Interesting piece on Brentford's moneyball approach

[Gwylan]

Surprised this hasn't been posted already

http://www.theguardian.com/football/blog/2015/feb/22/brentford-mathematical-modelling-denmark

I think it's a fascinating approach and show what happens when you get a mathematical whizz-kid with a love of gambling as a chairman. I wonder which other clubs could boast the same?

 

[Husty]

I have always found it a bit puzzling that Tony and co's incredible mathematical models can consistently return an (albeit small) margin betting on football and yet we can't pick a decent player to buy for Toffee. Where is the sharing on resources?

 

[perseus]

Play well and get relegated

I have advocated Danish attacking training methods before on NSC. That was based on a book about Danish attacking techniques and training methods to achieve this. Before we signed OG and Hyypia, I was in favour of a Danish coach.

However, I do think some teams are over coached, some like Mc Carthy's Ipswich seem over rigid in their shape. Overmathematical % game.

I don't trust the Who Scored player performance ratings as a judge of a good player or not. Certainly a guide and double check but some aspects are not quantified and omitted if not plainly wrong and false. I trust even less the FIFA player ability/skills ratings and some of these appear to be a mistaken ID.

 

[KZNSeagull]

Indeed. It is fascinating.

I have not read Moneyball, but have read Soccernomics which gives the example of Lyon's success in the noughties, which followed a similar sort of approach. Their transfer dealings, for example, almost always meant that they sold a player when they were offered more than they were worth and only signed undervalued players. The key to it was that they already had a replacement in the squad when they sold a player.

Early days I know, but a good example of signing an undervalued player could well turn out to be Kayal. I would imagine that our club uses similar modelling, but is probably not going the whole hog that Brentford want to do. It it finds bargains, then why not.

 

[Biscuit]

Interesting read, thanks for sharing, this section stood out,

Midtjylland’s success is also partly down to their excellent youth academy, which Ankersen helped get off the ground a decade ago: usually five or six starters in their games are home-grown. The emergence of the Danish under-21 international Pione Sisto, who has been linked with many top sides in Europe, is another factor.

I'm probably a stick in the mud but I'm really not keen on the continental structure with a head coach and sporting director - let the manager manage. By all means provide him with an amazing academy, good scouting network and texts at half time but the recruitment side should stay with the manager IMO,

 

[mona]

lunacy sheer lunacy.
Logically, in statistical terms, Baldock and O'Grady should have scored lots of goals this season.

 

[Silverhatch]

Do you think these models include or omit 'emotional' variables such as personality, lifestyle, attitude, physical/mental health, ego, drive etc.

If they are omitted then how do you account for them in the data? If you include them then how do you know if you're comparing eggs with eggs i.e. how can you accurately categorise the differences between so many individual/elite sportsmen?

The root causes for the financial crash of 2007/8 have been traced, in large parts, to the influence of Mathematical models that said the equities and reinsurance markets could not fail... That went well.

 

[father_and_son]

The key bit for me is...
Instead he has developed a deeper understanding about why teams win matches, constructed statistical models and used them to exploit inefficiencies and errors in bookmaker prices.

I can see easily how this drives success in betting. you are exploiting the fact you "know" more than the person you are betting against. But both of you are making judgement calls about a independent event.

I'm not sure that the same insight helps that much when the event is not independent. One team is better than the other and will win the match. Knowing this information upfront doesn't change the result.
The influence of what you say to the team is, IMHO, secondary to how you say it. And what you say to the media is completely inconsequential!

All clubs have analytics being completed continually on their own team and the opposition and this is fed into team selection and tactics. All teams do this before the game. They then measure performance during the game and this helps feed the knowledge base for the next game. However, "knowing" what will happen during the game because your model is very sophisticated isn't a step forward from what other clubs do because you don't have the degree of influence when the game is in motion.

 

[mona]

One wonders how much mathematical modelling was applied in the recruitment of players by David Burke.

 

[BBassic]

It's a very interesting approach. The data is only as good as the person analysing it though; it could easily be a different story if the analyst had been different.

 

[perseus]

Journeyman with a Creative Agent

It's a very interesting approach. The data is only as good as the person analysing it though; it could easily be a different story if the analyst had been different.

True. But the data is partly subjective and sometimes when it is more objective it is just sometimes plain wrong.

 

[Danny-Boy]

Palace were the first team to "go Danish" when they were in the old Div 1 about 1971.

Through a Scottish club (you couldn't import players directly from Europe then under FA regulations) they bought a fullback, Borge Thorup and a forward Per Bartram. Thorup vanished very quickly (christened "Throw-up") whilst Bartram played a few games in Div. 1 but didn't impress.

Later of course Charlton got Alan Simonssen who was a success.

 

[perseus]

"Ankersen won’t reveal everything about the club’s use of data, but does say that Midtjylland pay particular attention to what he calls “dangerous situations” in games. Interestingly, they are a client of E4talent, which tracks shots in the “danger zone” – an area that stretches from the start of the six-yard box to the edge of the penalty area – from which 77% of Premier League goals are scored."

I pay attention to these stats. Against Birmingham we had more chances from inside the penalty area than under Hyypia, probably than Gus and OG as well. So did Birmingham have a few, so we had more shots from within the box than B'ham had altogether. A good sign.

But I thought Birmingham (and Forest) were particularly naff teams. Leeds will be a different kettle. Very much the Danish style, get the ball in the box and then work on what to do on the training ground, with new ideas. The point even ordinary players if they are brave can learn how to do it. Calderon seemed to have the right idea about it. Best looks a possible although his goalscoring record is bad (only one word for it). I think Colunga would have been ideal.

 

[Gwylan]

lunacy sheer lunacy.
Logically, in statistical terms, Baldock and O'Grady should have scored lots of goals this season.

Not necessarily. I think the models that the article talks about are more sophisticated than "X scored goals in the past so will do again". Moneyball is specifically about how that sort of approach fails

Do you think these models include or omit 'emotional' variables such as personality, lifestyle, attitude, physical/mental health, ego, drive etc.

If they are omitted then how do you account for them in the data? If you include them then how do you know if you're comparing eggs with eggs i.e. how can you accurately categorise the differences between so many individual/elite sportsmen?

The root causes for the financial crash of 2007/8 have been traced, in large parts, to the influence of Mathematical models that said the equities and reinsurance markets could not fail... That went well.

I don't know enough about the models to know whether emotional factors play a part but I do know there were plenty of mathematical models that predicted the financial, so don't blame the techies for that one

 

[Rookie]

I would be surprised if we didn't operate a similar model. Bit in the programme on Saturday about the work the analysts do, a lot goes on before the scouting stage

 

[beorhthelm]

...Very much the Danish style, get the ball in the box and then work on what to do on the training ground, with new ideas. The point even ordinary players if they are brave can learn how to do it.

getting the ball in the box is hardly a revolution, what makes this "Danish" style different to English long ball or wing attack, which is essentially the same premise, based on very basic observation that most goals are score from just in front of goal, so get the ball there.

The root causes for the financial crash of 2007/8 have been traced, in large parts, to the influence of Mathematical models that said the equities and reinsurance markets could not fail... That went well.

the issue this highlights is that your mathematical models need to include all available information. the events of 2008 was caused by the realisation that the information was incomplete and therefore the pricing was probably wrong. not saying the model was right (they were almost certainly too simplistic), but the model itself is only as good as the data. economist know this, and there's a saying that the only thing known accurately about the future is the the predictions will be wrong: how wrong is where you make or lose your money.

 

[perseus]

getting the ball in the box is hardly a revolution, what makes this "Danish" style different to English long ball or wing attack, which is essentially the same premise, based on very basic observation that most goals are score from just in front of goal, so get the ball there.



the issue this highlights is that your mathematical models need to include all available information. the events of 2008 was caused by the realisation that the information was incomplete and therefore the pricing was probably wrong. not saying the model was right (they were almost certainly too simplistic), but the model itself is only as good as the data. economist know this, and there's a saying that the only thing known accurately about the future is the the predictions will be wrong: how wrong is where you make or lose your money.

Danish have specially developed training techniques designed to (they say) improve creating chances in the penalty area. I can't say that I understand why they should work. Reminded me of the good Swansea moves in European games which did not seem to work in the PL. It might suit Kemy Agustien's style of play.

Anyrate, the point is we seemed to be getting our chances more in the penalty area since Hughton has been in charge. This may be a good sign?

 

[Triggaaar]

Surprised this hasn't been posted already

http://www.theguardian.com/football/blog/2015/feb/22/brentford-mathematical-modelling-denmark

"Too often in football, the result determines the narrative – for managers, reporters and fans"

Amen.

 

[mona]

Like so many things, this should be a means to an end not an end in itself. I expect Benham will deeply regret forcing Warburton out when Brentford find things harder next season.

I found plenty of interest in Soccernomics but there were some strange ideas that you wouldn't countenance.

 

[Gwylan]

Like so many things, this should be a means to an end not an end in itself. I expect Benham will deeply regret forcing Warburton out when Brentford find things harder next season.

I found plenty of interest in Soccernomics but there were some strange ideas that you wouldn't countenance.

Yes, I do think the interaction in football makes it harder to rely purely on stats - although I agree they should play a part. Cricket is probably a better game for this sort of experiment as it focuses on lots of individual duals.

Warburton will be able to walk into a job in the summer - he has no worries