The way I'm reading it is us and Spurs are flukey outliers because we have a high completion rate of deep build up but also have a ton of dangerous give aways, because we have a disproportionately high amount of passes and possession in our own dangerous areas (which is dumb). Focus on deep build up without the progression meaning we are just keeping the ball in the deep area and opposition will eventually take it from us and get a big chance.
xG And Analytics Under Amorim
- Thread starter bosnian_red
- Start date
It's not that hard to hire a team of data engineers, data analysts and data scientists, I work with hundreds of them and I'm sure there must be data sources available for purchase.
For data analysis?
We still have too many players who can't cope with pressure, and if we come up against a team that can press really hard, we'll suffer.Ugarte and Bruno seem to be our only midfielders who can actually get past a pressing player, but Ugarte's passing is rather limited and Bruno's can be very inconsistent. Compare that to Newcastle's midfield. Their midfield trio can do all these things, they have the legs to press the opposition for 90 minutes, they can usually dribble past incoming press players and play a quick through ball to a teammate. Until we have at least 2 more midfielders of similar calibre to these three, I don't see us solving our midfield problems, which are largely responsible for these turnovers.Oof
Which all results in big chances for opponents when we cough up a chance:
I don't think getting the data is the problem anymore. Reading it and drawing the right conclusions from it is usually the issue. It's really baffling that it's not a thing at our club, but there are so many areas where we seem to be behind the times. When you look at the fitness levels of our players, I don't think our sports science department and fitness coaches are really up to the demands of modern football. Let's not talk about recruitment, we all know the issues we have there.
Yes, quite a lot out there these days.
You'd have the ones mentioned in this article for a start - https://analyisport.com/insights/how-football-clubs-use-data-to-sign-players/
Plus then there's good old Opta and a lot they can do in house too to produce their own.
Would hope you'd have someone who understands football heading it all up I suppose, although it's still all relatively in it's infancy. You can get jobs in football data/stats analysis with relatively little experience as far as I know. Was listening to someone about a year ago, think he ended up working for Everton after first doing some volunteer stuff with Tranmere. From what I remember he was just a keen amateur originally, a big football fan who wrote stats blogs and had a degree in something like data analytics or applied statistics.
Perhaps the fella from F1 isn't such a bad ideas as it's not an area where people have had 20 or 30 year carrers a lot of the time exclusively in football as it's relatively new. Still, would be benficial if he worked in the game of course.
Whether it is a good idea or not, suppose it would tie in with Ineos using Brailsford. They seem to like this sort of idea of using people from different sports.
Last edited:
From football data analytics, yes. It's not like analytics is a generalist methodology that you can apply seamlessly to anything once you've figured out how it works. I'm not saying someone brought from another field of analytics can't contribute, but there's a transition cost to be sure.
We're not talking about bringing someone in to get us some data. We're talking about putting analytics at the center of recruitment especially. This is how the likes of Brighton and Bournemouth get these players no one's heard about but who turns out to to just slot right into the roles intended for them. It's also how big successful clubs like Liverpool and City operate. That means skilled data guys who know football. Not some random engineer. If we don't have that, we are basically not a serious club, and no wonder we keep fecking up recruitment.
Actually, does anyone know where the club's at in that regard? You don't read much about this.
I'm pretty sure you can find qualified caftards
it really doesn’t matter if the data people ‘know’ football, how the recruitment team want to filter will already be decided and then you just build off that. Starlizard is easily the best example, it’s a sports betting company without anyone with a footballing background, and then Brighton use their database.
The thing I think people misunderstand about data based recruitment though, is it’s a bit pointless without having a relatively consistent style of play. Only then can you break each position down into enough detail to filter it and find alternative players to the best known ones. Otherwise you’re just doing what any club can do now via third parties (and what I suspect we do, pay for data).
Are you kidding? Brighton use their database, ok. But what does it mean that they use their database? That's just data. Analytics is what you do with that data. It's not like you can go on the Starlizard database, ask it who you should get to replace Moises Caicedo and out pops Baleba.
Of course it matters if the data people know football. They're not just providing some input to the recruitment people who handle the decisions. This has to be reciprocal process where the data people have an intimate understanding of what is needed, and where the non-analytics people understand clearly what the analytics tell them and what it doesn't. And this is still a frontier, where the analytics actually need to develop new methods and concepts in order to capture as precisely as possible the skills that are being sought. None of this is possible unless your analytics people also have a strong understanding of football. To work, analytics has to be at the heart of the recruitment process and everyone involved has to engage with it on some level.
You're very right on your second point though. You can't really find those pieces if you lack a clear idea of the puzzle. But I thought that was much of the point behind hiring Amorim.
Last edited:
No Brighton will have specific profiles of player they look for and use Starlizard to identify anyone in those parameters. Then they scout them and keep following them. Starlizard is stats based betting, it’s not full of football guys (hence why you don’t need football guys) so Brighton just define what they want to filter. Same for us - Ineos will already know what they want to do, the new data guy simply builds it. He might never have watched a game of football in his life, it makes no difference.
Suppose it's not only recruitment the data team can have input on, but providing suggestions for tactical tweaks, or persenting data on upcoming opponents etc. Depends club to club, how much a coach is invested in it all, and how much he or others choose to seek out from them or listen to them.
If you use it for a bit of that as well then obviously helps if those people understand football, or at least have someone who can act as a bridge between the coaching and data departments who does.
If you use it for a bit of that as well then obviously helps if those people understand football, or at least have someone who can act as a bridge between the coaching and data departments who does.
Last edited:
I agree, and I attempted to largely say the same the same thing. Must have worded it badly.
The only thing I'm not sure is the flukey bit. Flukey that we don't concede even more goals/chances/shots than we do or something else? I mean we have one of the highest losses of posession in the danger zone on the second graphic, around 29 per game, only Palace worse. We've conceded 361 shots this season (includes all managers), which is 7th best in the league so that's not too bad compared to a lot of our stats.
Could be lucky/flukey that teams don't get shots off more often in those sorts of situations if that's what you meant and would seem reasonable. Might also be a function of our formation, the 3rd central defender in Amorim's setup helping us to prevent shots a little bit more than we otherwise might having given the ball away in our own half so often. Might even be that we have smart defenders who take up good positions when we do, but I doubt that.
Last edited:
In other words, they just get their raw data from Starlizard. The point, and the actual work, is how do they know what they're looking for and what do they do with that data. For this you need football knowledge, and you need the engagement of the whole recruitment setup. Data is not self-explanatory, ready-to-go information. It's what you use for analysis, it's not analysis.
Yes exactly - see original post
Absolutely. Recruitment is just the most crucial and obviously useful area.
I can't believe we're even discussing this, it ought to be obvious by now that this is something you make the fullest possible use of in every pertinent area, and that having the capacity for that is a basic precondition for running a serious football operation.
What do you mean "see original post"? Which original post? We were discussing the wisdom of hiring a racing engineer for data purposes.
the post you replied to:
''it really doesn’t matter if the data people ‘know’ football, how the recruitment team want to filter will already be decided and then you just build off that''
I scanned through the bloke's LinkedIn and it didn't really fill me with confidence. The type of people that work on football data are PhDs in swarm analysis and such, not a rando with some experience visualizing complex data. Obviously some skills are very easily transferable - I'm sure he can look at data from StatsBomb or some other football data platform and make interesting visualizations for recruitment but football data is so much more than that and you really need hardcore math nerds.
The other angle where prior football experience matters is in how you influence coaches and leadership to use your data / recommendations. Football coaches and scouts are notoriously hard to work with for data people. In a sport like F1, the data culture already exists and everyone really gets it. You don't need to convince Lewis Hamilton that poring over tyre wear and performance data from FP1 will help him do a better job in the race, he already gets that. On the other hand, convincing Guardiola to change his tactics based on your what your data says? Much harder job.
https://www.linkedin.com/in/michaelsansoni/
Glazers, woodward like a lot of things.
That's what data analysts and data scientists do - data engineers stitch the data together yes, however, data analysts analyse it and interpret what it means, and data scientists build models to process large amounts of data to predict likely outcome based on past trends.
That's the reason I mentioned all 3 roles. Again, the data sources exist commercially, so we just need the skilled workforce, and as I said, there are loads of them. They're not that hard to hire. And we could take the approach of a few experienced folks along with promising grads to form a team.
We now have an experienced technical director and head of recruitment, so the football knowledge is not lacking.
I reckon they already have a lot in place - Jordan and his team were there for years - but our management team strike me as quite controlling and not people to really trust anyone outside their circle. I wouldn't trust Sir Jim and what he says in that interview for a moment about having 'no data analysis'. Basically, there's no use even having the greatest data team ever assembled if the boss can't understand the work you've done. Expect slick UI and much fanfare and then we announce Benzema and Gaitan on a free.
Re the second part, there was an interesting story on Sky about Klopp being convinced by his data team to take Salah over Draxler (I think it was him). City outsource their data requirements from what Blue Moon tells me, they hired someone much like Jordan around the same time but I have never seen much more on that front. I think their plan was basically Txixi knows exactly who suits the system, let him do his thing and really their recruitment has steadily got worse as time has gone on in my opinion i.e. an old school DoF.
Here's a well timed athletic article on what I meant in more detail if you're interested
https://www.nytimes.com/athletic/62...ampaign=iguk&utm_medium=social&onboarded=true
For what it's worth, I don't think one performance engineer from Mercedes isn't going to cut it and also don't think Jordan is all that
I think no data analysis means truly that -- we have no data and no analysis. Have many anecdotes over the years to suggest that this is true. Unless we start hearing about something more sophisticated than "we scouted 800 full backs before we signed AwB", that will be my position.
Let's see some good research on xT models coming out of the club or collaboration with Google deepmind and then we can talk again
Thanks, yeah am v familiar with the sector. Jordan had a team of like 4 or 5, that's a big data specific team + we have outsourced stuff too working with PhDs in education.
I think you're reading into the qualification too much, all the PhDs listed aren't necessarily better or worse than someone who did a very advanced engineering undergraduate. It's essentially extremely advanced math in many forms (biology, AI, physics, stats) but we don't really know what specifically this new rumoured hire will do. Many funds I work with pay for people to 'do' their math in specific instances as well as contractors, I have no doubt we can also just take this approach instead of hiring full time employees.
As an aside, that article takes quite a liberty with it's interpretation of the Nunez quote. It's trying to be pro data but to suggest Nunez was not a data backed signing based off that quote is about as lazy as journalism gets (it reminds me of what I used to do for essays to up my word count, making quotes fit that you know aren't right but hope the examiner is bored after reading 10 other papers).
I actually went and stalked Jordan a bit to see what he was up to these days (after leaving United) and it looks like he went deep on football data and is building https://twelve.football/ which is quite interesting and maybe he did know what he was doing. It's hard to get a sense of what they were doing at United and how much influence they had. I feel like we would've heard from guys like Crafton and Ornstein if they were influential in the recruitment process.
https://www.linkedin.com/posts/domi...han-almost-activity-7317231138316406784-xdxR/
Anyway, water under the bridge and Jordan got sacked, so it's hard to know what the state of that team is.
And yep I know there's a bunch of outsourcing / contract work here - was listening to an analytics podcast about this traveling data consultant person that helps somewhat random clubs in Africa spin up their data departments, show them the ropes and leaves. The data platforms themselves are quite good at providing insights and not just raw data these days, so that also helps.
All I want for now is something simple like going through StatsBomb OBVs before splunking 50m on a midfielder. All the fancy PhD stuff and new and innovative models can come later. Just validate what the scouts are saying with data and make sure we're not signing duds.
I know two people that work at Starlizard and have met a good few more at the pub. One of my friends works in the football analytics team. He doesn’t work directly with signings for Brighton (I’m not sure if any of them do), his role is much more about creating complex models to track different elements of the game. One thing I will say is that while they aren’t all into football. The people I have met are all massive sports fans in general. It’s like a hive mind of super smart people combining their interests in betting and sport. I imagine in some way that does help contribute to their success in those areas.
Yes, am sure the same can be said for Jordan, Shaw etc. discussed above but they wouldn't fall into the 'football people' idea someone was talking about. As in they love the sport but have no experience in it, other than getting into it via data. The article Teja shared above references it's actually better in the Liverpool guy's case that he came in without a football background, no preconceived ideas about what was important or not, he just thought of ways to break down the masses of data they had.
Starlizard is super interesting, my understanding from speaking to people who use them, though I don't know how I ever will 100% confirm it without you plying your friend with drinks and quizzing him, is Brighton 'pay' for their data (obviously a very loose interpretation of paying for something when you're paying your owner) and have a UI on the football club side that is separate. So if you worked at Starlizard you would never actually see or know what their front end looked like, I guess a smart move given betting is already pretty murky. It's unrealistic to think we'd ever build something like that but, now replying to you @Teja, I agree the first step is surely going to be some kind of in house filtering system that will likely start by simply scraping data from whatever Jordan and his team did + then likely paying for stuff for a few years as we build something in the background.
V interesting on the Africa story, it's a space that is going to balloon in size and is a real business opportunity in my mind. I feel like the back end is going to be impossible to really understand in detail, for me at least, but it would be super interesting to see what the club's actually look at. Didn't Inter Miami genuinely use FM for a while for some obscure latam leagues for example? I reckon that side of it is still way more basic than we realise at most clubs.
Leave this with me! If Utd get through the Lyon game I’ll be watching the semi with one of them so I’ll ask the question. Based on what I understand and the conversations I’ve had, I think your analysis will be pretty close to the reality.
There’s no chance Utd will be able to get a similar gig together to Brighton and it does speak volumes that the only other club I know of with a similar setup are Brentford. Funnily enough their owners used to be in bed together, had a falling out, went their separate ways to what’s become rival betting firms and with all the data that comes alongside that. What a lot of people fail to realise is that the data you get from gambling is huge and basically impossible to replicate in any other format. How else can you get access to huge sample sizes of people that can fundamentally stress test your models. The obvious question is how does that actually link to the football data and analytics used in recruitment (and I imagine beyond that!). What I will say though, is that it clearly does have an impact judging by the success of the recruitment at Brighton and to a lessor degree Brentford.
Thought this one was quite interesting.
Means we're ranked 8th in build up attacks, and that's what we seem to try to do most of the time, 13th for direct attacks. Full season data though, so includes ETH, might be even more skewed towards build up under Amorim alone.
Thing is, we've looked better against teams who attack us at times and the direct ones might well be more effective in those specific games.
You can roughly add up the total attacks off that too:
City - 212 (circa)
Chelsea - 198
Liverpool - 188
Arsenal - 172
Bournemouth - 141
Newcastle - 141
Fulham - 137
Spurs - 137
Brighton - 132
Villa - 132
Forest - 122
United - 122
The rest.
Attacks in this context must be something the lines of how many attacks you put together once you have the ball at the back as teams obviously have more shots than that in a season. Can't include the times you win it high or in midfield.
Amorim specific on that kind of thing, something we really need to improve on when favouring build up play the way we do. Buying a striker or two that lots are crying out for (including me) won't necessarily fix that alone, not unless the problem is purely down to movement of the forwards/strikers we have which could well be too simplistic. It can help some though, probably need players with better vision/better passers/more risk taking and or/ball carrying from midfield/defence when it comes to being able to construct attacks too.
Last edited:
We're 6th in terms of xGA, 10th in xG, and no one in the league bar Southampton underperforms their xG worse than us.
I'd be curious to know the breakdown of this for ETH's part of the season and Ruben's, I do remember Bruno and co squandering loads of chances in the first 7-8 matches to really create a disaster environment.
There are loads of issues with the team, but Ruben has halted the kamikaze football Erik went with and solidified the team somewhat, while he still has to contend with the same blunt attack. We probably need three new attackers, two of whom going straight into the first XI.
I'd be curious to know the breakdown of this for ETH's part of the season and Ruben's, I do remember Bruno and co squandering loads of chances in the first 7-8 matches to really create a disaster environment.
There are loads of issues with the team, but Ruben has halted the kamikaze football Erik went with and solidified the team somewhat, while he still has to contend with the same blunt attack. We probably need three new attackers, two of whom going straight into the first XI.
Please, how can Amorim be blamed when we have attackers, as a whole, do not take their chances. Whether that’s age, inexperience, low IQ is another topic.
Cunha is the first step to reversing this trend.
Cunha is the first step to reversing this trend.
It's worth calling out the xGA stats from that too. We've become reasonably solid at the back even though it doesn't quite feel that way. I'm becoming more optimistic about Amorim because recent data is clearly on an upward trajectory.
Amad getting fit again, Bruno's form in that roaming play maker role and a solid defence is what I hope will be the backbone for our improvement. I really hope they nail the attacking signings this year (before Bruno starts falling off due to age)
I think he'll have only 10 PL games next season to make his mark. If after 10 games this team hasn't improved or has gotten even worse, he'll be gone. It would be unfortunate because we didn't give him a proper timeline to make things easier, but at the end of the day you can only have so many chances in this league. If we're bottom half and struggling it doesn't matter he'll be gone
I notice that Arsenal have an xG that is only 7.3 better than ours, but they've scored 24 more goals. The impact of underperforming your xG vs overperforming it.
Understat has them 15xG better
This one makes me sad. Because it fits with what I'm seeing. I can't think of any other team who consistently gets so little joy from goalkicks taken short. It seems to always end up in aimless passing around the back until someone loses their nerve and hoofs it up the pitch (where we're basically guaranteed to lose possession)
Issue with the CBs first and foremost when you play 3 at the back + want high possession - there's no one good enough on the ball, bar Martinez, so we're automatically a player down to beat the press. None of them can push into midfield to become a passing option, none can really drive forwards with the ball if there are players to be beaten/nearby because they don't have the physicality/speed + few seem to be able to zip balls into good areas without making their longer passes super loopy.
All depends on your tactics/how you train - this is the big limitation of xG and why it's good to look at trends over a season and not really as something to hone in on to make/break an argument.
100%, and this is what we see on the pitch in nearly every game.
I'm not really into this idea of playing so much of our football in our own half. Doing so inevitably leads to mistakes leading to big chances and goals against.
It's a risk worth taking if on average, doing so would lead to more chances and goals for us, that very clearly hasn't been the case though, hence we are 14th.
Actually it shows Arsenal have better finishers than us and also have not been a great attacking side this season as they were last season.