Twitter

Sunday 26 April 2015

Week 34: Regression Of Scoring% & Save%

Key: Scoring% = goals for/shots on target for *100 | Save%  = 100-goals against/shots on target against

Just a helpful reminder about how Scoring% and Save% behave:



Early season scoring and save percentages are all over the place after 5 games (hit the play button), things tighten a little by the 10 game mark. At the 19 game mark things are pretty tight and the spread between teams has decreased.

At 34 games the spread is this: lowest save% is 61.5%, highest 74.8%. Lowest scoring% is 22.6%, highest is 36.5%.

The spread of scoring and save percentages isn't trivial, there is a real gap between the ability of teams to score and save the shots for/against on goal but over the course of the season most teams' conversion percentages are regressing toward a mean.

Not the mean, but a mean. Some teams may well exhibit an ability to maintain high or low conversion percentages over the course of what is, if we think about it, a pretty short season (38 games). The teams who maintain high or low conversion percentages may well do so due to shot location or luck, or score effects, or talent disparity. 

The make-up of scoring or save percentages contain many different variables, part skill and part luck, but the motion chart above gives us a quick and dirty example about how scoring and save percentages regress over the course of a season.

"Uh-Oh" - Mike Ashley

Definition of uh-oh:

"An expression of alarm, dismay, or realization of a difficulty". This is a quote that likely perfectly sums up Mike Ashley's current mood. His Newcastle club, seemingly so safe on in January when Alan Pardew was finally allowed to leave for London, is now just 5 points out of the relegation places. 

Four fixtures remain. A 5 point cushion.

"Alarm".

"Dismay".

Now, when Mike Ashley made the decision to appoint John Carver as interim manager until the end of the season, the prospect of relegation from the Premier League was unfathomable, almost impossible, some might say. Newcastle, at the time of Pardew's exit and Carver's appointment, were in 10th place - 7 points off the Champions League places & 10 points clear of the relegation places - and a gentle, sedate, and somewhat boring end to the season seemed to be the most likely future path for Newcastle. 

Calm waters, clear skies.



But those clear skies and calm waters were not to last. Newcastle under John Carver proceeded to take just 9 points from 15 games. 

2 wins, 3 draws and 10 losses. NINE freaking points in what looks very much like a controlled descent toward danger and potential doom.





These are Carver's 15 league games:






Just the 2 wins as mentioned and they came against poor Hull and Aston Villa teams. 

John Carver's detailed numbers from these 15 games:



 


John Carver's Newcastle aren't a terrible shots team (48.8%) but the shots on target numbers are brutal and no team can expect to pick up points on a consistent basis whilst taking just 38.8% of those on target shots. Injuries are also an issue for Newcastle as is a shitty PDO number (Scoring% + Save%).

Poor processes (SoTR) and some bad luck (injuries and some of the PDO) are a disastrous mix for any club and right now this combination is sinking Newcastle's season and could possibly place the club in some real danger.

Yet Newcastle and Ashley still have that 5 point cushion over the bottom 3. 

5 points, 4 games remain:



"Uh-oh".

Leicester & QPR as the remaining away games? Oh boy. A single win should be enough to secure Newcastle's Premier league status but where would such a miraculous result come from? Home games against West Brom and West Ham would seem to be the most likely and winnable fixtures but this Newcastle team is struggling and the home fans may well be nervous. 

And if I were a Newcastle fan I too would be nervous:






Newcastle are flat, other relegation threatened teams are picking up points.

I am leaning toward Newcastle retaining their Premier league status and why wouldn't I be when Newcastle have that points cushion with so few games left to be played. But this Newcastle team is sinking and their points cushion is being chipped away by the teams below who are picking up points .

Seven consecutive league losses now for Newcastle. Things are starting to get a little too close for comfort as the graph above illustrates.


"Uh-oh" - An expression of alarm, dismay, or realization of difficulty.








Saturday 25 April 2015

Cyborgs & Other Stories


Cyborgs

What is a cyborg?




What do they look like?



Scary. 

What do they look like with the human skin cover?




Scary. And Chelsea have quite a few of these guys!

The graphic below contains the 34 Premier League players who have played over 90% of the available minutes in 14/15. The list contains a ton of defenders (19/34) but the really juicy bit of info is that list contains SIX Chelsea players. Southampton have 4 players on that list, no other team has more than 2 players.

Health matters. It matters even more if the players that stay healthy also happen to be your core guys on the team. This health allows for some continuity in selection for key areas of the field - 3/4's of Chelsea's back four is on this list as is 3/4's of their midfield.

Cyborgs, man. Amazingly talented, injury-proof Cyborgs.

Cyborg Graphic




Manchester City Injuries

As mentioned, Chelsea are not just a ridiculously healthy club but they have benefited from consistent contributions from their core players. Manchester City, whilst not exactly injury-riddled, have not been able to lean on their core group to the same extent Chelsea have. 

Yaya Toure leads the club with 76.9% of the minutes played (top%), Gael Clichy is in 2nd place, Demichelis is 3rd, Jesus Navas is 4th, and Pablo Zabaleta rounds out the top 5. That list contains good players, but aside from the imperious Toure it looks nothing like the kind of core group that a team can rely on. (Manchester City Minutes LINK)

Chelsea's core 6 have played 93.8% of the Premier League minutes available. Manchester City's core of Zabaleta, Kompany, Toure, Silva, Nasri and Aguero have played just 66.2% of the Premier League minutes available.

Points is: Injuries and absences matter. Most Premier League teams have enough depth to cope with a certain amount of injuries but significant injuries to the core of a team can impact on performance and outcomes. And it would be silly to suggest that Manchester City haven't been impacted by the injuries to core players.

A question: If Manchester City's core 6 had played 93% of the available minutes where are they in the table? How many points do they have? How many pundits would be talking about a great City team instead of fawning over the 'invincibles'?

The Core, As A Gang Of Four

Anyone remember this?
If I am not mistaken I think around 45 minutes was added to that total on the final day of the 13/14 season. Still, it is a pitifully low number of minutes and Manchester City, quite simply, cannot rely on their very best players to be available for selection on a consistent basis.

Now, if we remove Nasri and Zabaleta and say that Manchester City's core contains just Kompany, Toure, Silva and Aguero wouldn't be interesting to see how many minutes these guys have played together in the Premier League this season?

504 minutes or 24.5% of the minutes available.


Chelsea's Core:













Minutes played by this Chelsea core 4? 1521 minutes or 76.8% of the minutes available.

Yaya Toure

For all the talk in the previous point about the impact of injury to Manchester City's core, it is becoming clearer to some that the most important player for the club may well be Yaya Toure. 

Much maligned at the start of the season due to the ludicrously named Cake-Gate episodeToure started the season sluggishly, misplacing passes, lacking spark, and most importantly for those media types, Toure's goal scoring rate had dried up from its unsustainable 13/14 level.


Despite his dip in form and some early signs of slowing in his game, Toure remains the most important Manchester City player in my opinion. And the Arsenal result hasn't hastened the author to reach the conclusion. Some tweets from over the Xmas period:






And this: 

A 'controller' is probably the most accurate way that I can describe the man who Manchester City miss so much when he is unavailable. Toure controls play from deep: he is an outlet pass option for defenders under the press, he is the link between the defense and the Silvas and Nasris of the world and most importantly he is the deep creator when Manchester City have the ball around the oppositions 18 yard box. Toure is City's unflappable quarterback.

He is not a player without faults, though. Toure can be suspect defensively, he can fail to track his man, he can be lethargic in the checking back but how much of this matters when in the context of the usual City game? In most games City's opponents have around 40% of the possession and only about half of that opposition possession will be in areas where Toure's suspect defensive work can be exposed. 

Complaints about Toure's defensive work are valid but those concerns occur relatively infrequently and when they are placed against all the crucial linking and attacking skills that Toure displays for the majority of games a sensible person is likely to come to the conclusion that Yaya Toure is the vital player for Manchester City.

Good Lucky Matrix

Green=Top 4. Red=Relegation. Size=Points
The further right a team is the higher their share of the shots on target. 
The higher a team is the higher their PDO is (PDO = scoring% + Save%)


Everton

Everton have taken 50% of the shots and 57% of the shots on target so why only the 23 points from 22 games? Their keepers can't make a damn save! 

Everton have conceded the 5th fewest shots on target (link) with only Chelsea, Stoke, Man City and Southampton having conceded fewer. The problem: of those 74 shots on target conceded Everton have allowed 34 goals. That leads to a pitiful 54% save%, which is way below average. 

Look:
 

Save% is just one of the issues troubling Everton and their increasingly beleaguered manager, Roberto Martinez, but this is a team that takes 57% of the shots on target and scores just 46.9% of the goals. 

The evil, oft-hated PDO is always the reason for the gap between SoTR and PDO - be it luck or systems. But right now the manager killing PDO is crippling this Everton team. Thing is, Save% and PDO have history with Martinez and not the good kind of history.

2 points in 6 games. 1 win in 10 games.

History





Experiments

Race For The Top 4




Ronaldo & Messi

Goals, assists & Points (goals+assists) on a game by game basis.

Goals on the x axis, assists on the y axis, Points indicated by the size of the bubble.


The Most One-Sided Game In Premier League History

The Most One-Sided Game In Premier League History

By one-sided game I mean one-sided shots count. And I am just guessing at the IN PREMIER LEAGUE HISTORY bit, although Manchester City's title winning game, where the home side outshot QPR 44-3, is likely to be the most one side game on record.

Manchester City took a shot against QPR's goal every 131 seconds during this game. The then soon-to-be-champions were shooting at a ferocious pace in this game and as the chart below illustrates score effects and the sheer urgency of the situation led to ever more frequent shooting from the home side.

The second half of the game on that title-winning day saw Manchester City take a shot every 90 seconds. Staggering stuff. It very nearly ended in tragicomic farce but for that goal




Full Size Chart (LINK)


That Goal


Height & Weight Of Premier League Footballers (2014/15)

Height & Weight Of Premier League Footballers (2014/15)

I don't really need to explain this too much.

I have taken the height and weight of every player to have played in the league (as of November the 12th, so a few players will be missing). I then split those players into three buckets: Defenders, Midfielders and Forwards. Goalkeepers are removed from this data.

This table lists the average height (cm) and weight (k) of each bucket of positional players for each and every Premier League team.

Table

As always, I publish the data for everyone to use, sort or scrape.



Ave Height (m)Ave Weight (k)D Height (m)D Weight (k)M Height (m)M Weight (m)F Height (m)F Weight (k)
Arsenal180.672.4182.574.9176.667.1182.775.2
Aston Villa181.674.418678.6177.168.4179.575.5
Burnley180.674.5182.175.1177.772.3181.775.8
Chelsea183.679.9186.885.2183.275.6181.178.4
Crystal Palace180.974.2186.978.9176.368179.875.5
Everton18075.4180.475.2179.975.5179.675.4
Hull City181.174183.174.9178.171.5182.876.4
Leicester18175.2181.778.717874.2182.473.3
Lge Ave181.575.118477.2179.272.6181.174.8
Liverpool180.574.7182.676.1177.473181.374.7
Man City180.475.618276.1178.174.318377.7
Man United180.472.9183.375.9178.570.5179.772.9
Newcastle181.173.2181.174.8177.87318171.9
QPR183.678.1185.977.4181.476.5183.581.3
Southampton181.875.7183.778.4179.872.2181.476
Stoke183.475.8185.578.3183.576.3181.473.3
Sunderland182.775.318676.518072.7182.477.4
Swansea180.674.2183.676.318375174.270.6
Tottenham181.974.8182.975.618274.3180.674.3
West Brom181.575184.877.9176.869.7181.776.5
West Ham182.876.4184.677.918275.318276.3


A fun little graph to finish things off. Guess who the 75 kilo outlier is?

A Collection Of The Worst Shots Differential Teams On Record

A Collection Of The Worst Shots Differential Teams On Record

The teams I will be looking at today are some of the most incompetent and downright terrible teams at controlling the shot count. Whether it was a weak offense or a porous defense all of the teams that are listed below were consistently and heavily outshot over the course of their respective seasons. Teams that are consistently outshot tend to find trouble. Trouble is defined here, at the very least, as flirtation with relegation.

The Teams


  • The mythical Derby County team of 2007/08.
  • The 7 worst shot differential teams of the Enlightened Stat Era.
  • 2 teams from the 14/15 season who are, as of game week 9, on course to become the worst and 3rd worst shot differential teams of the Enlightened Stat Era (2009 to present)

The Table

Sortable 


YearPOSTEAMPWDLFAGDPTSWin%Points%Goal%TSRShots +/- Per Game
2014/1510*Hull City925213131122.240.75034-8.8
2012/1319Reading38610224373-302815.824.637.135.8-8.2
2014/1517*Leicester92341115-4922.233.342.335.8-7.7
2009/1019Hull City38612203475-413015.826.331.236.3-7.6
2010/1118Birmingham 38815153758-213921.134.238.937-7.4
2013/1420Cardiff City3879223274-423018.426.330.237.7-7.2
2013/1413West Ham 38117204051-114028.935.14438.3-6.7
2007/0820Derby County3818292089-69112.69.618.337.4-6.2
2013/1419Fulham3895244085-453223.728.13239.7-5.8
2012/1317Sunderland38912174154-133923.734.243.239.5-5.7

These are the worst shots teams I could find. These are the incompetent bastards, the terrible teams who were consistently pulverized by the opposition in terms of the shots count. 

We have 10 teams on this list including two current teams: Hull and Leicester. All 10 teams had a TSR/Corsi (% of the shots count) of less than 40%. Six of the eight teams from the previous seasons were relegated and Sunderland in 12/13 should of been but for Di Canio and some luck. That Sunderland season is now known as The Costly Miracle. Why? At around Week 4 0r 5 of that 12/13 season I managed to get Sunderland at 18/1 to be relegated. Yeah.

So what does this table tell us? None of these teams were able to control the shots count and this led to pretty poor points totals, low win percentages and, in many instances, relegation. 

If you look at the Goal% and TSR column we can roughly guess at some of the PDO's these teams posted. If Goal% is noticeably higher than TSR then that team was posting a fairly high PDO. Birmingham, West Ham and Sunderland all posted fairly high PDO's and this likely helped those teams post higher points totals than their control over the shots count may have warranted.

Control of the shots count isn't the only important counting stat but it is a quick and easy was of identifying which teams may have poor processes. If your team is being outshot by the type of margins as the teams we see above it likely means that that there are issues with the ability of that team to control games. This may mean a team is unable to prevent the opposition from hemming them in their own end. It may mean a weak midfield, or a porous defense or an attacking scheme unable to maintain pressure in the offensive zone. 

In Reading's case, it meant all of these things!

What Does This Mean For Hull And Leicester?

Hull and Leicester are posting terrible Shots +/- numbers so far this season and the defense is to blame for both teams. Now, you may be sitting there reading this and thinking "Hull are 11th and Leicester are OK in 16th" and that is fine. Both teams have had decent-ish starts to the season, points are on the board, not too much to worry about. But there are things to worry about.

These teams cannot continue being butchered by the shots count to this extent and steer clear of any relegation trouble. Hull are currently being outshot by 8.8 shots per game and if these numbers were to continue then they would become the 2nd worst shots team on record (last 6 years). Leicester are being outshot by 7.7 shots per game and would become the 3rd worst shots team on record. 

These teams wont maintain good form whilst being outshot by these margins. Now, these negative shot differentials aren't really caused by score effects and shelling with an early lead (minutes winning/losingScore Adjusted TSR) they are more likely caused by poor processes, some tough schedules and a tiny bit of score effects.

There are two likely paths that Hull and Leicester may take:

1) The shots numbers continue to be terrible but their form goes from OK to bad as their PDO numbers regress toward their individual mean. Right now both Hull and Leicester are posting PDO numbers of above 110. It is very unlikely, although not impossible, that these teams can maintain those high PDO numbers. What is more likely though is that those PDO numbers regress back towards somewhere around 100. Hull and Leicester may well be the type of teams that can maintain high scoring and save percentages (PDO) on the back of their + talent, some luck and some score effects boost from spending lost of time in winning positions, but those words don't really past the smell test. 

If the shots ratios and the inability to control the shots count remains the same and their PDO's slide back a little toward normality then their respective Goal% numbers will begin to sit around their TSR numbers, which will be around 40%. From 2001 to 2014 51 teams posted a Goal% number of less than 40%. Their average points total was 34, average goal differential was -27, and their average position in the table was 18th.

Those are not good omens.

2) Both teams stem the flow of shots against, exert better control over games and reduce those horrible shots differential numbers. This is entirely possible by the way. Teams may have a softer run of games against weaker opponents than they have generally already faced, they may have key players back from injury, the manager could tweak tactics and personnel. Offensive schemes can be fixed. The lateral and vertical gaps in bad defensive schemes can be tightened and improved. It happens. It has happened before and it can happen again.

Final Thoughts

I worry a lot for Leicester and slightly less so for Hull. Why? Well, although I think option #2 may happen and those teams will post better shots numbers going forward I still think option #1 is far more likely to happen. I am not convinced that either Leicester of Hull can maintain above average scoring and save percentages over the course of the season and couple that with relegation type shots numbers and it can only lead to trouble.

It would be mighty wise for both teams to focus on the structural or personnel reasons that lead to being outshot by such margins. We know that both teams are being heavily outshot but why is that happening? What are those teams doing poorly, what are the issues? 

Once they have been diagnosed then the how becomes the focus. How can those weaknesses be strengthened. How can the processes be improved. How can the little, almost imperceptible things that lead to such ugly shots numbers be corrected and improved.

Those are jobs for the technical personnel and the management team.


The Worst Of The Worst



The 8 worst shots teams on record (and Hull & Leicester!)

A History Of Premier League Title Winners & What It Means For Chelsea

A History Of Premier League Title Winners & What It Means For Chelsea

During this cold and rainy evening in what is normally the mild and bright month of December in Barcelona, I decided to clean up a couple of tired and messy databases that I am in possession of. During this clean up I realized a couple of things: I only watch sports so I have an excuse to use a spreadsheet and I have data on all worlds of weird and wonderful topics.

One of the interesting topics was a database titled 'The History Of Premier League Title Winners'. This database was opened and there on the page in front of me was every Premier League title winner in the 38 game era and each teams cumulative points total over the course of the season.

Why was this database interesting? Well, let me show you....


(Bigger Link)

So, anyone see anything interesting there? I have highlighted Chelsea over the first 13 games to make it easier for you. Still nothing? Well, much to my astonishment I realized that there have been teams that secured more points over the first 13 games of a season than Chelsea have this year (33).

Three teams have achieved this feat: Manchester City in 11/12, Manchester United in 06/07 and Chelsea in 05/06. It is also worth noting that 2 other teams have matched Chelsea's haul of 33 points over the course of the first 13 games: Chelsea in 09/10 and Arsenal in 03/04.

The 5 teams mentioned were all darn good. The points totals for this gang of five range from 86 to 91 come the end of the season. We could just leave this mini-investigation here and say 'hey, teams with 33 points from their first 13 games are all really good teams and end up winning the league in that given year'. Yet this statement would ignore the teams who had 33 points or more from the same number of games who didn't win the league.

So I dredged through some old Premier League tables and found 8 teams who had 33 points or more from the first 13 games. Why did I go to such trouble? I wanted to see how those teams performed over the remainder of their respective schedules after they had gotten off to such hot starts.

Here are those 8 teams with their points totals in the first 13 games and their points totals over the remaining 25 games.



PointsFirst 13 GamesNext 25 GamesPPGFirst 13 GamesNext 25 Games
Arsenal 03/043357Arsenal 03/042.542.28
Chelsea 05/063457Chelsea 05/062.622.28
Man United 06/073455Man United 06/072.622.20
Chelsea 09/103353Chelsea 09/102.542.12
Man City 11/123554Man City 11/122.692.16
Man United 09/103457Man United 09/102.622.28
Arsenal 07/083350Arsenal 07/082.542.00
Newcastle 95/963447Newcastle 95/962.621.88
Average33.853.8Average2.602.15

The average points haul over the next 25 games is ~54. This number is dragged down by a single point by Newcastle's famous collapse in the 95/96 season. If we look across the table we can see that every single team saw a drop off in points per game in the next 25 games of the season.

Why did this happen? 13 games is a short span where luck can run hot, injuries haven't yet taken their toll, "form" is excellent, the schedule is soft and forgiving and all these things may have contributed to a points pace that was likely unsustainable over the remainder of the season. Injuries, fatigue, cooling off of some hot luck and a stiffer schedule may have led to a drop-off of the points pace.

The average points per game pace of this group fell by 0.45 points per game over the remaining 25 games of this season. In real terms this is a drop off of 11 points from any points projections that may have been made at the 13 game mark.

11 points is a significant number: Chelsea are on course for 96 points as it stands. The above table, while not concrete evidence, at least hints at some form of regression in terms of points per game pace.

Let's say Chelsea win 54 points from their remaining 25 games to finish on a total of 87 points from 38 games. Does that sound reasonable? It feels a touch low but if true then Chelsea's record over the next 25 games would look something like this:

Chelsea
Win 16
Drawn 6
Lost 3

Right now it seems impossible that Chelsea would drop points in 9 of their next 25 games but 54 points is the ball park figure for a team who recorded 33 points in their first 13 games. But you know, as much as that W/D/L line looks a little crooked it still results in Chelsea recording 87 points. 87 points is a lot of darn points and good enough to win the Premier League in 13 of the last 20 seasons.

You know what else 87 points is? Almost impossible for Manchester City to match.

Manchester City currently sit on 27 points from their first 13 games. It's a decent haul but it's a ways behind where Chelsea currently sit. This is the record Manchester City would have to post to get to the 87 points I have projected for Chelsea:

Manchester City
Win 19
Drawn 3
Lost 3

Now that is a difficult ask my friends. How on earth could Manchester City pull that off? I could find only six teams in the last 20 seasons who had 60 points or more from their last 25 games. Do you want to know who one of those teams was? Manchester City in 13/14: 61 points with a W 19 D 4 L 2 record.

With Manchester City's soft schedule over the next seven or so games I'd expect the papers/media to begin talking up Manchester City and chipping away at premature bullshit like this:



Friday 24 April 2015

Interactive: Leo Messi's Goal Contribution% In La Liga (2004 to 2015)

Most things are explained inside the actual chart. Goal Contribution % is the the player's goals and assists/number of team goals when on the field of play.

The point of posting these numbers is to see if we can capture any trend, positive or negative, in the amount of Barca's offensive output that Messi has contributed to.

The trend here is pretty clear: amazing growth and increased contribution to Barca's offensive output for a number of years. But the growth was slowing by age 24 or 25. Contribution peaked at 25/26 and actual dipped last season.

Messi is posting a very strong 56% goal contribution number so far in 14/15. We must remember that although Messi has fallen from his own previous peak, his goal contribution numbers remain outstanding. It's just, you know, very likely that Messi's influence and contribution to this Barca offense has peaked. He is likely not improving anymore but actually regressing for a number of reasons. It is possible that we have already seen the very best years of Leo Messi.

He is, and will remain, an outstanding footballer who contributes at a truly elite level.




Link 


Nationality Of Premier League Footballers (2014/15)

hover test
This is a pretty straightforward post really. The only rule I put in place in terms of eligibility was had a player actually played in the Premier League this season (14/15). Some guys have been injured all season, some guys just haven't played yet. Tough luck.

54 nationalities, 463 players.

42.1% are from the UK. 48.1% are from the UK & Ireland. 77.5% are from Europe.


Interactive Map

Map is fully interactive.


 
Full Screen Version (LINK)


Nationality By Team

Table is sortable by nationality and by team.


ArsenalAston VillaBurnleyChelseaCrystal PalaceEvertonHull CityLeicesterLiverpoolMan CityMan UnitedNewcastleQueens Park RangersSouthamptonStokeSunderlandSwanseaTottenhamWest BromWest Ham
Algeria11
Argentina112421221221
Australia1111
Austria11
Belgium1221122131
Benin1
Bosnia111
Brazil422211
Cameroon11
Canada1
Chile12
Colombia1111
Congo1
Costa Rica111
Ivory Coast11111
Croatia112
Czech Republic11
Denmark111
Ecuador1
Egypt11
France512111146112141
Germany311111
Ghana1
Greece1
Iceland1
Ireland2324511121411
Italy212
Jamaica111
Japan1
Kenya1
Mexico1
Montenegro1
Morocco1111
Netherlands2241112
Nigeria122
Norway1
Paraguay1
Poland111
Portugal1111
Romania111
Senegal121113
Serbia2121
Sierra Leone1
Slovakia1
South Africa1
South Korea11
Spain43333121211
Sweden111
Switzerland11
Togo1
United Kingdom710162169131383138111059991212
United States1111
Uruguay211
Zambia1

If any readers spot any errors or over/under counting please let me know.