Tagged: Predictability Indices

Expected Wins Five – Europe

In my previous series on Expected Wins Four – probably more appropriately entitled “Expected Points” – I’d taken a look at how the general tendencies of four primary Leagues in Europe (England, Germany, Spain, as the UEFA Champions League) compare to Major League Soccer – Is European Football Really Higher Quality than Major League Soccer?

This time I’m focusing strictly on Europe and offering up how things stand in PWP with the season coming to a close soon.  But before digging some things to share about PWP to date:

A reminder – PWP is about two things:

  1. The End State in that the final Index comes as close as possible to the League Table without using points earned in any of the calculations, and
  2. Recognizing that soccer is a game that is played in a free flowing environment – picture two amoeba fighting against each other in a confined space…. There is attempted control by the Head Coach that includes tons of preparation to set the stage for ‘an approach’ to earn three points – and then there is the game itself where there is but one time out (halftime) – no namby pamby huddles or official stoppages of play between possessions.  Meaning these guys play a full-on, in your face (sometimes literally), non-stop, constantly thinking and reacting to the game that can literally see the ball go in any direction at any time… not purely random but close.

Given that, PWP attempts to tone down all that volatility and parse out general tendencies that fall within the bell curve of activities – it’s not perfect – but it’s bloody good… and yes – I have made a few mistakes along the way (if you don’t work you don’t make mistakes).  The latest has been a technical mistake – the relationship of CPWP to the League Table is not an R Squared number (Coefficient of Determination) it is an R number (Correlation Coefficient).

For the stats followers that may be an issue… but even with the Modernized TSR (read here) the CTSR “R” is still generally lower (team to team) and certainly lower (table to table) than CPWP – meaning there still remains room for both statistical analytical approaches in a gmae that is played across the world…

Also, my thanks to some great research by Rob Lowe, a mate with the same passion for footy, who has asked to collaborate with me in the future.  He has done some additional regression analysis on the data points of PWP with respect to goals scored and points earned.  I should point out that his results show that not all six of the data points in the PWP equation independently-directly relate to goals scored or points earned.  For me that is okay – and actually great news for a few reasons…

  1. Both of my two new statistics (Passes Completed in the Final Third per Passes Completed across the Entire Pitch – Step 3 of PWP) and (Shots Taken per Completed Pass within and into the Final Third – Step 5 of PWP) did statistically relate to Goals Scored and Points Earned (independently).  Meaning those new statistics are relevant – both within the context of PWP and outside the context of PWP.  It’s this statistical regression type information that should solidify these two new statistics in the world of soccer.
  2. For both Possession (Step 6 of PWP) and Passing Accuracy (Step 5 of PWP) – as you will see a bit later – those two derived data points were never supposed to directly (independently) relate to goals scored or points earned as a matter of course I have advocated for quite some time that they shouldn’t.  PWP was built with the intention that the six derived data points only needed to relate to each other in a stair step relationship recognizing that in every game a team needs to possess the ball, move the ball, penetrate the opponent’s final third, take shots based upon that penetration, put them on goal, and score goals – all while preventing the opponent from doing the same thing.
  3. Another view on the outcome that Rob has noted – it’s unreasonable to analyze a game of soccer without taking those activities into account.  Rob’s positive feedback was that both possession and passing accuracy act as a “smoothing agent” within the Index – I agree but with beginning to learn the nuance of writing an Academic Paper I would put it this way.
  4. Possession and Passing Accuracy stats have limitations when vewing overall regression analysis relative to goals scored and points earned – but those limitations actually give the overall analyst of soccer a much better understanding about the context of activities that occur when a team is performing better than another team.
  5. In addition, Passing Accuracy statistics provide a coach a great measurement tool for how well some players may develop and progress into higher levels of competition – to exclude data of this import really ignores some of the most fundamental training aspects a team needs to do in order to improve.
  6. Also, there is excessive volatility in the percentages associated with Shots on Goal versus Shots Taken and Goals Scored versus Shots on Goal – if I only look at those two things then evaluating a game is all about (pass-fail) – granted winning and losing is pass-fail.  But to develop a “winning culture” a grading system perhaps more appropriate is A-B-C-D-F – in other words there are levels of success above and beyond pass-fail – especially when you are a team that isn’t at the very top of the league.
  7. By having Possession and Passing Accuracy in the equation you get a much larger (explanatory) picture on the culture of success – and as things appear to take shape, the Index itself, gives better clarity to that level of success for teams that are mid-table as opposed to bottom dwellers or top performers…

Now for the grist in Europe – first up – England: 

Note that the first two diagrams (in each four diagram grouping) highlight where the highest quantity and highest quality occurs within each competition – after some growing pains (earlier Expected Wins measurements) all four competitions now see the teams that win having the highest averages, in all categories, for both quantity and quality… proving (for the most part) that more is better and more results in more…

Barcleys Premier League PWP Data PointsBarcleys Premier League PWP Derived Data PointsEnglish Premier League CPWP IndexEnglish Premier League CPWP Predictability Index

All told the correlation, at this time, remains very strong – note that the “R” has replaced the “R2” in my third and fourth diagrams.

If I remove Possession and Passing Accuracy from the CPWP Index – the R value drops to .78 – statistically reinforcing that the Index, itself, better represents the standings in the League Table by including Possession and Passing Accuracy data.  Proving yet, another way, that goals scored and shots taken simply do not provide adequate depth on what activities occur on a pitch relative to earning points in the League Table!  And if you’ve read Moderning TSR this doesn’t mean ATSR/DTSR or CTSR doesn’t have value – it does…

As things stand today Chelsea take the League and since Man City, Man United, and Arsenal round out the top four (different orders) in both CPWP and CPWP-PI I’d offer it’s those four that advance to the UEFA Champions League next year.  The bridesmaid looks to be a two horse race (Spurs supporters may argue that) between Liverpool and Southampton.

Note that Southampton edges Liverpool in CPWP but that Liverpool edges Southampton in CPWP-PI – meaning when excluding Goals Scored – Liverpool has better quality than Southampton – so for Liverpool it’s more about converting Shots on Goal to Goals Scored – while for Southampton it’s more about getting clean sheets and scoring at least one goal; at least in my view – others may see that differently?

In retracing the earlier discussion on the data within the six steps of PWP – as you can see in both the first and second Diagrams (for all competitions) the Exponential Curve (Diagram 1) and well as Power Curve (Diagram 2) the stair step relationship between the data – point to point – are incredibly high…  Even more intriguing is how close those “R2” numbers are for both winning, drawing, and losing… really driving home the point, in my view, just how small the margin of error is between winning, drawing, and losing.

For goals scored (for or against) we really are talking about 5 or 6 standard deviations to the right of the bell curve…


 Bundesliga PWP Data PointsBundesliga PWP Derived Data PointsGerman Premier League CPWP IndexGerman Premier League CPWP Predictability IndexPerhaps the most intriguing issue this year isn’t the FC Bayern story – it’s the lack of goal scoring in Borussia Dortmund – when viewing the CPWP Predictability Index clearly Dortmund is offering up all the necessary culture the team needs in order to succeed – with one exception – goal scoring…. wow!

Another surprise may be Wolfsburg I’d pick them, and Bayer Leverkusen to finish two-three in their League Table – both show pedigree in team performance both with and without considering goals scored…


La Liga Premier League PWP Data PointsLa Liga Premier League PWP Derived Data PointsSpanish Premier League CPWP IndexSpanish Premier League CPWP Predictability Index

Barcelona and Real Madrid are locked in for the top team battle – my edge goes to Barcelona.  I’d offer more here but I’m simply not up on the La Liga as much as I’d like to be…

UEFA Champions League:

UEFA Champions League PWP Data PointsUEFA Champions League PWP Derived Data PointsUEFA Champions League CPWP IndexUEFA Champions League CPWP Predictability Index

The top eight teams that advanced are identified above – given the general success of CPWP relative to the top eight I’d expect FC Bayern Munich, BArcelona, Real Madrid, and Juventus to advance to the semi-finals.

In Closing:

My first of at least 4-5 Academic Papers is soon to be published – my thanks to Terry Favero for helping me work through this new experience – his support, patience, and knowledge in navigating all the nuance associated with writing an Academic Paper has been superb!

All four European competitions show more gets you more – this was not the case for Major League Soccer last year:

Major League Soccer Expected Wins FourWinners Expected Wins PWP Data Relationships Four

When more gets you more in MLS then I sense MLS has reached the BIG TIME – until then I think it’s a great breeding ground for Head Coaches that simply can’t get a job with a soccer club that has huge pockets of money.

Put another way – and many may disagree… I think a Head Coach who really wants to challenge their intellectual grit against another Head Coach can have greater opportunity to do that in MLS than they can by Head Coaching most clubs in Europe.

Why?  For at least one reason – a Head Coach in MLS really has to do more with less…

Errata – the first MLS slide indicates 654 events – the correct number is 646 events…

Best, Chris

COPYRIGHT – All Rights Reserved.  PWP – Trademark


CPWP Predictability versus MLS Results (Week 18 and 19)

Having been away on business last week I was unable to publish last weeks predictability versus reality results; in catching up here’s how things went in Week 18 and Week 19 versus the Composite Possession with Purpose Predictability Index (CPWP PI); excluding the Chivas USA v DC United match later this evening.

To begin here’s the CPWP Predictability Index for teams at Home, followed by, the CPWP PI for teams playing Away for Week 18/19…





Before digging into the results versus predictability note the significant difference in team performance at Home versus Away.

Pretty compelling evidence to reinforce what most believe, the home team usually does better… but… some teams can and will perform very strong on the road.

In reviewing the results… 

If you want the game by game comparison for Week 18 & Week 19 it can be found at the end of this article.

For now know that the CPWP PI accurately reflected five of the eight wins (draws excluded) for Week 18.

In addition, the CPWP PI accurately reflected seven out of seven wins (draws excluded) for Week 19.

If keeping track (after four weeks of leveraging the CPWP PI) it has been accurate in predicting 20 of 27 games (excluding draws); that’s a 74% success rate.

In Closing…

In general, the home team has won 74 games at home; while the away team has won 47 games on the road – the home team average percentage chance of winning based purely on results is 62%.

It would appear that the use of the CPWP, as a predictability model, gives someone a 12% better chance of predicting the outcome of a game then by purely picking the home team to beat the away team…

Perhaps others have a different view?

Best, Chris


Week 18:

San Jose, at home, lost to DC United 1 – 2.  San Jose, at home, has a .0368 CPWP PI while DC United, on the road, has a -.2174 – the CPWP PI was not accurate.

New York, at home, won against Columbus 4-1.  New York, at home, has a .1184 while Columbus, on the road, has a .1047 – the CPWP PI was accurate.

Toronto, at home, won against Houston 4-2.  Toronto, at home has a .0886 while Houston, on the road, is -.1706 – the CPWP PI was accurate.

Philadelphia, at home, drew with Colorado 3-3.  CPWP PI does not measure for draws.

Montreal, at home, lost to Sporting KC 1-2.  Montreal, at home, is -.0170 while Sporting KC, on the road, is .1112 – the CPWP PI was accurate.

New England, at home, lost to Chicago 0-1.  New England, at home, is .2516 while Chicago, on the road, is -.2241 – the CPWP PI was not accurate.

Vancouver, at home, lost to Chivas 1-3.  Vancouver, at home, is .1912 while Chivas, on the road, is -.1827 – the CPWP PI was not accurate.

LA Galaxy, at home, won against Real Salt Lake 1-0.  LA, at home, is .0476 while RSL, on the road, is -.1278 – the CPWP PI was accurate.

Seattle, at home, won against Portland 2-0.  Seattle, at home, is .2669 while Portland, on the road, is .0486 – the CPWP PI was accurate.

Week 19 (with the Chivas versus DC United game left to play):

Philadelphia, at home, defeated New York 3-1; Philadelphia, at home, is -.0107 while the New York, on the road, is -.0711 – the CPWP PI was accurate.

Columbus lost, at home, to Sporting KC 1-2;  Columbus, at home, is.0797 while the Sporting KC, on the road, is .1112 – the CPWP PI was accurate.

Toronto, at home, drew with Vancouver 1-1. (not measured).

LA, at home, beat New England 5-1; LA, at home, is .0476 while the New England, on the road is -.0565 – the CPWP PI was accurate.

Portland, at home, beat Colorado 2-1.; Portland, at home, is .0271 while Colorado, on the road, is -.0452 – the CPWP PI was accurate.

Sporting KC, at home, beat LA 2-1. Sporting, at home, is .3362 while LA, on the road, is .1393 – the CPWP PI was accurate.

New York at home, drew with San Jose 1-1. (not measured).

Columbus, at home, beat Montreal 2-1; Columbus, at home, is .0797 while Montreal, on the road, is -.0950 – the CPWP was accurate.

Chicago, at home, drew with Philadelphia 1-1. (not measured).

Dallas, at home, beat New England 2-0; Dallas, at home, is .0599 while New England, on the road, is -.0565 – the CPWP was accurate.

Houston, at home, drew with Toronto 2-2.  (not measured).

Real Salt Lake, at home, drew with Vancouver 1-1. (not measured).


Barcley’s Premier League – How Goes It?

In my latest installment of Possession with Purpose in Europe I have a number of diagrams to offer to include the latest on the PWP Predictability Index.

You’ll note that in every case the PWP Correlation to the League Tables for all four competitions has stayed the same or gotten better.

Also of interest is that a number of youth soccer teams, and another writer, have joined the queue in leveraging the PWP approach in analyzing soccer games – what remains, after publishing my Academic Paper (real soon as things go) is my ability to get data quicker and to set up a software system – probably using MS Access – to better enable match reporting.

It’s slow going – but that’s okay…  patience is a good thing…

Now for the grist in the English Premier League:

Last we spoke (after Week 26) here was the latest on CPWP Predictability;

  1. Eight of Ten
  2. Seven of Ten
  3. Eight of Ten
  4. Eight of Ten

In looking at Week 27 the CPWP Predictability Index was Six for Eight (hitting the 75% target).

For Week 28 the CPWP-PI had Man City earning at least a point vs. Leicester City, Chelsea earning at least a point vs West Ham, Man United earning at least a point vs Newcastle, Arsenal earning at least a point vs QPR, Everton earning at least a point vs Stoke, Spurs earning at least a point vs Swansea City, Liverpool earning at least a point vs Burnley, Aston Villa v West Brom dead even, Hull City earning at least a point vs Sunderland, and Southampton earning at least a point vs Crystal Palace.  Last but not least there was an off-game played between Spurs and QPR – the CPWP-PI had Spurs earning at least one point – they did.

  • In every case this week the CPWP-PI got it right with one exception – Stoke City took all three points against Everton!  So that made it ten for eleven in identifying whether or not a team would earn at least one point based upon the CPWP-PI.  In only two cases did the team expected to earn a point didn’t get three points – Aston Villa and Hull City.

For Week 29 the CPWP-PI had Chelsea earning at least a point vs Southampton, Everton earning at least a point vs Newcastle, Man United earning at least a point vs Spurs, QPR earning at least a point vs Crystal Palace, Arsenal earning at least a point vs West Ham, Hull City earning at least a point vs Leicester City, Aston Villa earning at least a point vs Sunderland, Stoke City earning at least a point vs West Brom, Man City earning at least a point vs Burnley, and Liverpool earning at least a point vs Swansea City.

  • Burnley had the upset of the week while Crystal Palace and West Brom continued their good, recent, run of form.  All told CPWP-PI correctly identified seven of ten teams earning points that week.

For Week 30 the CPWP-PI had Man United earning at least a point vs Liverpool, Chelsea earning at least a point vs Hull City, Everton earning at least a point vs QPR, Man City earning at least a point vs West Brom, Swansea City earning at least a point vs Aston Villa, Arsenal earning at least a point vs Newcastle, Southampton earning at least a point vs Burnley, Stoke City earning at least a point vs Crystal Palace, Spurs earning at least a point vs Leicester City, and West Ham earning at least a point vs Sunderland.

  • In every case but one the CPWP-PI correctly predicted what team would earn at least one point except for the loss Stoke City had against Crystal Palace – again – a team in good form since the coaching change!  That makes it nine of ten again this past week.

In summary:

  • Eight of Ten
  • Seven of Ten
  • Eight of Ten
  • Eight of Ten
  • Ten of Eleven
  • Seven of Ten
  • Nine of Ten
  • Totaling 57 of 71 for an 80% accuracy rating

Here’s the CPWP Index after Week 30:

CPWP Through Week 30 EPLHere’s the CPWP-PI Predictability Index for Week 30:

CPWP Predictability Index Through Week 30 EPL

For this next week CPWP-PI has:

  • Arsenal earning at least a point vs. Liverpool
  • Southampton earning at least a point vs. Everton
  • West Ham earning at least a point vs. Leicester City
  • Man United earning at least a point vs. Aston Villa
  • Swansea City earning at least a point vs. Hull City
  • West Brom earning at least a point vs. QPR
  • Chelsea earning at least a point vs. Stoke City
  • Spurs earning at least a point vs. Burnley
  • Newcastle earning at least a point vs. Sunderland, and
  • Man City earning at least a point vs. Crystal Palace
  • Another odd game has Aston Villa earning at least a point vs. QPR

In Closing:

Completion of my Academic Paper on Possession with Purpose nears…  another writer has asked to begin leveraging PWP analysis to their own team writing efforts and there are now three youth soccer clubs using the concepts and philosophy of PWP in trying to help their teams improve – both collectively as well as for their individual players.

Best, Chris


Valencia – Formula Won… La Liga

Most of the Headlines speak to the Real Madrid victory over the vaunted Barcelona; mine obviously don’t.  

For me Valencia is showing strong, and in my view, seems to have struck a great balance in attack and defense as they continues to impress.  And even though this early season run of form  might not last I do think it’s worthy to dig a bit deeper into their overall performance to see exactly why they are doing so well.

To begin – my standard Composite PWP Strategic Index:

CPWP Strategic Index Week 9

Why are Valencia so high in their overall team performance?

Is it their overall team attacking or defending performance?

At first glance you may think it’s their Attack – to review that here’s the latest Attacking PWP Strategic Index:

APWP Strategic Index Week 9

Even higher than Barcelona – one of the best attacking teams in the World!  Valencia are:

  • 7th best in overall possession – 51%; a full 17% less than Barcelona
  • 3rd best in overall passing accuracy – 85.97% – still less than Barcelona by 3%
  • 17th best (4th worst) in penetration per possession -19.71% – a full 13% below Barcelona
  • 9th best in Shots Taken per penetrating possession  – 15.84% – this time ~6% higher than Barcelona
  • 9th best in Shots on Goal per Shots Taken – 34.87% – roughly 5% lower than Barcelona
  • Finally, and perhaps the single greatest graphic difference is Goals Scored per Shots on Goal; at this point Valencia have scored a HUGE 60.83% of the time they’ve put a Shot on Goal – by comparison Barcelona sit at 31.68%..

In a phrase – Valencia ‘are’ the best team in performing the key indicators in possession with purpose.  They may not have the glitz and glamour of a Barcelona or Real Madrid but steady is good.

But before moving on to Defending I think it’s worthy to note their volume of activity not just the percentages above:

  • They match the league average in passes attempted (410) what skews that average is Barcelona and Real Madrid.  All told only six of the 20 teams in La Liga exceed the league average.
  • As noted above their passing accuracy is 3rd best in the league – with that their total completed passes across the entire pitch is 5th best at 349.
  • So by volume they are not what would be considered a dominating possession based team.
  • And in looking at their overall penetration into the final third Valencia average 107 passes per game – 13th best.
  • In other words they’re not really a possession based team, they are more of a counter-attacking team who simply wait for some extremely superb moments to take advantage of the opponent’s weaknesses in order to create ideal time and space conditions.
  • And to reinforce this view they are slightly lower (10.78 per game) than the league average (11.45) in Shots Taken – but slightly lower in Shots on Goal (3.78) versus the league average of (4.03).
  • And that ‘finishing touch’ sees them average 2.22 Goals Scored per game compared to the 1.34 for La Liga and just slightly lower than Barcelona’s average of 2.56 per game!

All told – Valencia are simply a team that is performing at an optimal rate.

But that’s not the complete answer for Valencia – here’s how they stand in the Defending PWP Strategic Index:

DPWP Strategic Index Week 9

They are 3rd best in La Liga in defending team performance; here’s how the key indicators compare to others as well as Barcelona:

  • Opponents average 48.68% possession – pretty much meaning the opponent has the ball as much as Valencia – opponents of Barcelona possess the ball just 31.18% of the time.
  • Opponents average 77.62% passing accuracy – and I’d offer that is more down to the amount of space Valencia cede outside their Defending Final Third – we’ll take a look at that when reviewing the volume of opponent activity.
  • In terms of penetration and shot creation from that penetration their opponents are 10th best at penetrating 24.09% of the time they possess the ball while also generating shots taken 16.21% of the time.
  • All told that leads to an opponent accuracy shot rate on goal of 35.53% with 21.67% of those shots on goal scoring a goal.
  • Bottom line here is that with average penetration (compared to others in La Liga) and average shots against, Valencia are 4th lowest in facing shots on goal and 4th lowest in seeing those shots on goal score goals.

It would appear they have a very organized defensive system and a very good Goal Keeper.

So how about the volume of attack faced from their opponents?  

  • At this stage they have faced, on average, the 8th fewest passes per game (388) compared to Barcelona at 300.
  • In terms of overall penetration, the opponents have offered up 117 passes per game in the Valencia Defending Final Third – with that being the 10th most in La Liga.
  • Statistics would seem to indicate that they do make it easier for their opponents to penetrate – which in turn appears to support what was offered up earlier.
  • When it comes down to shots faced they are 9th lowest in that category – while translating that to just 3.78 shots on goal (tied 8th best).
  • All told that added volume of penetration sees Valencia with a .89 goals against per game – 3rd best in La Liga.

Bottom line here – like what the percentages offer – Valencia cedes time and space outside the Defending Final Third while doing a great job of closing up shop as the opponent finally gains entry.

Is that the right mix to minimize the likes of Real Madrid, Barcelona, and Sevilla?

Hard to say at this time – but clearly – going into Week 10 against Villarreal it is likely they should get another three points.

Which brings me to my last Index – the CPWP Predictability Index.

In MLS this Index averaged a 55-65% accuracy in identifying the winner of upcoming games – at times the outputs were pear-shaped while others were spot on.

I have no idea how this will play out this year in Europe but here’s the Index itself and then a quick blurb on how to understand it:

CPWP Predictability Index Week 9

As noted Valencia take on Villarreal this weekend – note that Valencia has a higher number than Villarreal – simply meaning, with the law of averages considered, and the teams perform as they have in the past Valencia should win.

So in looking up the schedule for next weekend; Getafe should edge Deportivo; Real Madrid should defeat Granada; Atletico Madrid should defeat Cordoba; Barcelona should beat Celta de Vigo; Real Sociedad should defeat Malaga; Athletic Club should beat Sevilla; Levante should lose to Almeria; Elche should lose to Espanyol; and Rayo Vallecano should beat Eibar.

By the way – the Predictability Index is made up of all the PWP Data Point Relationships excluding ‘goals scored’ and ‘goals against’ – you really can’t develop a worthy predictability index using goals scored.

That should help explain why Celta de Vigo are higher up the prediction table than Valencia… based upon their overall run of play performances Celta should probably score more goals than they do.

All for now…

Best, Chris

COPYRIGHT, All Rights Reserved.  PWP – Trademark

You can follow me on twitter @chrisgluckpwp

MLS – Reading the Tea Leaves and Predicting Week 20 Outcomes…

A full slate of games for Week 20 that started off with a smashing win by San Jose over a very shabby Chicago – were you surprised; you shouldn’t have been. 

Chicago Fire have the worst average in overall team defensive performance of any team in MLS this year  – and it certainly doesn’t get better on the road.  Read my latest on the Attacking and Defending PWP Indices from Week 19 here.

In addition; San Jose – believe it or not – are one of the most frugal teams in Defensive Possession with Purpose this year (3rd best in my Index to be exact {2.2510}).

To start – a reminder of three things:

  1. The Possession with Purpose Predictability Indices work from PWP “without” Goals Scored or Goals Against – in other words I use the bell curve of five activities, not six, in order to offer an Index of prediction.
  2. To date, after four weeks, the PWP PI has been 74% accurate in predicting the outcomes of games – if you just pick the “home” side to win every game you’d have a 62% chance of being accurate.
  3. This Predictabiliy Index is not leveraged until after most teams have played at least 17 games – there is not enough historical data to generate worthy trends prior to the half-way point in the season.
  4. The two PWP Predictability Indices (Home) and (Away) are provided below for your consideration before I offer up the predicted results.







Let the games begin…

  1. San Jose at home to Chicago – results are already in – San Jose wins – PWP PI indicates San Jose should have won (.0368) to Chicago (-.2241).  PWP PI was accurate.
  2. Real Salt Lake at home to Montreal – Home PI for RSL =  .1374 / Away PI for MIFC = -.0170.  PI predicts RSL win.
  3. Colorado at home to Chivas USA – Home PI for CRFC = .1754 / Away PI for CUSA =  -.1827.  PI predicts CRFC win.
  4. Toronto at home to Sporting KC – Home PI for TFC = .1010 / Away PI for SKC = .0929.  PI predicts TFC win.
  5. New England at home to Columbus – Home PI for NER = .2516 / Away PI for CCFC =  .2047.  PI predicts NEW win.
  6. Vancouver at home to FC Dallas – Home PI for VWFC = .1912 / Away PI for FCD = -.2379.  PI predicts VWFC win.
  7. Montreal at home to Portland – Home PI for MIFC = -.0170 / Away PI for PTFC = .0486.  PI predicts PTFC win.
  8. Seattle at home to LA Galaxy – Home PI for SSFC = .2669 / Away PI for LAG = .1031.  PI predicts SSFC win.

In Closing…

That’s from a clinical/objective standpoint looking at the comprehensive ‘bell curve’ of activities that teams have offered in their first 19 weeks of play.

There are intangibles, as always, in soccer – with 22 players, 2 Assistant Referees, 1 Referee, 2 Head Coaches and the potential for 6 total substitutions most anything can happen that might turn a game on its head.

I can’t account for those intangibles but if teams have a propensity for making mental mistakes, getting red cards, or yielding PK’s, on a regular basis, than those intangibles will pile up and impact/influence overall team performance.

Good examples of intangibles at this stage include some:

  • New England are in a slump – seven straight losses
  • FC Dallas are on an up-swing
  • As are Chivas USA
  • Montreal have two games (in four days) against two of the best attacking teams in the Western Conference
  • Colorado have some injuries to deal with
  • Seattle is taking on an LA team that just got thrashed by Manchester United – but LA are simply one of the top performing teams in attack and defense (regardless of being home or away)
  • Toronto are really beginning to gel in attack
  • Portland are one of the best road teams in MLS this year and the addition of Liam Ridgewell does appear to have made their defensive scheme better.
  • Western Conference teams have simply done better against Eastern Conference teams this year (105 points now for the West versus just 69 points for the East in head-to-head competitions).


  • And they are considering moving Sporting KC to the West at some point in the future?  WOW – talk about an unbalanced Major League Soccer Conference scheme!
  • As much as it pains me to say it there should be one Conference and one League or a split to create four Conferences (two east {north/south} and two west {north/south}).
  • If the league is going to operate based upon the ‘entitlement’ that once a team is “in they stay in” (forever with no relegation) then the sooner this league organizes itself like other major sporting leagues in this country the better.

Finally, and perhaps the most controverisal of my views.

  • I don’t look at individual statistics
  • The game is played by a team… and teams win and lose – individual players don’t.
  • Actions, as much as “non-actions” both impact and influence the outcome of games.
  • And no… the statistics that folks should consider generating for this league, as a whole, should not mirror those of Baseball.
  • The further away from Baseball type statistics the better it will be for others (new to the game) to really understand how much of a ‘team game’ soccer really is.

That’s my soap-box rant for the day – a good podcast to listen to where I explain that view is here:  Yellowcarded Podcast.

  • The time hack to begin listening starts around the 3 minute mark and goes to the 35 minute mark
  • Towards the end of that 35 minute mark I respond to a question about the MLS Castrol Index that speaks to my views on the strengths and weaknesses of individual statistics
  • MLS Castrol Index – An individual Index (sponsored through MLS) that is so obtuse and inaccurate it’s mention is hardly worthy as I feel like it’s a backhanded recommendation to click on it and review the outcomes – don’t waste your time!

Best, Chris


World Cup 2014 Final; the two best teams? You bet!

If you’ve been following my adventures in Major League Soccer you’ll know that last year the PWP Index did a pretty good job in showing how the team performances played out in comparison to the League Tables (without) including points scored in my calculations.

To be honest, with such a small sample point I really didn’t think the PWP Indices effort would stand up against the Tournament (knock-out based style) of the World Cup.

But after taking a look at all the games (and inputting the team performance from said games) my Indices seem to hold up pretty well – wonder when Pepsi or another company that begins with “P” will consider sponsoring my work?  (just kidding – erhhh maybe not?).

Anyhow – here’s the lay of the land as it was tweeted earlier today:



NOTE:  All games are entered – and the comparison of these games does include the extra games played as the competition has headed towards the finals.

In other words Germany, Argentina, Brazil and the Netherlands all have six games worth of data.  In developing this I figured the more data points for a team the more likely their percentages would be watered down.

So for a team like Spain, who went out in the first round I figured they’d be pretty high up – well they are but the pedigree of the Netherlands, France, Colombia, Argentina, and Germany all put them past Spain EVEN with more games played!

If you’ve read my presentation at the World Conference on Science and Soccer as well as my Introduction into Possession with Purpose you’ll know my measurement methods and data source for this effort.  I can’t thank MLS Soccer.com enough for the publicly available data that allows me to generate my Index formulas.

Perhaps Prozone or someone else might help me obtain the data I need for all the European Leagues, to include the Champions League?

So with the overall accuracy (pretty compelling it appears to me) I’ve put my Composite PWP Predictability Index to test for the final (ahead of time)…

Before offering that Index though here’s how the teams compared against each other in Attacking PWP and Defending PWP:



From an attacking standpoint Germany are top of the table with Colombia 2nd, France 3rd, and Argentina 4th.

And when witnessing that blowout yesterday is that really a surprise, perhaps somewhat, but even prior to that game Germany were 3rd best overall in Attacking PWP – behind only Argentina and Colombia.

So how about the Defending PWP Index?

Notice (below) that Brazil is 17th out of 32 teams; prior to that game against Germany, Brazil were 12th.

So while some favored Brazil – the overall team performance indicators did show that Brazil were behind Germany in both the APWP and DPWP prior to that game.

The same cannot be said for Argentina and Germany – those two split top honors as you can see below as Argentina heads this Index; while Germany is a close 2nd.



Also note, if you’re a supporter of the United States, they were much higher in this Index (21st best) than they were in the Attacking Index (5th worst).

It is worthy (and most probably realistic) that if the United States had taken a stronger attacking stance against Germany, and perhaps even Belgium, they might have been the team getting embarrassed and not Brazil!

Finally, here’s the CPWP Predictability Index:



A pretty close call; in this one Germany has the slight edge in Composite Predictability in comparison to Argentina.

Argentina is #1 in the DPWP Predictability Index (not pictured) and  Germany is 4th best.

Germany is #1 in the APWP Predictability Index and Argentina slides all the way down to 16th best.

A distinct difference in Attacking and Defending Predictability based upon previous team performance while excluding goals scored…

In closing…

The overall Composite PWP Predictability Index indicates Germany is better in attack and Argentina is better in defense; the Predictability Indices indicate the same outputs.

For me, and my PWP calculations this should make for a brilliant final this weekend!

No personal prognostications from me – my objective team performance indicators point one way in attack and one way in defense; usually in games like these the better defensive teams win…

Best, Chris

Week 17 in MLS (2013 versus 2014) PWP; And what about DC United this year?

Over a year has passed since my first broad strokes about Possession with Purpose were applied to Major League Soccer; since then we’ve had one full year to look at it and how things have played out.

So how do things stack up today versus Week 17 last year, and, is something going on with DC United (besides the new strikers) that is different this year?

To begin; here’s a look at the teams after 17 weeks in 2013:



The top five Western Conference teams were Portland, Real Salt Lake, LA Galaxy, Vancouver and Seattle; the only team not to make the Playoffs last year was Vancouver.

Upon reflection, it was their defense that let them down, and the most probable reason why Martin Rennie got sacked.

In looking at the top five Eastern Conference teams they were Sporting KC, New England, New York, Montreal, and Houston – the same top five teams that eventually made the Playoffs.

So how about this year?

CPWP INDEX End of Week 17

CPWP INDEX End of Week 17

In looking at the Eastern Conference teams, the top five are Sporting KC, Columbus Crew, DC United, New England and New York – the odd one out, at the moment, is Toronto vice Columbus.

It should be noted that Toronto also have at least two, and no less than four, games in hand – so it’s not exactly “apples to apples yet” but should be in about 3 weeks time. As for the Western Conference, the top five so far are LA Galaxy, Seattle, Colorado, Portland, and FC Dallas.

Again the games in hand vary somewhat.

The HUGE, if not inordinately large question here is… Can the Portland Timbers turn their defensive nightmare of a season around with a healthy Norberto Paparatto, Pa Madou Kah and newly signed Liam Ridgewell, for three solid center-backs?  And, if so, does that fix the defensive issues?

Now an even tougher question…

Is the level of accuracy, last year, to be expected this year (nine for ten in teams last year making the Playoffs, based upon 17 games of data)?

I’m not so sure… And a good reason for that is the emerging clarity on how effective some teams have become (this year) in winning or drawing games with less possession…

In other words, playing to a counterattacking style, that sees some teams offering the opponent higher levels of possession, penetration, and shots taken.

So is there another way to try and answer the question about accuracy in the CPWP Index?

How about the CPWP Predictability Index – what does that offer after Week 17?

CPWP Predictability Index Week 17

CPWP Predictability Index Week 17

In looking at the CPWP PI, the numbers seem to indicate that Sporting KC, Columbus, New England, New York and Philadelphia have the best chances of winning, given historical team performances this year.

So the PI sees Philadelphia with an edge over Toronto… (reminder – TFC have four games in hand though)…

And does that Head Coach change, where Curtin is now in charge over Hackworth, reflect the Hackworth predictability of Philadelphia or the Curtin predictability of Philadelphia?  More to follow on that in a later article for sure…

As for the Western Conference; LA leads with Colorado, Seattle, Vancouver, and Portland – that sees FC Dallas dropping out with a smaller chance of winning and Vancouver sliding in…

And yet, neither Index has Real Salt Lake in the top five – could that be? Has the loss of Saborio, Beckerman and Rimando impacted RSL that much in such a short time span; and what does that say for the second half of the season? Lots of questions with no answers yet…

Now… take a look how far down DC United are in the Predictability Index (5th worst predictability in winning) – might that indicate how fortunate they have been in scoring goals or is that a reflection of something else going on?

DC United have the second best Goals Scored versus Shots on Goal of all the teams in MLS (42.12%); FC Dallas lead MLS in that category with 44.26%. Clearly the addition of Espindola and Johnson (even if they don’t play together) has added extreme value to this team.

Especially when their percentage for this same statistic, last year, was just 16.66% I wonder what the Expected Goals look like for DC United and how their shot locations may have changed this year compared to last year? Perhaps one or two folks who specialize in Expected Goals can help answer that one?

I did check to see if they have been awarded more PK’s than other teams – no – only 2 PK’s awarded so far this year.

As for Opponent Red Cards?

Perhaps that has created a positive influence in Goals Scored? Their opponents have had 5 Red Cards this year (two by FC Dallas in one game) – that is tied for 3rd highest (best/most advantageous) in MLS.

Has that helped?  I think so…

DC United have 10 points in the four games where their opponent has been red-carded and nine of their 24 Goals Scored have come from those games.

So, in retrospect – if the opponent’s for DC United “play-fair” it is (likely?) that will negatively impact DC United in the League Table.

That’s one advantage of the CPWP PI – it is not ‘doubly’ influenced by opponents being Red or Yellow Carded – it’s strictly five of the six primary data points of PWP.

In closing…

Still plenty to play for and any team, and I mean any team, can get on a winning streak – just look at Chivas USA their last three games.

How all the ‘defensive bunkering’ folds into the PWP Indices and Predictability outcomes has yet to play out. When every team reaches 17 games I’ll regenerate this article with updated information.

Best, Chris

Reflections of MLS Week 16; Predictable or not??? And what about Chivas USA these last three games… anything there in PWP to see?

As you know I’ve attempted to create a Predictability Index (PI) from my Possession with Purpose (PWP) analysis.  Here’s a link in case you missed the first article on PWP Predictability.

Before looking at the overall results here’s a reminder on where all the teams stand after 17 weeks:

CPWP INDEX End of Week 17

CPWP INDEX End of Week 17

Not every team has played 18 games yet so the Index is not equal – just like the MLS Table; Toronto have four games in hand over some teams in the Eastern Conference and the LA Galaxy have as many as five games in hand over some teams in the Western Conference.

When looking at the Western Conference CPWP (where all teams have played 14 games) the Index has LA atop (.2380); with Seattle 2nd (.2008); Colorado 3rd (.1578); Portland 4th (.0616) and Vancouver 5th (.0470).

All told that’s 3 of the top five teams in the Western Conference – not ideal but pretty close.

When looking at the Eastern Conference CPWP (where all teams have played 14 games) the Index has Sporting FC atop (.2219); with Columbus 2nd (.1578); DC United 3rd (.0807); New England 4th (.0347) and New York 5th (-.0416).

All told that’s four of the top five teams in the Eastern Conference – again not ideal but pretty close.

How does last year compare to this year after Week 17?  I’ll cover that in my next article…  For now since most teams have eclipsed the 17 game barrier I use the separate Home and Away CPWP Predictability Indices…

A reminder, of sorts, the CPWP PI is not intended to predict draws; it’s strictly an attempt to “test” how well it can/could predict wins.

The diagrams (along with individual Team Index numbers)  are provided at the end of this article.

Before kick-off; a reminder that last weekend’s games saw the CPWP PI had relevance in five out of six games where a team won/lost versus drew.

So for teams that won on the road this week we have:

Chivas USA defeating San Jose and DC United defeating Toronto FC.

The away CPWP PI for Chivas USA is -0.19; the home CPWP PI for San Jose is -0.04; the PI indicates Chivas should have lost – they won (inaccurate).

The away CPWP PI for DC United is -0.16; the home CPWP PI for Toronto FC is +0.09; the PI indicates DC United should have lost – they won (inaccurate).

So for teams that won at home this week we have:

FC Dallas defeating Philadelphia Union; Real Salt Lake defeating New England Revolution, Vancouver Whitecaps defeating Seattle Sounders and Chivas USA defeating Montreal.

The home CPWP PI for Dallas is +0.07; the Away CPWP PI for Philadelphia is -0.02; the PI indicates Dallas should have won – they won (accurate).

The home CPWP PI for Real Salt Lake is +0.04; the Away CPWP PI for New England is 0.00; the PI indicates Real Salt Lake should have won – they won (accurate).

The home CPWP PI for Vancouver is +0.18; the away CPWP PI for Seattle is -0.06; the PI indicates Vancouver should have won – they won (accurate).

The home CPWP PI for Chivas USA is -0.28; the away CPWP PI for Montreal is -0.11; the PI indicates Montreal should have won – they lost (inaccurate).

In closing… and that promised look at Chivas USA.

All told where there weren’t draws the CPWP PI was three out of six games.

Excluding draws that’s two weeks of (5 for 6) and (3 for 6); (8 for 12) = 66% accurate.

Clearly betting against Chivas USA at this time is not a worthy endeavor.

Here’s the differences in their Possession with Purpose indicators in the first 14 weeks compared to the last three weeks:

  • First 14 Weeks (APWP = 2.1425 / 2nd worst in MLS)
  • First 14 Weeks (DPWP = 2.5341 / 2nd worst in MLS)
  • First 14 Weeks (CPWP = -0.3915 / worst in MLS)
  • Last three Weeks (APWP = 2.2217 / 5th worst in MLS)
  • Last three Weeks (DPWP = 1.9502 / BEST in MLS)
  • Last three Weeks (CPWP = 0.2715 / BEST in MLS)

With that significant change in Defending PWP it’s worth a quick look to see what’s what in the first 14 Weeks versus the last three weeks…

  • First 14 Weeks Opponent (Possession 57.14%, Passing Accuracy 79.73%; Penetration 15.84%; Shots Taken per Penetration 19.34%; Shots on Goal versus Shots Taken 38.15%; Goals Scored versus Shots on Goal 43.21%)
  • Last three Weeks Opponent (Possession 57.96%; Passing Accuracy 79.67%; Penetration 19.21%; Shots Taken per Penetration 15.27%; Shots on Goal versus Shots Taken 22.92%; Goals Scored versus Shots on Goal 0.00%)
  • The differences?  Opponent penetration has increased while the number of opponent shots taken and shots on goal and goals scored have decreased.
  • Without having seen any of their games I would offer that Chivas has decided to open up the opponent opportunities in penetrating in order to tighten the screws a bit deeper inside the 18 yard box…
  • In other words they are not running two banks of four players atop and outside the final third – they have dropped a bit deeper and are now running their banks of four more within and around the 18 yard box.
  • Perhaps others who follow Chivas USA more closely could offer visual information to determine if that is an accurate assessment?

 As promised the CPWP PI Home Index:



As promised the CPWP PI Away Index:



Best, Chris

Next up Week 17 PWP in review…

PWP Predictability Index vs MLS Results Week 16

While I didn’t venture any predictions for this past weekend in Major League Soccer I thought it would be fun to see how the overall Composite Possession with Purpose (CPWP) Index fared compared to the results in Week 16.

As a reminder here is the CPWP Predictability Index from Week 15:

CPWP Home and Away Predictability Index Week 15 MLS

CPWP Home and Away Predictability Index Week 15 MLS

Vancouver at home to Montreal; result (draw) – CPWP PI predicted a win to Vancouver.

New York at home to Toronto; result (draw) – CPWP PI predicted a win to New York.

Portland at home to Sporting KC; result (Sporting KC win) – CPWP PI predicted a win to Sporting. 

DC United at home to Seattle; result (Seattle win) – CPWP PI predicted a win to Seattle. 

New England at home to Philadelphia; result (Philadelphia win) – CPWP PI predicted a win to New England.

Colorado at home to Vancouver; result (Colorado win) – CPWP PI predicted a win to Colorado. 

Chivas at home to Real Salt Lake; result (Chivas win) – CPWP PI predicted a win to Real Salt Lake.

San Jose at home to LA Galaxy; result (Galaxy win) – CPWP PI predicted a win to LA Galaxy.

Columbus at home to FC Dallas; result (draw) – CPWP PI predicted a win to Columbus.

Montreal at home to Houston; result (Montreal win) – CPWP PI predicted a win to Montreal.

Excluding draws – which the CPWP PI does not predict – where there were winners and losers the CPWP PI was five out of six in indicating who might win versus lose.

So where New England lost to Philadelphia – what, if anything, did Philadelphia Union do that was different from their historical averages so far this year?

Here’s my article freshly pressed to try and offer some answers to that question…

In closing…

I’m not sure how well the CPWP PI will play out this year – I won’t offer predictions prior to games using it – I still think and feel more games (data) is needed.

But I will look back each week and see how the CPWP PI plays out and look to see what was different, in a team performance way, that led to a result that didn’t fit the picture.

Best, Chris


Possession with Purpose – Predictability Indices – Major League Soccer

After some superb discussion at the World Conference on Science and Soccer folks back in 2014 I first wrote this article; I’ve since improved the Total Soccer Index (though I haven’t published the new algorithm).

Here’s my most recent article on predictability (Predicting Team Standings in Professional Soccer) – it was this article that led to me presenting the PWP TSI Predictability topic at the WCSS 2017 (Rennes, France) last year.

Anyhow…  back to this article.

To do this we agreed that Goals Scored needed to be removed from the equation.  In doing that here’s how the Composite PWP Index offers up an Expected Wins Index number for MLS teams (both home and away) and then separately for home and away games.

I took this extra step since most feel or think that teams who play at home have a better chance of winning than teams who play away from home; and indeed the average for home teams this year is 1.58 points, per game, versus away teams being 1.09 points per game.

Caveats prior to the diagrams:

  1. There will always be an issue with the sample size when looking at a predictability model based upon team performance (within a single game) as opposed to specific individual actions by players like shots taken and shots on goal – repeatability here is not those specific instances of activity but the ‘comprehensive instances’ of ‘team activities within a game’…
  2. For me that represents the primary bell curve of the game as opposed to, in my view, the 3rd or 4th standard deviations of the bell curve where goals are scored… I hope that makes sense???
  3. There are only 17 home and 17 away games – again that sample size is extremely small – but the volume of team activities within those games ‘should’ provide a general (bell curve) picture on what regular activities a team takes in order to score goals and/or prevent goals being scored.
  4. The Predictability Indices are the PWP Indices (minus) the percentages of Goals Scored versus Shots on Goal.
  5. No additional analysis to go with the CPWP, APWP and DPWP Predictability Indices – in my view there simply aren’t enough data points yet; when all the teams have reached 10 games played (both at home and away) I’ll break open the analysis as a real predictive tool to try and predict a win, draw or loss.
  6. For now consider they represent a ‘visual diagram’ that also includes the current Correlation (R) the Indices have with respect to total points or an average of points per game played in MLS.

The Composite PWP Predictability Index (PI):  The CPWP PI has a best Correlation with respect to Average Points earned in the MLS League Table (R) .52

CPWP Home and Away Predictability Index Week 15 MLS

CPWP Home and Away Predictability Index Week 15 MLS

The CPWP PI for Home games has a best Correlation with respect to Average Points earned in the MLS League Table (R) .66:

CPWP Home Predictability Index Week 15 MLS

CPWP Home Predictability Index Week 15 MLS

The DPWP PI for Home games has a best Correlation with respect to Average Points earned in the MLS League Table (R) .60:

DPWP Home Predictability Index Week 15 MLS

DPWP Home Predictability Index Week 15 MLS

The APWP PI for Away games has a best Correlation with respect to Sum of Points earned in the MLS League Table (R) .58:

APWP Away Predictability Index Week 15 MLS

APWP Away Predictability Index Week 15 MLS

In closing….

The visual diagram with the best Correlation to Points in the MLS League Table is CPWP-PI with an R of .66; and the second best is the DPWP-PI with an R of -.60.

So from a, point going forward approach, it would seem to me that the best visual diagram to use when offering up analysis later this year is the CPWP-PI compared with “Average” Point taken per game as opposed to the “Sum” of Points in the MLS League Table.

We will see if that holds true in a few weeks time; thanks for your patience.

Best, Chris