Things We Think We Know About Football – July 2013

Some days it’s good to take a step back from the daily deluge of information and try to organize what you think you have learned recently. I’ve been writing regularly on football analytics for about six weeks now (and off and on for seven months), and I’d say half the topics I’ve covered are based on poking around the data, and the other half are riffs on insight from other writers. I certainly feel a lot smarter about the game and about analytics than I did before, but what do I really know? And more generally, what do people in the analytics community think they know about the game that maybe they didn’t a month or two ago?

Crosses Are Bad
There are a couple of references for this that I really like. The first ones are from WillTGM, who looked at crossing generally as well as in the context of Liverpool, last summer. Will’s task was to look at whether Liverpool chose to play a heavy-crossing game tactically the year Kenny Dalglish was fired, and then to investigate whether that type of tactical choice is particularly effective at generating goals. I’m linking to the whole category of pieces he did, because they are really well written, explore a number of different aspects, and should basically be required reading for anyone interested in the topic from an analytics perspective.

Will concludes that open play crossing is extremely difficult (20.5% accuracy vs. nearly 34% from set pieces), and that most teams aren’t very efficient at converting them into goals. Liverpool did indeed choose to cross the ball an awful lot that season, but goals did not pour in as a result. In general, nobody is effective at turning crosses into goals. (I’m shortchanging the work a bit here when I summarize – he covered all sorts of useful stuff. Just go read them.)

theowalcott_crossThe second one is from @footballfactman, looking at Theo Walcott’s crossing, which is deemed a) generally inaccurate and b) surprisingly devastating. Paul digs a little deeper and discovers that most of Walcott’s assists are from short (presumably low) crosses and pullbacks. In fact, from having watched Arsenal constantly, I would guess the majority of Theo’s crosses period have been low and hard, and he had an almost psychic relationship with RVP and that near post run. This is dramatically different from what you would typically get from Golden Balls or Leighton Baines. Because of this style, they are easier for defenders to simply cut out and clear, but they also result in a reasonable amount of goals when completed. It would be interesting to see how many of the cutouts from this type of pass result in possession going right back to the attacking team, since they don’t see very controllable.

The point here is that crossing is really hard. Even Baines, who is one of the best in the Premier League at completing them, only succeeds in finding a teammate 30% of the time.

Headed Shots Are Bad
This one is fascinating because if you do the surface analysis, it looks like this.

heads1_colin

That image is taken directly from Colin Trainor’s seminal examination of the subject , and if you stop at the surface examination you suddenly think “WOW, headers are the way to go. They are just as accurate as ground shots, they are three times less likely to be blocked, 10% less likely to be saved (based on percentage comparison), and 33% more likely to score a goal. Those numbers are massive!”

Not so fast, my friend.

This… this is the area where good, detailed analysis comes into play.

When controlled for location as Colin did, headers are considerably worse at scoring than ground shots in every single spot on the pitch. The only thing that makes headers look so amazing is that they are all taken relatively close to the goal. Meanwhile, ground shots come from everywhere (even though they shouldn’t – that’s a topic for a different day).

Headed Passes Are Bad Too
This one came up as a result of the piece I wrote complaining about how big forwards don’t fit into Arsenal’s general strategy, so why does Wenger keep buying them?!? Anyway, it sparked a bigger discussion that not only deserves its own article, it also indicates a need to change the way we break out and display passing stats. Here’s the quick hit:

Player Ground PS% Ground Pass Head PS% Headed Pass Total PS%
Giroud 77.9% 430 39.3% 234 64
Dzeko 79.5% 517 32.1% 131 69
Ibrahimovic 78.5% 1165 41.4% 111 75
Van Persie 86.4% 831 39.6% 91 80
Lewandowski 78.9% 560 42.3% 123 72
Carroll 81.8% 340 38.6% 321 61

*pause for effect*

Headed passes from forwards are half as likely to be completed as ones on the ground.

There is sense to this. The hits that players take when trying to head the ball would be instant fouls if anyone did the same when they are standing on the ground. Additionally, the quality of the aerial ball itself is so much more important than the quality of a ball played to your feet. A headed pass is a first-time pass, every time. Oh, and unlike a pass along the ground, the trajectory of the ball now exists in three dimensions instead of two. Heading is just really bloody hard.

Andy Carroll has nearly equal amounts of attempted passes with his feet as his head – of course he’s going to look terrible. Maybe the problem isn’t with passing skill with the big forwards, maybe the problem lies with the approach?

Obviously it’s an important part of the game, but heading the ball simply isn’t something you want your offensive players to do a lot of if you can help it.

Passing Bleed Is Bad
I discussed this in brief when I talked about Olivier Giroud at Arsenal, but it deserves application on a broader scale.  Central midfielders pass the ball 50 to 80 times a match. A 5% difference in player completion percentage is 3-4 passes a match. A 10% difference means 6-8 more failed passes in the central midfield, and at that point it really starts to matter. It also has a trickle-down effect into what kind of passes your team has to make to be safe, and what kind they are capable of making to attack (think long, diagonal balls to wide forwards on the counter-attack).

Passing percentages vary widely due to tactical considerations, but if you control for those types of things, you want the best passers possible all over your squad. Choosing a player with a 5% lower success rate that makes up for it in other areas is fine, occasionally. However, 5% worse across an entire team that passes 500 times a match results in 25 more possessions for the opposition, every game.

Lesser passers bleed away possession to the opposition, and eventually that bleeding will lead to goals.

Football is Inherently About Percentages
It’s true, and the game does not give a damn whether people care about this or not, because it is imposed as one of the basic structures of playing football.

I know that’s statty/geeky as hell, but it is a simple, obvious truth. The sooner this is accepted, the sooner people can go about applying the principles to make their teams better.

Conclusion: Strategies that revolve around crossing and heading are hugely inferior strategies.

Look, this isn’t my conclusion, this is what math says.

Crossing is hard.
Heading is hard.
Passing the ball as a header? Also really hard.

So why would any manager choose to do it regularly?

I have two theories on this, from two different angles.

1)      This approach makes sense if it is rare and teams are unprepared for it. If the vast majority of the league plays normally, and your squad employs a physical, aerial approach, teams may be uncomfortable playing that style of football and you have an advantage. However, once a number of teams play this style, counter-strategies (like putting four centerbacks in the lineup across the back line and/or playing all your tall players) come into play that destroy this.

2)      It’s cheaper to play this way than competing with other teams for “normal” players. I might believe this if I didn’t see how much Sam Allardyce and especially Stoke had spent on players in the last five years.

So uh yeah… I don’t think either of my theories make logical sense anymore. Therefore I honestly don’t know why any smart manager, and especially an analytically savvy one, would choose this avenue for their team.

Mixed Strategies Are Not Only Good, They Are Vital
So this is where it gets tricky. Crosses are bad. Balls in the air are bad. Headed shots are inferior to ones on the ground from the same location.

BUT

Image from thetimes.co.uk

Image from thetimes.co.uk

You need to be able to perform all of these at a reasonable level in order to make your opponent respect they are part of the arsenal. You need to threaten from wide to keep the defense from simply packing the box so tightly nothing ever gets through.  You also really need to be able to head the ball well to have a chance at scoring goals from longer free kicks and corners.  So even when they aren’t deployed as a primary avenue of attack, these things need practice, and work, and players who have the skill to turn them into threats.

However, when you bring it back to percentages, they don’t have to be the primary or even the secondary options. You need to enact them just enough that your opponents don’t know what’s coming all the time.

The interesting thing is, when applied in this fashion, the effectiveness of these lower percentage plays often goes up. Teams get so focused on shutting down the central passing and lateral movement, that they overbalance and leave the break to the byline open. Suddenly defenders are scrambling to prevent a free break on your keeper, and the cutback to the penalty spot is completely open.

Central defenders’ legs get tired from chasing speedy guys around for 75 minutes, and they no longer defend aerial crosses to the far post as well. Chicharito/Dzeko have a late-game feeding frenzy.

The short corners you keep taking pull out not just one, but two extra men from the box, and that was the guy that was supposed to mark the player on your team who just got an open header.

So yeah, all those things listed above are bad, but… you can’t just play it on the ground all the time, or your team becomes predictable and easier to stop . Mixed strategies aren’t just recommended, they are vital when it comes to success. This is true whether your team is trying to pass the ball into the back of the net every game, or whether your manager has procured Andy Carroll as the pack horse for their own special brand of hoof ball.

Brain dump, finished.

Class dismissed!

–TK

About these ads

10 responses to “Things We Think We Know About Football – July 2013

  1. Good article, I enjoyed it.

    One problem I have though, is that I’m reasonably certain the vast majority of managers know that crossing and heading from 6 yards out is inefficient as against shooting with your feet from 6 yards out, but it’s very hard to get the shots with one’s feet.

    Often, as you correctly pointed out, teams prioritize defending the center of the field, and the only place on the field that can profitably be attacked is out wide. Then, once the ball is played out wide, the most efficient thing to do probably IS to cross it in, as against the other options.

    Your article seems to be saying that teams are choosing crossing incorrectly against the other options, and I’m not sure that’s true.

    When a central midfielder is looking to play a penetrating pass, let’s imagine (because of the central priority of defenders) that he can put his striker through on goal for a 1v1 with the keeper 5% of the time, while putting his wide player through for a crossing opportunity 80% of the time (numbers completely made up obviously). We all know that the 1v1 will result in goals vastly more often in a vacuum, but is it still true that the central midfielder should only be looking to put his striker through?

    Of course not. If the striker will convert 60% of 1v1s while the team will score on crossing opportunities just 20% of the time (again, numbers made up), than the central midfielder MUST pass the ball wide for a crossing opportunity.

    Your article seems to gloss over the very real fact that it’s much safer in terms of keeping possession to direct the build up play wide and get crosses, than it is to get close shooting opportunities through the middle. This is an extremely important consideration .

    Similarly, it’s obvious that Ibrahimovich will pass better with his feet than his head, but it’s very difficult to get the ball into his feet! If Ibra is 20 yards from goal and central, the other team is likely to defend him in such a way that a simple ball to his feet will be cut out, precisely because he is so much more efficient, shooting and passing, with his feet.

    His team then, may only be left with the option of playing to his head where he can attempt to play a difficult flick on. Again, if a ball played to Ibra’s feet is intercepted 90% of the time, but one played to his head is completed (and allows Ibra to attempt a headed pass to a teammate) 50% of the time, than the ball to his head may be the best option.

    To analogize to basketball, I could give the conversion percentage for layups against the conversion percentage for 18 foot jump shots and ask why even the best teams with the smartest coaches sometimes take jumpers when layups are better, but of course that’s silly. There are shot-clock considerations, defensive considerations, etc. that mean they have to. They would love to get layups all the time, they just can’t. Just like soccer teams would love to get shots on the ground from inside the area all the time, but they can’t so they settle for crosses.

    Basically, I don’t think managers are eschewing analytics and making large tactical errors as much as you haven’t considered a critically important variable.

    Good, thought-provoking article though, and I enjoyed it!

    • Stak,
      Good post. I guess what you are saying is that somehow you need to consider “reality” (i.e. the playing pattern of the opposing team) when setting up your own game plan? The examples of Ibra and basketball were illustrative.

      Though I have to say, and statistics could prove me totally wrong here, that my feeling is that some, successful teams try to avoid the long crosses from outside the box intentionally to be able to create a shorter pass, delivered with higher accuracy.

      Barca play the ball from one side to another, and back again (and back again!), patiently waiting for that second of non-concentration of the defender enabling them to deliver that penetrating pass. Of course they do cross, but they seem to favour possession to get that 1v1 pass through.

      With Bayern it’s not uncommon for Ribery or Robben to try to beat their man and then get into the box in order to create havoc (lure one CB towards them and so create space for Gomes/Mandzukic) – instead of going down to the touchline for the long cross.

      Of course with the best people in your team everything is possible and as I said this is a feeling I got which I cannot back up with statistics.

  2. This post reminded me of an issue I’ve often wondered about: I feel like long-range goal kicks result in turnovers more often than not and it seems like a shorter pass, from keeper to midfielder, would be more productive.

    A long ball from a goal kick seems to favor the opposing team, given that the defenders are taller and facing the oncoming ball.

    Thoughts? I’ve never seen this particular stat tracked.

    • Carson

      Take a look at the following table of info taken from whoscored.

      It shows the ‘top’ 20 players for worst passing accuracy in the Premier League. 16/20 are goalkeepers, and the other 4 are target men strikers, which suggests to me that even the successful passes that hoofball keepers make are then wasted as the target man can’t find a colleague. Really can’t see the logic to that tactic, as it’s not used for variety purposes.

      R Name Team Pos PS%
      1 Mark Bunn Norwich GK 34.2
      2 Adam Federici Reading GK 40.7
      3 Simon Mignolet Sunderland GK 41.4
      4 Julio César Queens Park Ran… GK 42.6
      5 Tim Krul Newcastle United GK 45.7
      6 Brad Guzan Aston Villa GK 48.6
      7 Ben Foster West Bromwich Al… GK 49.3
      8 Mark Schwarzer Fulham GK 50.4
      9 Jussi Jääskeläinen West Ham GK 51
      10 Hugo Lloris Tottenham GK 51.3
      11 Asmir Begovic Stoke GK 53.3
      12 Kenwyne Jones Stoke FW 54.7
      13 Tim Howard Everton GK 56.5
      14 David De Gea Manchester United GK 56.8
      15 Artur Boruc Southampton GK 57.7
      16 Petr Cech Chelsea GK 57.7
      17 Peter Crouch Stoke FW 58.5
      18 Carlton Cole West Ham FW 59.7
      19 Joe Hart Manchester City GK 60.5
      20 Andy Carroll West Ham FW 60.9

      • Btw Beez,couldn’t we get Ted to post more often on TTT? I’ve only seen him there once.

  3. Great analysis, the answer to this maybe less about attack and more about defence.

    I think there is still a fear in football about giving away possesion near your own goal. For this reason a lot of tactics are based around getting the ball away from your own goal. Football managers prefer to see the ball in the oppositions half or in wide ares with the majority of there team behind the ball, or goal side of the ball. This makes sense to a certain extent if you are a poor team, with players who are likely to give the ball away.

    But surely in the premiership it is not that hard to find a group of players who can keep possesion.

    Would love Alan Shearer to read this article. Doubt that he reads much about football in his spare time thouhgh.

  4. Nice article, I must say. But I strongly disagree with the conclusion. Let me try to put it in a very simplistic way.

    Lets define ASoPP as an Attempted Switch of Player Possession.

    Basically, every ASoPP yields a goalscoring opportunity, which may vary from zero proximity to 100 proximity, in percentages. A short pass from the goalkeeper to the right back being something like 0.00001%, while a perfect through-ball being a 99.9%.

    So now graph the different types of ASoPP. First of all, you will find that “bringing the ball into the box” type of passes are most likely to be converted.

    Lets now see how many of the low ground passes are into the box. Not many, we call them trough balls. On the other hand, a huge percentage of the high ground balls are into the box. Which is what we are aiming for. They may have a lesser convertion rate expectation, but are, by a degree, easier to execute.

    So yes – I believe Kenny was right to spam the ball into the box. Its just that he got unlucky(see the woodwork counts for example) and was probly lacking a finisher.

    • Misho,

      I’m not sure I agree or understand what you are saying. I don’t think Ted was advocating to stop use the long cross or the high ball into the box, he was merely saying that using them “at the right time” was a lot more efficient than using it as a general tactic.

      And obviously Kenny wasn’t right in ordering the spam cross attack, since goals were not scored. Whether this was down to bad luck, which I don’t believe in (already Arnold Palmer knew that “the more I train the more lucky I get”), or missing the right personnel in the box is in this perspective beside the point. It didn’t work, so it was a wrong game plan long term.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s