It’s time to offer up another revised version of my Possession with Purpose Analysis.
My intent here is to:
- Provide an update that may help simplify this effort, and
- Update new links to articles most have found to be of great interest in the last year.
To begin… Possession with Purpose (PWP):
The End State, as always this is good to know up front:
Create an objective Strategic Family of Indices, with publicly made available data, that has relevance and helps identify (explain) the strengths and weaknesses of team performance ‘outside’ the realm of Points in the League Table.
Of note; this analysis has been presented, and received with great interest, at the World Conference on Science and Soccer of 2014. So it’s not a fly-by-night attempt to offer up analysis that can’t translate back to the soccer and science industry or help inform the general, or well educated, soccer community (both here and across the pond) about Footy…
Create a Family of Indices that measure the ‘bell curve’ of strategic activities that occur in a game of football (soccer); recognizing that in order to score goals the following activities usually need to occur:
- Gain possession of the ball
- Move the ball
- Penetrate the opponents defending final third
- Generate a shot taken
- That ends up on target and,
- Gets past the keeper
From a statistical (measurement) standpoint those activities are organized into these six categories:
- Possession percentage
- Passing Accuracy across the Entire Pitch
- Passing Percentage within and into the Opponents Final Third compared to overall possession (i.e. = Penetration)
- Shots Taken per Percentage of Penetration
- Shots on Goal per Shots Taken
- Goals Scored per Shots on Goal
It’s not a secret formula but I do retain Copyright.
The Family of Strategic Indices – there are three of them:
- Attacking Possession with Purpose (APWP)
- Defending Possession with Purpose (DPWP)
- Composite Possession with Purpose (CPWP)
APWP Index: How effective a team is in performing those six process steps throughout the course of a game. Example:
DPWP Index: How effective the opponent is in performing those six process steps, throughout the course of a game, against you. Example:
CPWP Index: The mathematical difference between the APWP Index and DPWP Index. Example:
Simply stated, the analysis stemming from this effort is a comparison and contrast between how a team performs (in the bell curve of these activities) relative to other teams in their league “without” including points in the league table.
Last year the CPWP Strategic Index Correlation (relationship) to Points in the Table, for Major League Soccer, was .77; this year, at the end Week 26, the R is .85.
In returning back to the End State:
“Create an objective Strategic Family of Indices, with publicly made available data, that has relevance and helps identify the strengths and weaknesses of team performance ‘outside’ the realm of Points in the League Table.”
Given the very high level of Correlation these Indices have, I’d say this Family of Indices has considerable statistical relevance; and I should point out that although the PWP approach is an Explanatory Model it can also be leveraged as a Predictability Model.
After speaking with a number of folks at the World Conference on Science and Soccer (2014) it was agreed that the most effective way to turn this into a Predictability Model is to remove Goals Scored (in both Indices) and ‘see’ how the Composite Index takes shape after that.
Here’s an example of what I mean:
A word or two of caution…
From a purely statistical viewpoint I do not see this as a Predictability Model that has direct relevance yet… why?
For the simple reason that there have not been 15 games played for all teams both Home and Away – teams show a tendency, for the most part to behave slightly different at home versus on the road…
Why the number 15? I suppose it comes down to Confidence Level in the number of samples that are needed in order to forecast the future based upon the past… with 34 games played in Major League Soccer you really need 15 games to reach that 95% Confidence Level limit in samples…
All that said, it is extremely inviting/inticing to see that even when Goals Scored (both for and against) are removed the CPWP Predictability Index still has a correlation (R) of .84…
Links to articles that have had extensive views over the last year and a way to get a taste of how PWP analyses might be able to help you, as a writer (through collaboration with me), better inform your audience about the nuance of soccer:
- Chicago Fire
- Portland Timbers
- Consistency of Purpose – Attacking Standard Deviations
- La Liga – Simana 2 ( I can offer translation of my articles from English to Spanish on special request)
- Bundesliga – Bayer Leverkusen (I can offer that translation request to German as well, on special request)
- English Premier League – Chelsea
- Colorado Rapids
- LA Galaxy
- Sometimes what doesn’t happen on the pitch has more value than what does happen
- New Statistics – Open Shots – Open Passes
- FIFA World Rankings – Time for a change?
- Expected Wins
- Passing – an oddity in how it’s measured (Part I)
- Passing – an oddity in how it’s measured (Part II)
- Expected Wins 3 – My deepest dive yet into the average performance of what winning teams do in Major League Soccer, the English Premier League, Bundesliga, LaLiga, and World Cup 2014.
- My original Introduction and Explanations (detailed) to Possession with Purpose Family of Indices
- 2014 End of Season Analysis – Houston Dynamo – Dynamic Dynamo Demagnetized as Dominic Departs
- West Ham and Aston Villa – EPL– Going in two different directions
- 2014 End of Season Analysis – Chicago Fire – Candle Burned at Both Ends
- Getting Better as a Youth Soccer Coach
- The Comforts of Home in Major League Soccer
- Seattle Sounders – Road Warriors in 2014 MLS Regular Season
- Portland Timbers End of Season 2014 – Defense Wins Games & Better Defending Leads to Better Attacking
- Valencia – Formula Won – La Liga
- Getting More From Less – Peeling back the statistical differences between teams that Direct Attack versus playing a Counter-Attacking Tactic as part of a Possession-based System.
- Expected Wins 4 – Is European Football Really Higher Quality than Major League Soccer?
- Seeing Red!!! Toronto FC
- World Cup – Two Best Teams? You Bet!
- UEFA Champions League – Some Great Games Coming
- Busting the Myth of Moneyball in Soccer Statistics…
- Scintillating Saints of Southampton Stay Strong
- Hurried Passes
- Catching up with Europe (CPWP and initial discussions on TSR)
- Redefining and Modernizing TSR
- Expected Wins Five (Europe – Pucker Time is here)
- Passing – More from Less – Barcley’s Premier League after Week 30
- MLS 2015 – Control or Lack Thereof
Others in mainstream media sometimes offer up subjective opinions that may not be substantiated with objective data; I won’t do that.
Every shred of analysis offered here will include some sort of objective data to support an opinion or conclusion.
Like any other mainstream business; statistical analysis provides objective data as a tool to leverage when looking to make business decisions. It is not a substitute for the seasoned leadership needed to make final decisions.
I don’t advocate that this analysis is the ‘answer’ or the only tool that substantiates one view – in a soccer match, with 40,000 supporters in attendance, I’ve learned that those 40,000 supporters have 40,000 sets of eyes that see things differently.
On this site, this information and analyses presented, is merely my view, from my eyes, in how I see the game – hopefully, in order to make my future articles of better value, others will add their comments, thoughts, and questions.
Finally, I’m not sure how this will develop but I’ve been approached to provide a manuscript for this analytical effort – for publication in a Sports Science Journal. More to follow on how that goes.
COPYRIGHT, All Rights Reserved. PWP – Trademark
NOTE: All data used to generate this analysis stems from OPTA through a number of open/public websites across Europe and America.
My thanks to OPTA and all those open websites for helping to facilitate my own analysis and potential improvements that may arise from this effort.
While most were probably focused on some other battles this past weekend – and rightly so in some cases – the Timbers might just have shaken the Western Conference a wee bit to reinforce, that when they get their defense right, they will be a team to reckon with.
Before diving in though; here’s a link to my pre-match thoughts on all the games this weekend; some thoughts are smack on – while some are way off target; so it goes.
Back to the Timbers.
I don’t offer this lightly, for almost 80% of this season the Timbers defense has been downright deplorable (just three clean sheets) last year they had 10 clean sheets after 25 games.
Only now – with a major shakeup in the back-four, after that resounding Sounders smack-down, have the Timbers acknowledged that defense is first and played like it!
The star of the match, and I don’t do this often since team is always first, was a young lad by the name of Alvas Powell – here’s a great picture of him post game with the ever present, and highly entertaining Pa Madou Kah, in the background – picture courtesy of Little Imp (@stretchiegirl)
So before digging into some specific statistics about the Timbers here’s a link to my post-match article, about that game, and then the Composite PWP Strategic Index for Major League Soccer after 25 weeks:
One other technical detail that’s probably new for many – the yellow stars indicate which teams have already sacked their manager this year.
I’ll offer up a reminder a bit later on all the stars present at the end of last year.
And if you are interested in some details about why Toronto FC sacked Ryan Nelson – I’ve included this article published by MLSSoccer.com for your reading pleasure.
To summarize, based upon what I took away from the article, Ryan Nelson was sacked due to poor team performance. I’m not sure what that means to the Toronto front office but it’s meaning (could?) be intuited based upon this Index. I’ll leave that for others to decide.
So now on to overall team performance:
LA Galaxy, Seattle Sounders, and Sporting KC continue to lead the overall CPWP Index – others moving up or staying put in the top half include Columbus, DC United, Portland Timbers, FC Dallas, New England, Real Salt Lake; while New York, Colorado, and Vancouver took slight dips this week.
On the outside, looking in, the list is much shorter. Of note to me, is that only two of those teams performing on the trailing end are Western Conference teams.
Can some conclusions be drawn from that? Perhaps – but I’ll save those thoughts for when the season is completed.
Attacking PWP Strategic Index:
For the statistical types; the R2 between the APWP Strategic Index and Points in the League Table is .74 – that’s also pretty good.
Leading the league are the LA Galaxy (no surprise I’d expect). On the tail end there’s Chivas and the ever shocking Dynamo, especially for some, after beating Sporting KC this weekend. Somehow I don’t think Houston is entirely out of the Playoff picture.
With respect to Portland they are sixth best in possession percentage, passing accuracy within the final third, and goals scored per shots on goal – pretty consistent in three critical attacking indicators.
With regards to overall passing accuracy they are in the top ten at 8th best. When converting possession to penetration they are also 8th best – and in shots on goal per shots taken they are 7th best.
In looking at shots taken, per penetrating possession, (a percentage number usually better when lower than higher to infer patience) they are 11th best.
So all told, in attack, they are very consistent, and good, compared to others.
Their downfall has come in Defending PWP – here’s how the teams stack in that Strategic Index after Week 25:
For the statistical folks the DPWP Strategic Index R2 is -.66 – again pretty good but there is a tricky quirk about defending.
There remains a challenge in measuring what doesn’t happen (for the attacking team) based upon how the defense plays.
In other words some positional activities that the defense executes are never measured – what gets measured are actual events as opposed to non-events; i.e tackles, interceptions, clearances, etc…
One of my recent articles was published with the intent to push professional soccer statistical companies to begin tracking and differentiating between Open Passes and Hindered Passes, as well as Open Shots and Hindered Shots, to help measure what doesn’t happen.
“Well an attacking player decides he can’t make a pass to a player in a forward position because the defender has the passing lane closed (hindered) – so the attacker passes elsewhere (an open pass that is unopposed).
In counting the number of Open Passes versus Hindered Passes statistical types can begin to plot maps on what areas the defense is inclined to leave open (cede) versus what areas they are inclined to hinder (defend against).
When graphing those Open Passes versus Hindered Passes you can now infer (statistically measure) what doesn’t happen; i.e the ball is “not being passed successfully here”…
Put another way – if a player has the ability to make an Open Cross – that is completed. What didn’t happen is the fullback didn’t close on the winger and the center-back didn’t clear the ball.
If the Cross was a hindered cross then the value of defending can be determined even more. If it was a Hindered Pass that results in a shot taken then the fullback was not positioned properly to block the cross – nor was the center-back positioned correctly to clear the cross… Again – a statistical measurement of what doesn’t happen…
As a Youth Head Coach that type of information would be extremely critical to know when developing training plans between games… in considering how much money is involved at the professional level I would have thought the value would be even greater. Perhaps others may have a different view on that?
I’m not sure how clear that is but I’ll try to provide a few more examples as time passes… for now my early thoughts also include differentiating between an Open Throw-In and Open Cross versus Hindered Throw-In and Hindered Cross.”
In looking specifically at the Portland Timbers this year – they 10th (mid-table) in the DPWP Strategic Index – not bad by all accounts.
In peeling back the Defending Indicators they are 4th best in limiting their opponents passing accuracy (75.73%); they are 6th worst in preventing their opponents from completing passes in their defending final third (66.75%).
In terms of Possession percentage; teams average 47.38% – 6th lowest in MLS.
When looking at opponent shots taken per penetrating possession it’s 8th worst (18.85%)- and the percentage of opponents shots taken being on goal is 9th worst (36.72%).
Most critical (the weakest link it appears) is that the percentage of opponent possession leading to penetration is 26.48% (the worst in MLS). What this means is that over 25% of the time that the opponent has the ball they penetrate the Timbers final third… All told the final indicator (goals scored per game) is 3rd worst (1.65).
So how about the game against Vancouver?
- Vancouver had 45.57% possession – lower than the Timbers average.
- Vancouver passing accuracy across the entire pitch was 82% – higher than the Timbers average.
- Vancouver had 73% passing accuracy within the Timbers final third – higher than the Timbers average.
- Vancouver had 28.49% of their overall possession result in penetrating the Timbers final third – higher than the Timbers average.
- Vancouver had 10.27% of their shots taken per penetrating possession – lower than the Timbers average.
- Vancouver had 33.33% of their shots taken being shot on goal – lower than the Timbers average.
- Vancouver had 0% of their shots on goal result in a goal scored – lower than the Timbers average.
In conclusion: Here’s what happened in simple terms.
Portland ceded some space outside and slightly higher, within their defending third, in order to minimize the time and space Vancouver had in having their shots taken end up in the back of the net.
So while Portland didn’t park the bus they did get behind the ball, as much as possible, in an attempt to minimize risk… not rocket science – just good defensive team management.
Every game, for almost every team, is a ‘must win’ at this stage of the season – the ironic thing is that phrase has really been an accurate phrase for every game this season.
The earlier you consistently win games the less ‘must-ful’ they become as the season ends.
The exceptions to this, at this time, are probably Chivas USA and Montreal Impact.
Neither have a credible chance of making the playoffs – so those early season and mid-season games they lost were really their MUST win games – and of course, they didn’t win them.
As promised a reminder on coaching changes from last year; here’s the End of Season CPWP Strategic Index showing all the teams (stars) that had changes in Head Coaches during or after the season:
Note that five out of the six worst teams in PWP team performance saw coaching changes – and seven out of the bottom ten. Will we see that sort of house-cleaning again this year?
COPYRIGHT – All Rights Reserved. PWP – Trademark
Over my two year course of analyzing/researching team performance in soccer I’ve come across a number of general issues I have with modern day statistics in this game and think there is room for improvement.
That’s not to say most of the current statistics have weaknesses – they don’t; the majority of them have function and purpose but in considering recent discussions with others at the World Conference on Science and Soccer, plus my own analyses and that of others I sense (think and feel) there needs to be a better way to answer this issue.
Sometimes what doesn’t happen on the pitch has more value than what does happen. And my previous article entitled with that phrase is the impetus for this follow on article. Another article previously written (Hurried Passes) also attempts to capture more background on these potential improvements.
- Much of what doesn’t happen in this game is as much, if not more critical, to the outcome/result, as what does happen.
- Question – Is there a way to deductively or intuitively measure what doesn’t happen? I think so.
- Question – Do you feel or think it would be better to know the consistency of a striker scoring from ‘open shots’ versus a ‘hindered’ (not open) shots? This gives you two separate attacking data points for strikers – their success rate in scoring goals from open shots as well as their success rate in scoring goals from ‘hindered shots’. I do.
- Question – Do you feel or think it would be better to know the successful consistency of a team/ individual player being able to make an ‘open pass’ or a ‘hindered pass’? This gives you two separate attacking data points for everyone – their success rate in making open passes as well as their success rate in making passes while being ‘hindered’. I do.
- Question – Do you feel or think it would be more helpful to better quantify and qualify team and individual defending statistics that can be used to support what did and didn’t happen? I do.
- If you answered ‘yes’ to any of these questions please read on…
Potential new statistics for soccer (football):
- Open Shot – An open shot is when the striker can take a shot, directly towards goal, with no defender within a square yard or so that can hinder, or impeded the pure technical ability of a striker to strike the ball. This offers a number of ways to analyze a condition – 1) the team has had the skill necessary to generate and create a clean shot on goal. 2) the defending team has been put into a condition where that open shot clearly indicates they were out of position, and 3) the defending team, when ‘not’ giving the opponent an Open Shot has done their best to be in a position to deny a clear shot on goal; in other words this is a deductive way of measuring ‘what didn’t and did happen’ with respect to good/poor team / individual defending and good/poor team/individual attacking.
- Open Pass – An open pass is when a teammate can pass a ball, directly towards any other teammate, where no defender, within a square yard or so is present to hinder that pass. This offers a number of ways to analyze a condition 1) the team has had the skill necessary to generate and create a clean pass to a teammate, 2) the defending team has either made a consicious decision not to challenge a pass from that area (note the location of the pass needs to be graphed) or the team was not in a good defensive position to hinder that pass, 3) the defending team, when ‘not’ giving the opponent an Open Pass has done their best to be in a position to deny a clear pass to another teammate; in other words this is a deductive way of measuring ‘what didn’t and did happen’ with respect to good/poor team / individual defending and good/poor team / individual attacking.
I truly believe more effective and efficient analysis can come from defining passes and shots as being ‘open’ versus ‘hindered’ and by doing this it creates a more effective way to filter and help better determine what statistically doesn’t happen versus how current approaches are taken to measure what does happen.
And with this approach, other ‘did happen’, statistics like tackles, interceptions, blocked crosses, blocked shots can add additional clarity on the ‘did happen’ while the what ‘didn’t happen’ can now be more precisely graphed and plotted to better track good/bad zonal defending schemes versus good/bad man-to-man defending schemes – further identifying individual performance indicators that plot strengths and weaknesses of individual performance as as well as tactical coaching performance.
From an operational standpoint it merely means adding two new statistical categories (Open Shot and Open Pass) – the current statistical categories (Shot and Pass) would merely be redefined as being as the terms associated with shots and passes that are ‘hindered’.
And yes, it will be slightly judgmental (nothings perfect and even the refereeing in this game still remains judgmental) but with modern day technology I’m sure the video analysis programs can be tuned to generate that statistic based upon the physical presence of a data dot (of the player) relative to the other player making the pass or shot… – especially with the advent of GPS.
For more explanations on this concept read here (about 2/3rds of the way through the article)…
If you’re in the world of soccer statistics and you think or feel these improvements add value please retweet. In addition, under any circumstances 🙂 please add comments, as appropriate.
COPYRIGHT, All Rights Reserved. PWP – Trademark