck in games affects the outcome, but how the game hesitates to recognise this. I had the three Euros games on yesterday and having read this my ears pricked up with the number of times pundits said the word luck without even thinking about it and how this tied into the book.

Analysis of games using xG is discussed, and how this leads into 'the table does lie', both with examples.

It's an evolving relatively new area, but with the ability to analyse aspects such as throw ins, crossing the ball as a goal scoring method, corner kicks, as well as a myriad of other in game possibilities, then a more informed decision on managers, players and tactics can be utilised. That is apart from being used to identify undiscovered gems when signing players as per Moneyball.

An example that I've often scratched my head at, that I assume analytics could be used to give information on, is a free kick towards the end of a game we are chasing from behind. It seems an inordinate length of time is spent with the lining up the wall, discussions between the players as to who is going to take it with bluffing techniques thrown in, jostling in the box with the ref then delaying play to speak to the players concerned, only for the free kick to be blasted into the wall or row Z. With the clock ticking down would it be better to try a different routine with the ball passed short and played on from there. I've seen some examples in perhaps more innovative women's football. Analytics could give an answer as to the success of free kicks taken in particular areas with the personnel at hand, versus time taken to take them and probability of scoring from attacking open play in similar positions.

One reason I really like this move by the club is that we know it will be very hard to gain promotion from the Championship (if that is indeed the aim but then that's another philosophical debate...) given the unlikely-to-be-resolved-anytime-soon influence of parachute payments given to clubs relegated from the PL so they can bounce back up. Trying to gain an advantage using analytics hasn't at present been outlawed by the EFL, so is a smart way to try and progress rather than financial manipulation given the hazards we know that involves.

I'm sure DC in his business dealings doesn't just go with his gut but uses feasibility studies and similar to base decisions on, and I see the use of analytics at Derby being not dissimilar to this. PW as well is a shrewd cookie, and beneath the bobble hat and jokey exterior he's pleased to use this option for the next step of Derby in the Championship. 

Erudite as always WKR, that's a great summary. Lots for folk to digest then. I think the cricket background probably induces some cognitive bias of its own and that's understandable, but for my money, I love the fact that we are looking to gain advantage in ways that traditional thinking alone may overlook. I've cited Benham before, in defence of the use of xG as I think he's super shrewd and 'built' the new Brentford off the back of xG and other forms of analyses, though I had no idea Bloom was a man of a similar school of thought.

Thank you for sharing your thoughts! I'm intrigued to see how this will play out, though the benefits will probably only be known if disclosed by the club. 

[RoyMac5]

17 minutes ago, WestKentRam said:

This links into two other examples in the book with regard to managers.

When Klopp left Dortmund to join Liverpool amateur analysts at the time had shown that his last 'poor' season at Dortmund was actually due to bad luck with regard to results of matches versus his team's stats. If this was realised at the time then who knows he might never have joined Liverpool.

With Alan Pardew at Newcastle in 11/12 his team finished 5th in the PL but based on the stats such as goal difference, distribution of goals and net shots on target, they should have finished lower. He was given an 8 year deal on the basis of the 'lucky' season, thinking he'd cracked it, but next season they finished 16th and he didn't last much longer. However the stats for both seasons were actually remarkably similar so he got a higher placing than expected in 'lucky' season 1 but lower in 'unlucky' season 2. If analytics had been used then more rational decisions concerning his contract might have been made.

Another interesting aspect of this is outcome bias, ie the perceived wisdom after the event to make the narrative fit the outcome rather than being a cool analysis of the facts. So season 1 Pardew is praised for attractive play and courage to play youngsters, in season 2 criticised for having a too gung-ho style of play and lack of experience in the team. 

Outcome bias at work in the two examples you quoted? 😄 

Whilst data analysis undoubtedly plays a big part in most sports there's also lots of 'snake oil' sellers around. Time will tell how these appointments work but I'd have been more impressed with similar appointments of people but with a football background.

[Chester40]

As a 'non-football' man (other than as a fan) I can see why Clowes would welcome some analysis that is based on 'facts' that can be explored and explained, then reviewed for their effect and accuracy later on.

Rather than being bamboozled by 'snake oil salesman' like Harry Redknapp who would dismiss him as a know-nothing with no worth other than signing cheques for players he has a 'gut feeling' for.

In that context I can see why Warne with his emphasis on physical energy and attitude to the team is a good fit.

[WestKentRam]

54 minutes ago, Comrade 86 said:

Erudite as always WKR, that's a great summary. Lots for folk to digest then. I think the cricket background probably induces some cognitive bias of its own and that's understandable, but for my money, I love the fact that we are looking to gain advantage in ways that traditional thinking alone may overlook. I've cited Benham before, in defence of the use of xG as I think he's super shrewd and 'built' the new Brentford off the back of xG and other forms of analyses, though I had no idea Bloom was a man of a similar school of thought.

Thank you for sharing your thoughts! I'm intrigued to see how this will play out, though the benefits will probably only be known if disclosed by the club. 

No problem. Always good for me to have someone to discuss these things with!

The book says Benham and Bloom reportedly used to work together but had a falling out and parted ways. Both then went on to set up similar very successful football gambling enterprises based on analytics.

I see how Derby could use this is by reverse engineering what the gamblers know. If there are aspects of play, that might not be immediately obvious or based on conventional wisdom, that give a team an advantage, then incorporate these into how we play.

 

[Comrade 86]

5 minutes ago, WestKentRam said:

The book says Benham and Bloom reportedly used to work together but had a falling out and parted ways. Both then went on to set up similar very successful football gambling enterprises based on analytics.

Suspect I'll be grabbing a copy myself! Had no idea that there was any link between the two men, but both have developed models that seem to allowed their respective teams to punch well above their weight. Interestingly, Benham 'only has a net worth of £220 million, though that's nearly 3 times the figure of only a few years back and I think this is probably something Clowes has noted, given he too has a fairly modest pile compared to most Championship club owners. We will certainly need to be creative in our thinking, so again, this seems a positive angle to me, though only time will tell.

[WestKentRam]

59 minutes ago, RoyMac5 said:

Outcome bias at work in the two examples you quoted? 😄 

Whilst data analysis undoubtedly plays a big part in most sports there's also lots of 'snake oil' sellers around. Time will tell how these appointments work but I'd have been more impressed with similar appointments of people but with a football background.

Very good 😄 I would argue not exactly outcome bias, but possibly as I suppose most analysis after the event could be prone to. A bit more of a narrative bias maybe. Apparently there are 188 forms of cognitive bias, and I think I use most of them in my reasonings...

Klopp apparently felt he'd done all he could at Dortmund as results had slipped, and it was time to move on. However an independent amateur analyst had shown, as the season was going on, that the poor form was due to results not matching the actual play.

It is supposition as to whether Klopp would have stayed if he'd have known he was doing better than he thought. Perhaps he wanted to move on anyway to a new challenge, so, yes, using the stats to explain why he moved, whereas he might have wanted to go even if they won the league again.

The Liverpool analysts had gone through the data to reassure themselves before appointing him that the poor season wasn't due to bad form. It was news to Klopp when presented with the data at Liverpool. He knew what it felt like on a game to game basis, with games they 'should' have won, but analysts unknown to him had proven this to be the case statistically.

With Pardew it is using stats to show Newcastle over then underperformed. Yes it is done in hindsight given the results at the end of the season, and analysing this years later gives the look of an outcome bias with an explanation fitting the stats, but I would say that's what conclusions are. League tables of where teams should have been placed based on performance stats are produced, sometimes giving rise to anomalies such as this.

A similar theme to Brentford's Table of Justice I believe they used, hence why they stuck with Thomas Frank early doors when he got off to a dreadful start with them

There is mention of the bias of using stats to prove your argument, whereas others that lend weight to a different view are quietly ignored. Similarly using stats based on shakey evidence or not collected properly, without looking at the core data. A good reason to employ someone at the club who knows what they're doing in this field. Roll on the actual football...

.

[KBB]

I will say with regards to scouting and my experience is that some scouts have "biased eyes" ie. They will expect or want a player to be good based on their feelings and instincts. Ai or analysis software doesn't have this bias. It will spit out data with no prejudice or bias. The reports are "clean" so to speak and then the hands on scouting can begin based on desired qualities.

This reminds me of the James bond, new Q quote along the lines of "I can cause more damage on my laptop before my first coffee then you can do in 10 years in the field mr bond"

Scouting, player recruitment and identification has changed and is still evolving. We are moving with the times and looking at new ways. I'm fully behind any new ideas and find it exciting.

[DarkFruitsRam7]

3 hours ago, WestKentRam said:

It was news to Klopp when presented with the data at Liverpool. He knew what it felt like on a game to game basis, with games they 'should' have won, but analysts unknown to him had proven this to be the case statistically.

That shows how much the game has changed in recent years. There’s not a chance any manager would be unaware of the underlying data these days. xG is a thing, @Day, and David Clowes knows it!

Edited June 22 by DarkFruitsRam7

[EtoileSportiveDeDerby]

On 21/06/2024 at 19:42, MadAmster said:

The decision will likely have been David Clowes' to make. The real question is who came up with the idea and then managed to sell it to DC?

Through my work I speak to quite a few people doing AI in many fields One of them worked for a leading retailer in the UK, left his job to head up a data science (AI by any other name these days) team at one of the top 6 clubs.

[DarkFruitsRam7]

6 hours ago, EtoileSportiveDeDerby said:

Through my work I speak to quite a few people doing AI in many fields One of them worked for a leading retailer in the UK, left his job to head up a data science (AI by any other name these days) team at one of the top 6 clubs.

I’d be interested to know how AI is starting to be used in football. Feel like it has the potential to be a game changer given how advanced it’s getting.

[richinspain]

"David, we've been in touch with Arnold Schwarzenegger, he thinks it's a brilliant idea "

"That's fantastic, with the amount of money he's got that will take a lot of the financial burden off of me"

"No David, he's our new number 9!"

[RoyMac5]

Can we not call this the Sporting Intelligence Department please, why Unit!

[RoyMac5]

9 hours ago, EtoileSportiveDeDerby said:

Through my work I speak to quite a few people doing AI in many fields One of them worked for a leading retailer in the UK, left his job to head up a data science (AI by any other name these days) team at one of the top 6 clubs.

Who is our AI expert at the Club? It's a rapidly expanding field, but as can be seen from the PO IT/Fujitsu debacle there needs to be people at the top who understand what they're asking for and how it's implemented.

[EtoileSportiveDeDerby]

3 hours ago, DarkFruitsRam7 said:

I’d be interested to know how AI is starting to be used in football. Feel like it has the potential to be a game changer given how advanced it’s getting.

Where there is data there is AI, simple as that. Player performance, recruitment, fans engagement, computing video analysis are all applied domains but the more interesting one for me is tactics. As I understand it Liverpool were the first and still the best into this, closely followed by Citeh. Not really surprising since those AI folks (the good ones) aren’t cheap. If you head to training ground guru website there are a few articles in there. The scousers signed a partnership with Google DeepMind ( AI division) recently and they published a paper this year, there is stuff on LinkedIn in too. One of the articles on the guru websites ( if I remember correctly) explained how they helped Mane, Firmino and Salah make certain type of runs when counter attacking to each find themselves in the best areas to get the highest statistical chance of scoring.

 

[Mihangel]

13 minutes ago, RoyMac5 said:

Who is our AI expert at the Club? It's a rapidly expanding field, but as can be seen from the PO IT/Fujitsu debacle there needs to be people at the top who understand what they're asking for and how it's implemented.

I'm not sure how you came to a comparison between, essentially, a computerised till and the use of AI in sport but.... Royal Mail has 5,000+ employees and 10,000 franchises - Corporates are all open to this sort of debacle if they don't have the right people and processes in place. Clowes and his senior leaders will be very close to the process - it's the nature of only having a couple of hundred employees  - he doesn't need to be an AI 'expert'. The risk is absolutely minimal compared to an accounting platform that can be corrected by an IT nerd!

[Turk Thrust]

39 minutes ago, RoyMac5 said:

Can we not call this the Sporting Intelligence Department please, why Unit!

Or Football Forensics Section. FFS.

[EtoileSportiveDeDerby]

34 minutes ago, RoyMac5 said:

Who is our AI expert at the Club? It's a rapidly expanding field, but as can be seen from the PO IT/Fujitsu debacle there needs to be people at the top who understand what they're asking for and how it's implemented.

I don’t know. But it sounds like Mo could be leading the charge. The PO IT sh!tshow is a completely different thing, but you are right to a point, there needs to be people who understand how it is implemented and not just at the top. With AI, sometimes the people who implement it struggles themselves to understand the reasoning as the data patterns,  the maths behind it, the scale/volumes and parameters that drive the models are far too big and complex for the human mind.

[RoyMac5]

33 minutes ago, Mihangel said:

I'm not sure how you came to a comparison between, essentially, a computerised till and the use of AI in sport but.... Royal Mail has 5,000+ employees and 10,000 franchises - Corporates are all open to this sort of debacle if they don't have the right people and processes in place. Clowes and his senior leaders will be very close to the process - it's the nature of only having a couple of hundred employees  - he doesn't need to be an AI 'expert'. The risk is absolutely minimal compared to an accounting platform that can be corrected by an IT nerd!

Who do you think made the decisions at the PO about dealing with Fujitsu? The IT ITK had an opinion on them (Fujitsu) decades ago.

As you say you need need the right people in the right places. So who is our AI expert that will be able to relate the use of AI to football and who at the top will understand what to ask for and know if it is being implemented. The numerical risk is less but the risk to the Club is what is important.

Edited June 23 by RoyMac5

[jimtastic56]

26 minutes ago, Turk Thrust said:

Or Football Forensics Section. FFS.

If the football recruitment fails - we could have a real good cricket team on our hands.

[RoyMac5]

3 minutes ago, EtoileSportiveDeDerby said:

I don’t know. But it sounds like Mo could be leading the charge. The PO IT sh!tshow is a completely different thing, but you are right to a point, there needs to be people who understand how it is implemented and not just at the top. With AI, sometimes the people who implement it struggles themselves to understand the reasoning as the data patterns,  the maths behind it, the scale/volumes and parameters that drive the models are far too big and complex for the human mind.

As you are no doubt aware with all IT that is still the case, and was the reason I pointed out probably one of the worst cases in recent years, although by all means throw the NHS systems in there too. 👍

But yes the point was why have them if you don't know how to use them - spend the money on a striker! 😄😉 

Edited June 23 by RoyMac5