The folks at Prost Amerika Soccer have been kind enough to ask me to join the team. To help prepare you for the type of soccer analysis/reporting I’ll offer this year I think it best to give you a taste of my most recent article on soccer statistics.
- As you read through what is offered there will be a few links to some of the latest analysis, as well as my basic approach that was published in England last year — Possession with Purpose.
- Please note, I rarely dig into individual player analysis – my belief has always been – team first.
- With that here’s my latest article to get your feet wet.
- “Moneyball 2 – Soccer Statistics – Taking it to the next level”.
- Thanks in advance for your patience.
Probably the biggest leader in soccer analytics (Prozone Sports) have released two products in the last year, or so, that should begin to change the landscape of soccer statistics.
I applaud their actions but I’m still not convinced they scratch the right itch, or at least my question; what are the best statistics available to better understand team performance and how individual players influence the outcome of team performance?
Warning – before settling down to read this, get a cup or pint of your favorite beverage.
My reason for wanting to know the answer to this question is three fold:
- The first is knowing, or having a better idea, of what information I would like, as a senior leader/owner, to best understand my teams’ performance relative to my opponents/competitors.
- The second is knowing, or having a better idea, on what types of players I would need to buy/sell in order to improve my team.
- The third is knowing, or having a better idea, on what the true value of a player is when looking to negotiate a contract.
Before digging into what might scratch my itch here’s a summary on what Prozone Sports has done of late:
Styles of play (Possession) = (tactical profiling). Dr. Hector Ruiz and Prozone Sports have identified eight styles of play (what I almost call eight styles of possession). They are:
- Direct
- Counter-attacking
- Maintenance
- Build-up
- Sustained threat
- Fast tempo
- Crossing
- High pressure
Each of these styles of play have definitions and video examples on what they mean and how they are calculated. It’s a great presentation and opens the eyes on many different ways we can view the game.
Now my own thoughts on Tactical Profiling:
- When I first met Dr. Hector Ruiz, at the 2014 World Conference on Science and Soccer we, Hector, myself and Ben Knapper (lead statistical analyst for Arsenal FC) had some enlightening discussions on what new ideas might come forward after seeing how team performance analysis, using Possession with Purpose (compare and contrast the attacking team to the defending team), had shown trends that could help categorize team behaviors.
- Their tactical profiling approach begins to do that but it mostly ignores the defending team and what actions defenders take that may influence those seven different styles of attacking play/possession.
- Note that I offer ‘mostly ignores’ defensive actions.
- In truth, Tactical Profiling offers seven styles of play/possession not eight.
- “High Pressure” is not an attacking style of play/possession, it’s a defensive tactic – and it belongs on the other side of the equation.
- If Prozone Sports were to build a comprehensive style of play listing then they would need to include the low-block, bunkering-in, a high block, a single pivot, a double pivot (central defending midfielder schemes) or any other type of block involving three, four, five, six, seven, eight, or nine players.
- They don’t, and I think they saw that flaw; hence their creation of Game Intelligence. A way to begin to measure the un-measured statistics in soccer.
Game Intelligence; A product developed by Prozone Sports that is similar to what has been offered in the NBA (gravity). They have identified eight of them; here they are with their brief definitions provided by Prozone Sports:
- Defensive Balance: Automatic recognition of imbalances created by the attacking team
- Numerical Dominance: Identify overloads and underloads
- Defensive Press: Pressure applied by a defender to the attacker in possession
- Player Vision: Number of passing options available
- Player Attraction: Number of defenders synchronised to an attackers movement
- Player Space: Estimated region of the field a player has ownership over
- Time On Ball: Time of nearest defender to intercept attacker
- Offensive Channel: Quality of access an attacker has to goal
Now my own thoughts on Game Intelligence:
In short, this product is an attempt to address the weaknesses in current statistics – there isn’t a great way to measure the un-measurable. In particular the unmeasured statistics on the defensive side of the pitch.
Like Tactical Profiling I believe Prozone Sports have taken a great leap forward in doing this; but it still lacks.
A good example on why it lacks is the statistic called “Player Attraction”.
- Prozone defines “Player Attraction” as the number of defenders synchronised to an attackers movement.
- I would submit this isn’t a realistic metric as it assumes all defenders are, or can be synchronized, to defend against just one player.
- Even if they were, I would find it very hard to believe a Head Coach would be willing to offer that tactical approach publicly.
- So in returning to what Hector, Ben and myself discussed I would think a more appropriate title would be “Gravitational Pull”.
- And the definition would be ‘how to measure the space and time a player creates elsewhere’.
- This way that statistic isn’t prescribing a pre-determined coaching decision – it is only prescribing a team behavior that may be seen relative to one or multiple players versus another player.
- A more accurate (team) metric, working from this, is one that answers the question —> “how effective is the team in using that time and space to create or take a shot that winds up on target and in the back of the net”?
- I would call this metric (“Push-Pull). And the higher that metric the better.
- In other words the individual metric is “gravitational pull”; and the team performance metric is “push-pull”. Where the value of “gravitational pull” only has relevance where the team metric “push-pull” is high.
- After all – the true value of a player with high “gravitational pull” only adds value if the team is great at executing “push-pull”..
In conclusion:
- None of these new statistics adds value in tracking team performance unless the output from those individual statistics shows a positive or negative correlation to possession or penetration that then translates to a positive or negative correlation relative to shots taken, shots on goal, and goals scored.
- And like my analysis has shown in the article on Busting the myth of Moneyball statistics in Soccer – those correlations may differ from team to team and league to league – meaning one size does not fit everyone.
- For me this is the biggest fatal flaw of these two new approaches by Prozone Sports.
So what’s next?
I think the number of styles of play/possession by Prozone are excessive. Its’ simpler than that. For me there are only two styles of play (possession). They are:
- Possession with the intent to possess, and
- Possession penetration into the attacking third.
What does Possession with the intent to possess mean?
- This style of possession does not have an objective end state of penetration; the objective end state is to control the ball so the opponent doesn’t control the ball.
- This style of play occurs outside the attacking final third and many call it defending with the ball – I do too.
- Two Tactical Profiling statistics fall into this category – maintenance and build-up.
- I suggest four things can influence success or failure of this style:
- Opposing players who apply pressure in a certain area,
- The Head coach of the team with the ball,
- The Head Coach of the team without the ball, or
- Attacking players making in-game decisions based upon how they read the opponent’s actions.
- For me this is where newer statistics like Game Intelligence and Tactical Profiling come in – and where there are high positive or negative correlations of that data relative to this style of possession I then begin to narrow down what individual statistics best support those higher correlations.
- Given my Moneyball analysis (link provided earlier) I expect those correlations to vary from coach to coach, team to team, league to league, and year to year.
- In other words one size of statistic, team or individual, does not fit all teams in all leagues, year in and year out.
What does Possession Penetration into the Attacking Final Third mean?
- This style of possession has one objective – score a goal. If you aren’t penetrating there is no real intent to score a goal.
- In looking at what Prozone Sports has developed I think five of the other six styles of play (possession) fit here; direct, counter-attacking, sustained threat, fast tempo, and crossing.
- For me, the same four things that influence Possession with the intent to Possess influence possession penetration into the attacking final third.
- I should note here that PWP does not include dribble penetration. Why?
- Because most statistics show that a player only travels with the ball about 180 meters per game – versus traveling without the ball almost 6-8 kilometers per game.
- Meaning, on a histogram, ~85%-95% of the game involves passing; and since I don’t have big data I choose to ignore dribble penetration. It is what it is and if I had access to big data I probably would include it.
- Anyhow, the critical point here is determining what team statistics have a higher positive or negative correlation to successful penetration that results in more goals scored.
- This is where newer statistics like Game Intelligence and Tactical Profiling come in; where those next level of statistics show higher correlations to goals scored then you can begin to narrow down what individual statistics best support those higher correlations.
- As noted in my article on Moneyball statistics in soccer – those are expected to vary from coach to coach, team to team, league to league, and year to year.
- Again reinforcing that one size of statistic, team or individual, does not fit all teams in all leagues, year in and year out.
Back to my original question:
What are the best statistics available to better understand team performance and how individual players influence the outcome of team performance?
For me, the first statistics to help answer this question are those used in Composite PWP: The difference between how the opponent attacks against you and how you attack against the opponent. With these team statistics being the most important:
- Overall possession percentage
- Passing accuracy across the entire pitch
- Percentage of passes outside the attacking final third versus those within the attacking final third,
- Shots taken per completed pass within and into the attacking final third,
- Shots on goal per shots taken, and
- Goals scored per shots on goal.
Put differently – any or all of those Tactical Profiling or Game Intelligence statistics should be evaluated for how well they correlate (influence) these team statistics. If there is a strong negative or positive correlation to any of my PWP team performance indicators then they add value.
Next – what current individual statistics, in this list below, show a consistently strong negative or positive correlation to the team statistics above?
- Dual’s won/lost
- Tackles won/lost,
- Aerials won/lost
- Successful/unsuccessful dribbles,
- Successful/unsuccessful passes,
- Successful /unsuccessful crosses,
- Passing accuracy,
- Fouls,
- Yellow cards,
- Red cards,
- Total passes,
- Blocked shots,
- Interceptions,
- Recoveries,
- Off-sides,
- Clearances,
- Headers,
- Shot location (seriously – every one knows the closer you are to goal the more likely you will score – the same basic logic applies to basketball as well)
- Key passes,
- Assists,
- Shots taken,
- Shots on goal, and
- Goals scored.
In reality – none of them show strong consistency from team to team.
- In some cases combinations of those individual statistics do, like combining clearances, with blocked shots, recoveries, tackles, and unsuccessful passes, may show better positive/negative correlations but they still ignore the influence of un-measured statistics.
- In other cases, shots taken, or shots on goal, or goals scored, by individual players show good positive correlation but not on a consistently strong basis across all teams.
- And there are instances where shots taken, or shots on goal against may show good negative correlations to points earned. Meaning the higher the opponents’ statistics in those areas the more likely the other team is to earn points. Translating to a team that sits back and cedes possession is more successful in earning points than if that same team played a more possession-based game.
- All this circles back to the point about a need to understand both teams activities before trying to determine what individual statistics add or detract value in analyzing that team performance.
Here’s a recent example that may better describe what I mean.
- The Audi Player Index, sponsored and offered on the MLSSoccer.com web page identified Goal Keeper, Bobby Shuttleworth (New England Revolution) as the second best player in MLS last week even though their team lost 3-nil to Philadelphia Union.
- When I offered, in a tweet, that this didn’t pass the giggle test a few people responded by saying – hey – Shuttleworth had two PK saves.
- So I then asked – so what about the Union and the players who took those PKs?
- Might the fact that Shuttleworth made those two saves come down to the Union players offering poor PKs as opposed to Shuttleworth being solely responsible for those saves?
- The response (to paraphrase) was… well the Union guys have a history of not taking good PKs…
- So even though the Revolution lost their game 3-nil, and it’s likely at least one of those PKs was a poor attempt by a Union player, Shuttleworth still got individual recognition as being the second best individual player last week!?!?
In closing – this game is about two teams playing, not one; therefore all relevant team statistics, and subsequent individual player statistics must account for both teams activities, not just one.
- In addition, many of those team statistics can first be influenced by in-game decision making (mentality) by any player, coach or assistant coach, unplanned injuries, time left in the game, score-line/game-state, pre-game tactical formations in attack and defense, weather, pitch condition, referees, or playing at home versus on the road.
- So if I’m a senior leader/owner in a soccer organization don’t tell me the individual statistics my players have – tell me how the team statistics relate to the bottom line and then give me the relevant individual player statistics that directly correlate to those team statistics.
- Then I’ll know how my team compares to other teams, what types of players I’ll need to buy or sell, and how much value I put in the purchase of new players to make my team better.
- The un-offered fourth reason why I’d like to know this information – I’m an American, and I firmly believe taking this more in-depth approach will help our country compete at the international level. There is nothing more disappointing to a soccer supporter than to see their country lose a World Cup game.
- And seeing outputs from our most recent game against Guatemala it seems reasonable to me there are fatal flaws being made in tactical approaches as well as player selection and I wonder how much of that is due to misunderstanding the value of individual statistics relative to team performance.
- Finally, for what it is worth, Prozone Sports do not own the intellectual rights to the ideas behind those new statistical approaches.
For Prost Amerika Soccer readers:
- Future articles on this site will include PWP MLS Power Rankings – an objective view on who’s best/worst in team performance, both week to week, by conference, and cumulative rankings,
- On occasion I’ll offer up the latest in soccer analysis, and
- In-Depth team performance analysis on selected teams in MLS throughout the year.
If you have a favorite team you’d like me to review let me know. Comments are always encouraged! And retweets are awesome 🙂
Best, Chris
COPYRIGHT – All Rights Reserved. PWP – Trademark
You can follow me on twitter @chrisgluckpwp
My home site is www.possessionwithpurpose.com
