[mantralux]

juvehero, on 12 February 2011 - 02:10 PM, said:

I feel dirty and worthless

I do too.

[latzee]

mantralux, on 12 February 2011 - 10:40 AM, said:

And just for argument's sake: our stats over the last 7 days.

stats.png

Several days worth of work to cater to 0.8% isn't really my idea of efficiency.

Considering that IE users get undelayed meta refresh before Google Analytics loads, to a page not tracked by Google Analytics, that is not all that surprising those 0.8% are only IE9 users who visited the site after IE9 was cleared. All other IE visits simply do not get caught by GA

Try checking server logs with Awstats or something instead, it would give a bit more realistic results, although they will certainly be skewed to favour other browsers simply by the fact that the site won't really attract repeat/prolonged visits by IE users

[TomTwice]

:popcorn:

[mantralux]

latzee, on 12 February 2011 - 02:50 PM, said:

Considering that IE users get undelayed meta refresh before Google Analytics loads, to a page not tracked by Google Analytics, that is not all that surprising those 0.8% are only IE9 users who visited the site after IE9 was cleared. All other IE visits simply do not get caught by GA

Try checking server logs with Awstats or something instead, it would give a bit more realistic results, although they will certainly be skewed to favour other browsers simply by the fact that the site won't really attract repeat/prolonged visits by IE users

Yes they were actually, as it's only just recently I've removed the analytics tracking from the IE page.

[keefa]

juvehero, on 12 February 2011 - 02:10 PM, said:

I run IE8

I feel dirty and worthless

So you should.

[kin_lin]

juvehero, on 12 February 2011 - 02:10 PM, said:

I run IE8

I feel dirty and worthless

Well you are in good company juvehero, I am a Gunner and I run IE8 . . . therefore I must be almost as dirty and worthless as a Liverpudlian :'(

Seriously, I will wait until IE9 has been formally released with the officially released bugs and security issues, instead downloading the Release Candidate version so that MS can benefit from my to debug it for them. Then I will get to see your good work Mantralux. Good luck with your website

[mantralux]

Our post-match tactical analysis of the 2-0 win against Wolves is now up on the site if anyone's interested. http://www.arsenalre...nal-2-0-wolves/

[TomTwice]

I would, but I'm on IE8 here at work, and it's being haughty and refusing to go...

[mantralux]

TomTwice, on 13 February 2011 - 10:28 PM, said:

I would, but I'm on IE8 here at work, and it's being haughty and refusing to go...

Quit that job!

[latzee]

mantralux, on 13 February 2011 - 03:00 PM, said:

Our post-match tactical analysis of the 2-0 win against Wolves is now up on the site if anyone's interested. http://www.arsenalre...nal-2-0-wolves/

As I actaully wathced the game, I'll be critical again as one thing really stands out this is actually completely wrong analysis of Wolves, they certainly didn't play 4-3-1-2. They played bog standard 4-5-1/4-3-3, with Doyle playing a poor man's false 9 role, which is why his position seems a bit deeper than the wingers. Wingers played standard, (relatively) wide roles. Their seemingly central-ish average positions are an artifact of the fact that they often swapped sides. It is pretty obvious, looking at their passing patterns:



So, yes, although the stats were correct (average positions graphs), the interpretation was completely wrong. Like some of us said, everybody makes mistakes

[mantralux]

latzee, on 14 February 2011 - 06:11 PM, said:

As I actaully wathced the game, I'll be critical again as one thing really stands out this is actually completely wrong analysis of Wolves, they certainly didn't play 4-3-1-2.

Just to clear this up; I rarely talk about formations from a defensive approach. Instead, I put much more value in the actual average positions, and I use those average positions to explain a formation in numbers. Many don't do this, as it's pretty standard to explain a formation from a defensive standpoint. I fully understand that swapping wingers will create a warped reality where the average positions end up more narrow, but on the other hand, it is the average position after all - and therefor I call it a 4-3-1-2.

If I would name the formation from a defensive standpoint, I'd definitely say 4-5-1 in this case.

[latzee]

The problem is that average position on the field during the game is not the position players played in. They played clear wingers, both of them spent most time wide. Just because they swapped sides a lot so a average over time ended up centrally, doesn't mean that they played centrally.

Average position is just a tool - like many others it can help analyse the game. It can be useful but it can also be misused, like in this case. The fact is that they played on the wings. Not centrally. Average position is meaningless in this case as it gives the middle value out of two extremes, while that middle value was not really used. And that didn't change in offense or defense. So yes, it was 4-5-1 in defense, but it was more of a 4-3-3 offensively. It was never 4-3-1-2. That's just a meaningless artifact in data produced by wingers swapping sides.

[mantralux]

latzee, on 14 February 2011 - 08:22 PM, said:

The problem is that average position on the field during the game is not the position players played in.

I understand what you're saying, but the defensive positioning isn't the positions the players played in either. The closest we'll get to where players actually were playing, is by looking at average positioning. But, as you say, in this case reality got distorted, due to the constant positional swapping between the wingers.

However, just so there's no misconception: I never said the wingers were playing as two central forwards, I just said that the formation was a 4-3-1-2, which, according to the average positions, it was. Average position is what I go by in terms of formation, even though I realise many don't.

As Doyle was being pushed back, his average position became lower than the average position of an attacking midfielder. But he didn't play as deep as a central midfielder, or a holding midfielder. On his flanks, Hammill and Jarvis were playing wider then the average positions imply, but they were also playing much higher up the pitch than Doyle. Those two factors are the reason for me labelling it as a 4-3-1-2. The numbers are indicating sectors of players on the pitch, in this case 4 defenders, 3 midfielders, 1 attacking midfielder and 2 wingers (who, as the highest playing positions become labelled as forwards). But the forward sector (last number in a formation) doesn't automatically imply the positions are played centrally.

If you describe Wolves' formation as a 4-3-3, you're essentially saying that Doyle, Hammill and Jarvis were at such a similar height on the pitch that they should be regarded as one sector in a formation, and I don't agree at all. Yes, their swapping distorted the average positioning when it comes to how wide they were playing, but it didn't affect the lateral positioning. And if you look at the average positions, you see that Doyle is playing deeper down than Fabregas does for Arsenal, while Hammill and Jarvis were playing higher up than Arshavin and Walcott do for Arsenal.

If Doyle would have been positioned higher up the pitch, I would agree that Wolves' formation was a 4-3-3.

And I agree I could have explained this better in the article, but for me it was still a 4-3-1-2. The only reason I wouldn't describe it as a 4-3-3 is because of Doyle's behaviour, dropping deep and making the wingers the most offensive positions in Wolves' formation.

Quote

Average position is just a tool

I'm with you.

[Millie]

I would have to agree that average position does not actually show the tactical formation, but that's a personal perspective. I know the lads on the Times podcast were also deriding the average position data a couple of weeks ago.

As latzee says, average position shows just where they were as an average of their touches. I actually think that starting positions are a more appropriate way of viewing a team. Average position does not actually show formation. It shows where people averagely touched the ball.

For instance, going on average position from Manchester United v Wolves, Doyle was playing behind O'Hara, which anyone who saw the game would know immediately was bollocks.

Average position is not a great indicator of position within a team system, I would argue. But clearly you feel differently.

[mantralux]

Millie, on 14 February 2011 - 09:49 PM, said:

Average position does not actually show formation. It shows where people averagely touched the ball.

What I'm saying is that 'formation' for me doesn't mean the visual presentation usually seen on TV before kick-off. Those formations are just presenting the viewer with a broad guess at where players will line up in their starting positions. But in my opinion, they rarely show how players will actually behave or position themselves within the context of a match. And I don't think starting positions are at all relevant to analysing a match, as players rarely stay in their starting positions anyways.

In my opinion, 'formation' is better determined by average positions, coupled with actually observing the game of course. But the positional data can reveal things you won't be able to focus on for 90 minutes.

In your example with the Man Utd game, if the average positioning show Doyle playing behind O'Hara, there must be a reason for it. Either Doyle played very deep, or O'Hara played very high. The fact is that O'Hara spent most of his time in a higher position than Doyle, but if you had a different opinion during the match, then it's time to figure out why the average positions are showing something different to what you experienced watching it.

And that's the fun part.

[Millie]

Well, it was clear that Doyle was coming deep to receive some of the higher balls that were sent his way. Also, Doyle played much longer than O'Hara did, and therefore was playing much deeper towards the end of the match.

I completely agree that the TV formation sheets are rough guides where people basically line up compared to each other. In that sense they aren't that reliable. But I think it's also unhelpful to simply say that the average touches helps you to understand the actual formation any better. Unless you can break the average position chart into clearly defined periods, then it becomes difficult to get proper meaning from it, especially if a player switches position during a match or the team take on a different posture.

Movement, in my opinion, is far more appropriate for judging formation than average positioning. Do teams play with a "flat" line at the back on the defence? Where do people run to? Where do they go to pick up the ball? Where do they go to cover once the opposition attack? Who, if anyone, are they marking?

Those are the sorts of things that go into formation, in my opinion. Average position does not show that. And that, to me, doesn't show what a football formation is.

I prefer to use average positioning to explain trends rather than using it as the formation tout court. At the very least, I would want my average position charts broken down into 15 minute blocks.

[mantralux]

Millie, on 14 February 2011 - 11:24 PM, said:

But I think it's also unhelpful to simply say that the average touches helps you to understand the actual formation any better.

I'd argue that average positions are exactly what gives you the actual formation. It shows you how the players actually positioned themselves, which to me is the actual formation. Starting positions are misleading and mostly based on players preferred positions, not their actual movement....but average positions actually shows player behaviour.

Quote

Unless you can break the average position chart into clearly defined periods, then it becomes difficult to get proper meaning from it, especially if a player switches position during a match or the team take on a different posture.

I fully agree. That's why it's important to mix statistics and numbers with actual observations of the game. And hopefully we'll have access to a more detailed data feed from Opta for the 2011/2012 season, at which point you'll be able to select average positions in a timespan.

[latzee]

mantralux, on 14 February 2011 - 11:40 PM, said:

I'd argue that average positions are exactly what gives you the actual formation. It shows you how the players actually positioned themselves, which to me is the actual formation.

Except that in this case it really really really doesn't? Average position graphs show both wingers centrally because that is what you get when you average extreme left and extreme right, but they were by no means positioned centrally. You even call it a very narrow 4-3-1-2 in your analysis, and it was by no means narrow, they both hugged the touchline most of the time (just look at their passing charts). You simply screwed up and read too much into average position stats. It's not the end of the world, we all make mistakes. Just (wo)man up to it, we promise not to gloat too much

[latzee]

Millie, on 14 February 2011 - 09:49 PM, said:

I would have to agree that average position does not actually show the tactical formation, but that's a personal perspective. I know the lads on the Times podcast were also deriding the average position data a couple of weeks ago.

Do I get a prize for harping on this exact issue months ago?

http://forums.fm-bri...post__p__145247

[Millie]

mantralux, on 14 February 2011 - 11:40 PM, said:

I'd argue that average positions are exactly what gives you the actual formation. It shows you how the players actually positioned themselves, which to me is the actual formation. Starting positions are misleading and mostly based on players preferred positions, not their actual movement....but average positions actually shows player behaviour.

Respectfully, no it doesn't. While I agree that starting position in isolation is also a poor indicator (especially for a player like Tevez who starts forward and comes deep; or Ashley Cole who makes a lot of forward runs), I think that arrows on a formation diagram show formation far better than average positioning does.

It also does NOT show where the players actually positioned themselves. It shows an average of where they touched the ball. This is not the same thing. It does not show where they tend to place themselves when they defend and it doesn't show where they tend to place themselves when they attack. Furthermore, if they ever change position they completely skew the data (as has been pointed out in the thread).

For Tevez, for example, I think that in general a player in roughly the "centre forward" position with a backwards arrow tells you far more than the average position chart could. If you wanted statistical "proof" for that observation, you would need both a passing chart and some way of observing where players are when they don't have the ball.

The average position chart shows the average of where a player touched the ball. It is static. It does not show movement - it starts to break down if players swap positions - and it doesn't show what the coach told the player to do or what the player has decided to do.

Average position, like heat maps, passing charts, tackling stats and so much else is a tool - it is not the proof of formation but one piece of evidence amongst many to ascertain player behaviour.

It's like saying that the number of people claiming job seekers' allowance is, absolutely, a figure of unemployment. Of course it gives you a very good idea of the numbers, and can be used as one of the main tools to find out the actual number. But it is not proof sin qua non.

I hate to harp on about Zonal Marking, as I'm pretty sure you've expressed doubts about Michael Cox's methodology before, but I think his description of his formation diagrams provides a better analysis than relying on average position charts:

Quote

Most match reports in national newspapers now feature a ‘diagram’ alongside match reports, demonstrating the positions occupied by the 22 starting players in the relevant match.

Unfortunately, they’re often incorrect – with players in wrong positions or teams in wrong formations – it is not unusual to see a side which lined up 4-3-3 depicted as 4-4-2. This should not be acceptable – if a national newspaper frequently got the name of goalscorers wrong, it would soon lose credibility. The formation played by each team is equally as important (if not more important) to the overall result as the names of goalscorers, so it’s disappointing that such little care is given to the positions of players.

The diagrams uses on this site follow the following rules:

1. They show the players in their ‘actual’ positions on the pitch, generally in the defensive phase of play. Any significant changes when in possession are depicted by movement arrows.

2. The two sides are both displayed across the whole length of the pitch, rather in their own half. This shows their real positions on the pitch, helping to explain which players were up against each other, and which players were generally ‘free’ or in space.

3. The pitches are vertical, rather than horizontal. This makes it easier to imagine how the players will shift when players make natural forward runs, as the play would move ‘vertically’ rather than ‘horizontally’.

4. If a team switches formation halfway through a game, it is the initial starting formation that is used, unless otherwise stated.

5. If a player is substituted before half-time and his replacement plays more minutes of the game, the replacement is shown on the diagram. Otherwise, the starting XIs are shown.

6. It is difficult to represent players who switch position on a diagram (frequently the case with two wingers), so these are simply shown in the position they most regularly occupy, with the switch described in the text.

7. The colours shown are the colours worn by each side in that particular game, unless the teams are hard to differentiate between on the diagram, in which case the colours may be altered slightly.

I've emboldened number 2 as that's the key one. Obviously, number 1 is less helpful, since he is using a subjective reading of the match rather than taking any statistical evidence to accurately position players.

I think that the average position chart obscures what is going on. As latzee has pointed out (I will have to take his word for it - I didn't see the match) the charts have led to a conclusions which appears to defy observation taking in other factors.

Formation is a pretty abstract construction anyway. It's not about where people were "on average", but how they behaved on the pitch relative to their team mates and opponents in the defensive and offensive phases.

It is not static. There is no "one" formation. And for that reason, I think that arrows offer more to the reader and more for the understanding of the match than simply taking an average.

If we have a different definition of formation then that's fine.