Chelsea-Norwich Passing Analysis (And Introducing Radial Graphs)
by Graham MacAree on Sep 4, 2011 10:16 AM BST
in Statistical Football Analysis
Email
Print
As you may or may not have noticed, We Ain't Got No History has been a little quieter than usual over the last couple of days. No, it's not because the transfer window killed me or anything (although it came close) - I've actually been playing around with a better mechanism for showing passing in specific games than either a flat chalkboard or my initial attempts at a radial passing map. Since it's taken a lot of effort to build this thing, I might as well put it to some use - let's dig into that fun Chelsea vs. Norwich City match from last weekend once more...
Figure 1: Chelsea radial passing maps vs. Norwich City, 8/27/11. Powered by Tableau.
Why yes, that is pretty cool. Go ahead and hover over the question bar for help, because that should explain pretty much all there is to the graph. With this, we can see who everyone was passing to, where they were receiving the ball from, and all sort of interesting goodies. How useful this actually is is anyone's guess - since they take so long to build we don't really have anything else to last Saturday's to, and thus we have no baseline for anything analytical - but there's value in description for description's sake anyway. Plus, really really cool.
Anyway, a few things stand out to me here, those caveats aside...
- Frank Lampard was actually a key playmaker for Chelsea. He was the guy responsible for switching play to the fullbacks, especially keen on spraying the ball out to Jose Bosingwa on the right flank. If Chelsea were to play Lampard a little deeper, he might have some luck launching fast breaks down the wings, especially with Daniel Sturridge's speed on the right.
- Speaking of Jose Bosingwa, he was a key outlet for the midfield. The trio of Lampard, Ramires and John Obi Mikel passed to him 37 times, while the equivalent Florent Malouda-Lampard-Mikel troika managed only 22 passes to Ashley Cole. Bosingwa's re-emergence as an attacking threat who can do his job at the other end of the pitch as well is highly welcoming.
- Mikel passed the ball forwards (i.e. in the forward quandrant) 22 of 56 times. Just saying.
- Despite the difference in who was distributing them the ball, both Didier Drogba and Fernando Torres received it in roughly the same part of the pitch - on average. Drilling down further into the data we see that that was (surprise!) an artifact of the fullbacks supplying Torres high up each wing and the midfielders giving Drogba the ball in the centre. Neither striker did much with their service, though.
- All hail Juan Mata!
Anyway, there's a lot to work with in there and I don't think I've scratched the surface of what we can find in this sort of data. Why don't you give it a play and see what you think? I'm sure there're lots of interesting things to discover. Don't forget to use the filters to find the information you want!
39 comments
|
9 recs |
Do you like this story?
Comments
HOLY CRAP THIS IS THE GREATEST THING EVER
I’m sure you’re already bored to death by the amount of praise going your way, but brace yourself for a whole lot more. You make post-match analysis a lot more fun.
Second
This thing is amazingly cool and actually very helpful in analysing games. Genius.
Have you thought about a job as a performance analyst? Bit more of this and AvB might give you a call…
Well I sure wouldn't mind
But this is more ‘Graham draws stuff’ than performance analysis proper, so I think you’re giving me too much credit :)
by Graham MacAree on Sep 4, 2011 4:31 PM BST up reply actions
Mikel passed the ball forwards 22 times?
Back in college we used to take a shot ever whenever Mikel passed the ball forward! :)
Hilario didn't pass backwards
Not even once.
by Al Benson on Sep 4, 2011 10:43 AM BST up reply actions 1 recs
mikel is very accurate
he just doesn’t always move it as quickly as i would like. he is one of my favorites though
i'm very interested to see what applications you come up with for this
Seems like quite an exciting tool.
Out of interest, did it take up so much of your life because that’s how long it takes to put one of these together; or would you be able to sort one out quicker in future now that you’ve got the framework together and just need to add the data?
I will be able to do one quicker in the future
However, I will never be able to do one quickly. Data collection takes a very long time, but I kind of got in the zone and nerded out.
by Graham MacAree on Sep 4, 2011 4:18 PM BST up reply actions
Interesting - but only part of the story I think
There’s been a lot of criticism of Frank and Mikel concerning sideways and backwards passing and the lack of those cute little passes that catch defenders out.
I think one of the biggest issues with the way Chelsea have played the past year or so is not so much where the passes are going but the speed of movement of the ball and the player.
Frank is a Chelsea legend but I think the view forming that he and Mikel (and Anelka) slow the game down and make us predictable has a lot of truth in it.
No idea how you could capture that in a diagram though :-)
I agree!
There are two clear weaknesses with modelling based purely on the data we have. The first, as you’ve pointed out, is timing. The tempo of some of our midfield players is very slow. Mikel’s more guilty of this than Lampard is, I think, but Frank doesn’t exactly have the quickest release in the world.
The second is accuracy – far too many of Mikel’s (I’m picking on him a lot. Sorry lil’ buddy) don’t go straight to a teammate but force them to cut back, breaking up the tempo there as well.
I could probably find a way of representing tempo if I had access to the data to play with… but I don’t. It’s always good to have a reminder that the tools we’re producing are very limited, so thanks for the comment!
by Graham MacAree on Sep 4, 2011 3:54 PM BST up reply actions
Where do you actually get all of the data from?
Do you record it all yourself just from watching the game over and over, or are you friends with Opta or something?
Would it be of any use to look at the average time between a player receiving the ball and then playing it on? Obviously these durations will change depending on the area of the pitch and whether there are any team-mates in space, but I would have thought that the average would give some sort of insight into who tends to slow the play down.
To put it in context, you might not mind, for example, that only 20% of Lampard’s passes against Norwich went forwards if he was only spending a second or two on the ball and then releasing it to whoever is open, whereas Mikel’s 39% of passes in a forwardly direction might seem less impressive if it takes him 7 or 8 seconds to pick out each pass.
I would have thought it relatively straightforward (although quite time-consuming) to collect this data with a game-tape and a stopwatch.
It might also be interesting to see the average time taken to make a pass in each direction; ideally one would see players looking for forward passes first (and therefore getting them off relatively quickly) and only looking backwards if these are unavailable, meaning that backwards passes take longer because they are a last resort.
The Guardian chalkboards have all of the information
You can just overlay a pixel mapper onto the passes, going through them one by one and pulling out x1 y1 x2 y2 coordinates as well as the passer and the receiver. That’s all the information you really need to build out the angle/distance information.
Unfortunately, that doesn’t give you nearly enough to get tempo out of things, and it also takes forever. Adding the data-collection feature on top of that would be an awful lot of extra work for me, and I doubt I’d be able to do it.
by Graham MacAree on Sep 4, 2011 4:58 PM BST up reply actions
Some things I noticed
- Ivanovic didn’t complete many of his longer (15+) forward passes through the centre of the pitch, but was decent in playing to the fullbacks. Perhaps Luiz would be better playing balls to the midfielders?
- Lukaku and Anelka, on average, played higher up than Torres or Drogba. Maybe the passing sample is skewed because they played fewer minutes, but it seems to imply that we were pressing higher up after the substitutions.
- Mata played more centrally and higher up than Malouda, but still played mostly on the left, despite his roaming. Also, he is very good at passing.
On your points
I think Luiz will just be a better passer than Ivanovic in general – much more of a ball playing centre back than Ivanovic ever will be. But with that comes the risk we’ve seen too often already.
Am I wrong in interpreting that Torres and Drogba only linked up three times during the game?
Yeah
At the moment I guess both Luiz and Ivanovic (based on his last game) give some form of risk in the back line, but I’d rather have that risk and a capable attack than nothing.
And no, you’re not wrong – Graham talked about it in the first analysis which was quickly buried by transfer coverage.
For some reason, despite such a strong defensive showing in pre season (yes, they count for nothing, but still, one goal conceded in seven games…) our centrebacks have looked horrificly off the pace. Perhaps the lack of any real attacking sides meant they got complacent?
Drogba and Torres – I just don’t get it. I get the vision that they work together perfectly in my head, but for some reason in real life it just doesn’t click. I suppose it’s the big-name thing more than anything else.
Definitely the wrong place for this but given our 'experiences' with Zamparini this summer it might be interesting to some to see this from the Guardian:
Palermo owner Maurizio Zamparini:
• February: sacks coach Delio Rossi for “having no balls” and replaces him with Serse Cosmi, “a real man, a man of grit”.
• April: Sacks Cosmi for “fundamental weakness” and reappoints Rossi.
• May: Says Rossi is “like my wife. I want him all for myself.”
• June: Rossi resigns, Zamparini appoints Stefano Pioli and pledges stability: “Piolo’s a great, humble guy. I’ll make a bet now that he will be here with me for three years of this contract, then a further two.”
• Last week: Sacks Piolo before his first match. “It wasn’t working. I want Rossi back.” Rossi: “Absolutely not.”
(Pioli’s verdict when he took the job: “I am sure there will be moments when my opinion and that of the president contrast, but that’s normal. It all seems really positive.” 37: number of coaching changes Zamparini has made since 1987.)
One big reason why we didn’t get Pastore right there ^
You can use the FanPost/FanShot feature for this kind of thing
Zamparini really, really needs a psych evaluation…
The new model is bloody amazing by the way, never seen anything as comprehensive before!
Very helpful in understanding the team dynamics too – noticed, for instance, that Mata was always receiving passes from behind, played the shortest passes of any player and played further forward than Malouda. Also he played more passes to Torres in 10 minutes than Mikel did in the full 70 he was on for.
Wow.
How did you get the angles?
AAARRRRRRRSSSSSHHHHHHHHAAAAAAVVVVVIIIIIIIIIIIINNNNNNNNNN!
Dream goal!
Apparently some people are having trouble loading this
If you are one of them, please let me know so I can try to figure out why!
Orientation of Fields
Great data. A very interesting way to describe passing.
One criticism: for the radial map, up is the opponents’ goal; on the mini-map (for average field position), right is the opponents’ goal. Looking at a winger (Ashley Cole), it is disorienting to see all the passes go right but see his average field position on the “top”. If you have room, I recommend rotating the mini-map counterclockwise 90 degrees to match the orientation of passing data.
Trayton
Fan of: Cardinals, Blues, Sounders, Yellow Jackets, Wolverines, Rams, and Blazers.
Done
Obviously I had originally oriented the mini-map in a more sensible position but it just too up too much space – I’ve figured how to shrink those tables down so it’s back to vertical
by Graham MacAree on Sep 4, 2011 6:53 PM BST up reply actions
I love this, and I think you definitely should be being paid[ a lot] for this.
Is it possible to have a breakdown by passing location? I’m sure it would be hard to implement, but it would help to know where a player is making most of their passes. If the wide players are making forward passes in defence and sideways or backward passes in attack and midfield, they’re doing well, but if those are reversed, they’re a liability.
Again, though, this is a great tool for analysis. It’s great to see confirmation, or not, of the performance I thought I saw on the day.
That is a great idea.
I’ll work on implementing something like that
by Graham MacAree on Sep 4, 2011 8:09 PM BST up reply actions
You know, probably better than any of us, that stats lie.
Thank you for considering adding that capacity! Every new comparison we add helps us better interpret the data. I am afraid, though, that adding too much to this analytical tool will lead to you getting a well-paying job making them somewhere else. [I’m only joking, that would be awesome for you. : ) ]
That is beautiful, sir!
It’s more than I was asking for! I just meant a simple Attacking/Midfield/Defence breakdown, but this is awesome! This tool has made me much happier about buying only Meireles. You can see how effective the Lampard/Ramires/Mikel midfield actually was. As I said at the time, Lampard and Ramires were playing as an attacking double pivot of sorts, and I think this holds that up. Ramires was the more defensive of the pair, yet a huge factor in linking the midfield together, especially in the first half. Lampard was the main creative outlet in the first half, and actually, as he tired in the second half, he let Ramires take over a bit. Mikel never looks good, really, though, but he looks solid here.
Obviously, there are issues with their speed, but Lampard is still healing, and I think Meireles and McEachran can help relieve him as he tires, and take a few games off him, allowing him to heal more completely. They’ve also only had a few games against Premier League opposition, and AVB is still searching for the perfect formation to fit the players.
This tool has given me a great deal of optimism for the coming season. It’s allowed me to see more definitely that our midfield is starting to come together effectively, and that a big signing like Modric or Moutinho would have just made things worse.
Thank you, man. You’ are a legend. [I’m so glad you weren’t born an Arsenal supporter : P]
by Kevin Kostka on Sep 5, 2011 12:00 AM BST up reply actions
Graham this is great.
We are all very lucky that you are doing this. I’l say it again, this is the best Chelsea site out there.
I read the other comments - they were all positive - we must bring balance to the force!
Your mother was a hamster, and your father smelled of elderberries.
Must say this is extremely impressive
How long does it take you to create this report for one game?
PS – Please continue your work, time permitting of course, it is very much appreciated
"Eat Well, Stay Fit, Die Anyway" - Ancient Chinese Proverb
Stellar Tableau work
This is great great Tableau work – you’re becoming a real expert in this, and using it to come up with really innovative ways of parsing the data. We’d love to get you involved in a Tableau User Group/conference sometime – interested? Try me on twitter (@acotgreave)
by Andy Cotgreave on Sep 12, 2011 10:53 AM BST reply actions
Comments For This Post Are Closed
Secondary Sidebar
User Tools
FanPosts
Community blog posts and discussion.
