Top, top lad is Dommy
United appoint Dominic Jordan as director of data science
- Thread starter Tom Van Persie
- Start date
I had no idea that this was a thing in football.
I've watched half of that video. The only way I see that being useful today is to help decide which shape to play vs specific teams and threats etc. But is it really better than just being good tactically in football as there are innumerable variables to try and disect this with data?
How can this really be actionable for a football manager or a player? It's early days and it's probably something we should be investing in for the future to not be left behind should it prove useful but damn, I'd rather we just hired a competent football manager and coaching team than have one that's trying to learn on the job.
This reminds me of when Liverpool kept Brendan Rodgers and changed his coaching staff in an attempt to improve things. He still eventually got sacked after spending alot more money on players anyway.
I see, not necessarily a major addition but another brick in the wall nonetheless. Thanks!
That's right mate, another brick in the wall which will hopefully narrow the gap to City, Leicester and in particular Liverpool who have the most extensive setup in that regard in Europe, from what i've read. Below is also a article from the excellent Training ground guru site which provides detail on the role in question. And the guys at Training Ground Guru frequently have high level people on their podcasts who work at clubs in the EPL and European football which provides fantastic insights on the inner workings at clubs like United, Liverpool, Leicester, City etc.
I'll also throw the below tweet in for anyone that's interested.
A great Data Scientist provides the business guys with the right amount of information necessary to make a decision, helping to alleviate analysis paralysis to some extent.
Data Scientist sounds a bit too trendy for me. I like the term "Decision Support" better. Effective analytics is not only about understanding correlations, it's about identifying the numerous and most important objectives, adequately quantifying them, and producing a solution that maximizes this group of objectives. All in a way that's actionable upon by managers and understood by players.
Indeed.
For example in Liverpool's case, see the below shape they took up when 1-0 up against Spurs in the 87th minute:
You can see how forming into that very narrow central block would force Spurs to play through areas of lower goalscoring potential. And with just three minutes of the game left, taking that shape at that time could be the difference between coming away with three points or one.
I mention this because it's an example brought up on Liverpool's own website when discussing pitch control and the impact of their data science team.
I know what a data scientist does and there purpose, i have two reporting into me. For me the term data scientist fits perfectly because not only do they provide information for decisions based on objectives, they build a stream of analysis starting from a hypothesis, use statistical models and end with an outcome. They then use scenario analysis to derive, prove or disprove an objective. The role is science led, i would only hire someone with a statistical and scientific mind - other people do view this different though.
Analysis paralysis is not the result of a data scientist, its the result of an obsession the receiver has over data. Its good to make decisions on data, but you also need guts and common sense. My point is that Ole has shown me, he has become obsessed with statistics. Reviewing data is both qualitative and quantitative, his version of qualitative is saying the qualties mcfred has is passion desire and energy.
I didn't... say you didn't
Congratulations on having two report to you
So the reason why data scientist doesn't fit exactly, is because a lot of data science today focuses on predictive models i.e. what is the relationship between x and y. Like plots showing trends between xG and average distance of shots from goal. Now that's all useful information, but it's hard for a manager who isn't too data savvy to convert that to actionable information. Now decision models (which incorporate data science models as well as techniques from other fields) come in and say, "now that we have these relationships ironed out, this is how our players should be best positioned to eliminate counterattacks, or exploit weaknesses of a certain opponent..." If you consider that Data Science then good. But that's a gap missing in popular Data Science commentary today.
And yes, part of the traits of a bad data scientist is providing too much information, or irrelevant information, or too indepth information that leaves the decision makers confused as to what route to take.
There's so much data "out there" that's untapped.
There's so much data that isn't yet recorded.
For sure this role has the potential to take us up a few notches
It's gonna take time to answer questions as to why we leak so many goals or give the ball away so much or fail to convert 27 shots
There's so much data that isn't yet recorded.
For sure this role has the potential to take us up a few notches
It's gonna take time to answer questions as to why we leak so many goals or give the ball away so much or fail to convert 27 shots
SAF did it for more than 30 years in top level football management without any of these technologies. This makes me have more respect for him.
Did you write his CV summary?
No, I am Dominic Jordan
Everybody did it that way in those days.
I am quite sure that SAF would employ a data scientist in his team if he would be coaching today, his longevity was only possible because he was able to adapt to changes. He would have adapted this one as well.
Indeed.
For example in Liverpool's case, see the below shape they took up when 1-0 up against Spurs in the 87th minute:
You can see how forming into that very narrow central block would force Spurs to play through areas of lower goalscoring potential. And with just three minutes of the game left, taking that shape at that time could be the difference between coming away with three points or one.
I mention this because it's an example brought up on Liverpool's own website when discussing pitch control and the impact of their data science team.
It's useful but is this anything new that Ole shouldn't know already? Mourinho is renowned for creating shapes which are very compact off the ball (narrow and deep) cutting off any obvious spaces for the opposition.
LVG also used still images ALOT including still images from the training sessions to show players when they make decisions or took up positions against the tactical plan.
I genuinely think Ole isn't as aware of the nuances in tactics at the very top level and the only reason I say that is from the evidence of how poor our shape is during matches.
The biggest example of this from this season would be Bruno playing far too high up the pitch as a second striker which isn't helping us contain the oppositions deepest midfielder when defending and he's just another attacker with his back to goal when we're attacking. Leaving the other two midfielders outnumbered all too often.
That would never happen with LVG or Jose who would use their number 10s for alot more when off the ball like Muller, Sneijder, Oscar, Lingard.
If this is a last ditch appointment to make it work with Ole at the helm then I hope it works. But I won't be holding my breath if we need a data scientist to tell us to be compact when we lose the ball.
I agree with basically all of that, but need to defend Ole just a little bit on this point: Bayern often played something like a real 4-2-4 when Müller played as a 10 (from van Gaal over Heynckes, Pep to Flicks sextuple winners and further...), so he was sometimes even more advanced on the pitch then Bruno regularly is. So the problem is not that Bruno's position is unreasonably high in comparison, but that United is executing it poorly and that United don't have the right players behind that front four.
Bayern is much better organized in pressing than United (where Bruno often seems to lead the press on his own, while the rest of the team is maybe 10m to far at the back), and always relied on a much stronger double pivot. Fred/McT are not terrible, but they are far behind prime Schweinsteiger/Martinez or Kimmich/Goretzka, especially in their attacking contributions.
Adding this on my CV, thanks.
Good point about Bayerns 424 at the time. It can work but if it's not working for us then I'd like to see different ideas. The 424 we end up playing at the moment isn't working. Usually Pogba get's scapegoated for what's happening in midfield but they're outnumbered too often, then the defence is exposed to counter attacks even from poor sides.
I can't think of a 424 that worked in the Premier League unless you count the 1999 United team playing 442 as similar as it had second strikers in Sheringham and Yorke playing.
Will send my Venmo over for 0.5% of the lucrative offer the CV will nab you
We dont or wont play with any patterns now. Is this really going to change how we set up and play? Surely its more a Pep thing to tighten up his passages of play, rather than our lump the ball as far up the pitch that you can. I live in hope.
I'm not going to get involved in discussions about Ole so I cut out the rest of your post, but about this point: I agree. That's a pretty simple idea. So I would suggest that there is a bit more behind the change in shape than the article suggests.
To offer some ideas: maybe the data division had analyzed data on Spurs' patterns of play from previous games, and identified that they were primarily focused on going through the middle - meaning that playing compact like this would be a good defensive tactic here. (But probably not for the entire game. Liverpool use width a lot for their own attacking play, and you don't have time at the brief moment of transition (where nowadays the main opportunities lie) to first regain your width. So you have to maintain some width in defence as well.) That would have allowed Klopp to prepare this particular tactic in pre-game training. Alternatively, the data division may have noticed that Spurs was really focused on the middle in this particular game, and told Klopp so he could make the switch. (Which seems less likely to me, as it's hard to communicate such a specific shape to the entire team in such a hectic phase of a match.)
This is a fair bit of speculation on my side, of course, but it makes more sense for the data angle, and might also explain why they're not sharing all of that narrative. (It's maybe a bit more than they would want to reveal about their methods.)
On the red/blue graph about where opportunities lie (post #242): I think this isn't about in-game management. You can't tell people to change approaches or where to play the ball on the fly, it's too complex. But it's great for devising overall patterns of play or tactics for a particular match. For example, Liverpool's wide play with its full-backs is quite special in world football, and might be the result of opportunities found through this sort of data. Teams might also develop various scenarios of patterns of play that follow what these graphs say about opportunities. Those scenarios might even depend on the next team they're playing, as an advanced algorithm could juxtapose their and your tactics and calculate where there are spaces to exploit.
In that case, I would not be surprised if an element of American Football would enter soccer, in which teams prepare a couple of in-game scenarios and managers (with the help of data scientists) have signs indicating to their team which scenario to use. (Although that would rather apply to broader tactics than individual instances of play, given how quickly soccer cycles through instances of play.)
I'm just saying stuff off the top of my head here and not thinking this through completely, so you can probably poke holes in some of the above. But I do think that with this kind of thinking (except done much better
) is where there are lot of opportunities for data analysts (and probably in many other areas as well).And he did it for 30 years in top level football management competing with other teams and managers who also didn't have any of these technologies. So that's a bit like having more respect for Alexander the Great because he did all he did without having tanks and firearms.
On the general issue; I don't think we really need to understand exactly what Jordan's role and function will be in order to acknowledge that this seems a shrewd investment. Analytics are proving to be a great tool in all kinds of areas, football (and team sports generally) is among the ones most obviously suited to what it can offer and United have not been among the teams who were early leaders in this field, and need to catch up.
Last edited:
Data Science along with many other things help the overall team output. It's a percentage of the whole and it's good we have invested in this.
But it's still a game with eleven players trying to put the ball in the back of the net. A lot of that is still instinct and individual skill.
Put great science and great skill/intuition together and you will get great outputs.
But it's still a game with eleven players trying to put the ball in the back of the net. A lot of that is still instinct and individual skill.
Put great science and great skill/intuition together and you will get great outputs.
The concerning thing for me is why are we only appointing this guy now? Other clubs have had these analysts in place for years. Even a club like Brentford who are just newly promoted seem to be at the forefront of this stuff. I remember when Utd used to be innovators in the game, now we’re just painfully lagging behind. Sad times. Better late than never I suppose.
It sounds like we started improving our data science capabilities since Moyes came in. We just went and got a whizz to lead and streamline the dept. It can only be a good thing.Relatively cheap hire and could have a very positive knock on.
Last edited:
I don’t know why people think that we are just now getting a sports science department. I read somewhere that fergie sold stam as the sports science guys told him that he was on the downward slide as the data told them so. We have had this department since the fergie days.
Spot on mate. It was actually Moyes who got the ball rolling as far as modernising the club.
As someone who is currently in the data science industry (albeit not for a sport's organisation) I feel too often data scientists (or the teams that I oversee at least) are keen to tell the entire story with quantitative data, regarding it as the gospel truth for strategic decision making. But one can learn a lot through qualitative aspects and interviews ...eg. with players/staff. It is a delicate bridging exercise between analytics and qual data.
But but... pattern of play and tactics don't matter.Indeed.
For example in Liverpool's case, see the below shape they took up when 1-0 up against Spurs in the 87th minute:
You can see how forming into that very narrow central block would force Spurs to play through areas of lower goalscoring potential. And with just three minutes of the game left, taking that shape at that time could be the difference between coming away with three points or one.
I mention this because it's an example brought up on Liverpool's own website when discussing pitch control and the impact of their data science team.
What a moronic post.
Sport science and data science applied to sports are completely different things. I can pretty much guarantee that whoever adviced Fergie was not using machine learning on large data, and almost guarantee that they were not using (advanced) statistical models.
Heck, even as something revolutionary as moneyball in baseball (2003, so a couple of years after Stam was sold) was just logistic regression in small data, something that is medieval by current standards.
I guarantee that Liverpool with all their fancy Data Science etc, will be back in wilderness when Klopp is gone. All these things are just fancy tools.
And Pep, even with only chalk and board on hand, will still be dominating the League because he's an excellent manager/coach and can spend billions.
And Pep, even with only chalk and board on hand, will still be dominating the League because he's an excellent manager/coach and can spend billions.
Last edited:
The data is valuable and can be used as a feedback but if the manager has a certain philosophy and the feedbacks override that philosophy and or preferred playstyle then those feedbacks will not be used. I'm fairly certain of that.
Agreed. Those articles that are using Liverpool and City as examples of data science success, massively excluding (on purpose) the main point of their recent success.
This. Perfect storm : finding all the things that just worked seems a bit simplistic. Bit of luck. Lot of hard work, trial and error and being totally right in the right place at the right time. Just because Liverpool got it right for a few years doesn’t mean they’ve sown it up. Get even better data people. More suitable high potential players. Keep doing the right thing at the very top end of the market until it all comes together. At least we are on the right path and seeing the different angles. We used to be cutting edge with Fergie’s scouting and man management. We can get back to that once we find a fresh angle and keep investing. Do our best to streamline the hiring process. Get the best across the board. Every little tweak hopefully helps get closer to that again.
What everyone was using 15-20 years back will be medieval by today’s standards. But to say we never had this department at all is either being naive at best or willfully ignorant at worst.
The sports scientist back then also used data. Might not have been the models and the metrics which we use now. Nonetheless they were analyzing data and coming to decisions.
The point here, I think, is that sports scientists are more like medical professionals that look at the athletic and bodily aspects, while the data scientists we're talking about here, are rather looking at the way the game is being played
So what i wanted to say was that even back in the day we were using data to analyze player performance and take decisions accordingly. Now we are looking at game performance and improving it based on data. This appointment was made to lead the data science team. Why people on here are saying we never had a data science team prior to this appointment makes no sense to me.
Yeah, it's definitely not a new group that's being created now. I would say it rather looks like this group is gaining in weight by creating this Director position to lead it. But I don't know if these people would work with the sports scientists, so this data work might be a separate development from the use of sports science earlier on. That's speculation until someone provide some article on the subject though.
Says the article. I think they mean 'more complex' but anyway it's a much harder task to predict football than stocks. Stocks go up or down numerically and there are a number of known numerical inputs likely to influence this. What are the football equivalents? Should we sign Luke Shaw or not? What are the relevant inputs? Will we score?
Now it might be possible to use Data Science to answer these types of questions but it's much harder. Data Science has a tendency to be over sold.
Tbh the credentials of dom is a bit underwhelming as compared to the data guys at city and liverpool