Are Man City muscling in on Liverpool's analytics crown?

|| We glean a lot from a little as we look at two new job adverts ||

Hello reader! Grab a drink and put your feet up for a few minutes. Get Goalside! is here.


In Captain America: Civil War (stick with me), the MacGuffin that drives the plot towards Cap and Iron Man throwing punches is a proposal: superheroes should register with a government authority. There follows, naturally, disagreement.

The AI-made-flesh character Vision pipes up, pointing out that since Iron Man arrived on the scene, the number of ‘enhanced’ people and the number of aliens wanting to smush Earth to dust had risen exponentially. Maybe the two were linked. “Our very strength,” he says, “invites challenge; challenge incites conflict; conflict,” a thespian melodramatic pause, “breeds catastrophe.”

It wouldn’t be a surprise if, somewhere on the internet, there’s an image of Liverpool FC’s very own Avengers: Michael Edwards and Ian Graham at the head of an unconventional team of ‘enhanced’ minds, their faces superimposed onto Chris Evans and Robert Downey Jr’s bodies. And, like the Marvel team, their very strength has invited challenge.

Manchester City are hiring for their very own best and brightest (see, I told you to stick with me).

To be more precise, City Football Group are hiring. Two job adverts recently went up for a computer scientist and an AI scientist, two roles not dissimilar to ones you’d find in Liverpool’s famed department. You can learn a lot about a club just from adverts like these — linked here and here, for the duration they’re online — so let’s dive in to what we can glean.

Share

The starting point of who’s doing the hiring is important to note. These will be CFG hires specifically. Yes, they’ll be assisting Man City, but they’ll also be helping out the wider group of clubs.

(At present, that group stands at ten: ownership of Manchester City, New York City FC, and Melbourne City, along with investments in Yokohama F. Marinos (Japan), Montevideo City Torque (Uruguay), Girona FC (Spain), Sichuan Jiuniu (China), Mumbai City FC (India), Lommel SK (Belgium), and Troyes (France))

The nature of CFG means that it’s kind of a surprise they haven’t invested more heavily in this kind of brainpower sooner. The job descriptions for the computer and AI scientists state that they’ll be working on things like identifying talent, informing development pathways, helping manage injury, and helping evolve tactical principles. While clubs at different levels may have different needs and situations, a research and analytics team like the one CFG are now hiring for could produce insights that can be used across the board.

And, of course, all paths in the City Football Group lead to The Big Club back in England. If CFG can improve development patterns at Montevideo and Mumbai, that will be in service of the City brand globally and in presenting pathways, possibly, to Manchester itself.

The AI scientist advert is the more interesting of the two. It, and not the computer scientist one, has the point in the job description that reads: “research & develop AI models that will evolve the tactical principles utilised by our teams across CFG”.

I speculatively wondered aloud, only being half-serious, nearly a year ago whether Liverpool’s analytics department could be behind the team’s tactical switch to using full-backs as crucial attacking outlets. Trent Alexander-Arnold being damn good at kicking a football obviously helps, but there were puzzle pieces to put together if you were after a conspiracy theory to latch onto.

How this would work in practice is probably that the analytics/research team would do a project and then findings would be presented to and filter up the hierarchy. At Liverpool, this would be first to head of research Ian Graham, and then maybe sporting director Michael Edwards. We know that Edwards (who reportedly nudged Liverpool to hire Graham) is best buds with Jürgen Klopp — thanks to The Athletic, we know that they play padel together so often that they arranged for a court to be built at the new training complex (yes, the one that the women’s team might or might not be allowed to use). There would seem to be a pretty direct research team -> Graham -> Edwards -> Klopp line of communication at Liverpool men’s team then.

Sadly, we do not have reporting on whether City’s Director of Data Insights & Decision Technology, Brian Prestidge, plays, I don’t know, golf with Guardiola or checkers with Txiki Begiristain. I would imagine, though, that the path from research department to manager is a lot less smooth at Manchester City (or, quite frankly, any club) as it is with Liverpool’s men.

Share

Sign up to the newsletter to get insight in your inbox every Tuesday

Both the AI and the computer scientist roles have this phrase in the job description: “Play a key role in developing our Narrow AI platform to identify talent, inform development pathways and understand & predict the football transfer market”. ‘Narrow AI’ is a specific term we don’t need to worry about*, but what we can all take from this is that CFG have a platform of sorts, that incorporates AI, to do transfer stuff.

To pre-empt any rash headlines though, ‘AI’ is a very easy term to BS about. I’m pretty sure a spell-checker is ‘AI’. CFG are probably using something more sophisticated than Clippy, but it’s not necesarilly HAL 9000 either.

*it just means that it’s focused rather than a broad use of AI

The mention of ‘predict the football transfer market’ is particularly interesting. I think that everyone reading this newsletter will be aware of how analytics could help in identifying talent, and developing player pathways is also something that has been talked about publicly (Memphis Depay used Dutch consultancy SciSports to help him choose where to go after Manchester United).

But ‘predicting the market’ is something else. Is this about finding where the value in the market is going to be in, say, a year’s time? Picking the best moments to buy or to sell? Maybe, particularly in the present landscape, it could be predicting where market values are likely to settle and therefore work out what a ‘good price’ is likely to be.

Another interesting point from the AI scientist description is this: “Drive further development of our machine learning models that enable our Human Performance practitioners to better manage fatigue, injury & illness.” The word of interest here is ‘further’, implying that this is something that the CFG are already doing.

However, the thing that catches my eye the most from these pair of ads is in the one for the computer scientist: “Play a collaborative role within the DIDT team in advancing our computer vision, simulation & reinforcement learning environments.”

‘Computer vision’ is really what it says on the tin: feeding images into a computer for it to ‘see’ things. I may be simplifying things slightly here, but camera-based tracking data works by using computer vision algorithms to ‘see’ where the players and ball are on the pitch.

Traditional data companies like Opta and StatsBomb are also using it to help supplement their data gathering. The latter company have written about their method and its use in helping the accuracy of their location data — and ‘freezeframes of surrounding players — for shots.

Are Manchester City developing similar systems to produce their own data? This wouldn’t necessarily be in creating an in-house Opta, it might simply be to collect metrics that other providers don’t collect, linked to existing data sources. For example, we know that Guardiola is big on counterpressing: you could link a feed of event data to the video to pinpoint moments when a City player loses the ball, and then use a computer vision system to look at how many teammates are in a close enough vicinity to counterpress.

Share

Sign up to the newsletter to get insight in your inbox every Tuesday

There’s a bunch of stuff in the computer scientist job ad that’s functional but important. Lines in the job description about ‘placing the power of our AI models in the hands of stakeholders’, ‘conducting research through applied experimentation and modelling’, ‘productionize our AI-based solutions’. This all seems like ‘you’ll do research, then help us make sure people within CFG can understand it’. Fair.

Meanwhile, the AI scientist role is more of a senior job, requiring more years of experience in a related field than the computer sciencist job and it being desirable to have “previous experience in playing a leadership role within a team.”

The two jobs represent a new stage in a rapid expansion of the team. Over the summer, Ravi Mistry was hired as Football Intelligence Officer and John-Mark Sisman as Performance Physicist (short bios on the Training Ground Guru website here). Sisman presented at the 2018 OptaPro Analytics Forum on pressing styles. I don’t know where I made notes, and didn’t tweet about it, so here’s someone else’s:

With these two new hires added in, CFG’s DIDT department will be pretty bulky, and will cover: physics, computer vision, machine learning, productising the insights, and data communication. Four members will have joined the team (who below to a ‘department group’ called ‘Football Know-how’, that I find too amusing not to mention) since around July.

The challenge to Liverpool’s analytic-minded strength is coming. Yes, it’s coming from one of the richest clubs in the world that also own a stake in nine other clubs across five continents. Yes, that might be a disturbing sign about football’s ecosystem. But yes, this does also mean that there will be a new contender in the Premier League for the biggest collection of brainboxes stuck in a room somewhere who may or may not be listened to by the famous people that everybody else cares about.

It was… inevitable.

Sharing and subscribing helps encourage me to keep writing this newsletter. If you’ve liked this, please consider doing one or the other. Thanks for reading!

Share


Shout-out corner

David Sumpter continued his magic trick of helping free tracking data come to light last week

Or, if you just want the data itself, here’s the link to SkillCorner’s open data set. Also a link to SkillCorner’s website for good measure. (Coincidentally, given the subject of this newsletter, SkillCorner has had a partnership with Liverpool for just over a year)

Elsewhere, Toronto FC’s Devin Pleuler added some more to his analytics handbook, a great resource.

Also, there was a paper published about the frequency and effects of scanning in football too.

Finally, a plug for last Friday’s Mark’s Notebook newsletter, where I looked at how Aston Villa beat Liverpool 7-2


Thanks for reading. Times are, to be honest, rubbish, so look out for people and if you’re able to give some time or money to a local good cause.