Shots Overall vs Shots on Target

Shots Overall vs Shots on Target

By Aaron Nielsen

In my last post I mention my dislike for the stat shots off target especially as a positive stat since it leads to the end of the offensive possession.  Although not keeping up with all the advance analytics in soccer I was unaware of the core of a new statistical evolution or fully aware of the stat TSR - http://grantland.com/the-triangle/what-is-total-shots-ratio-and-how-can-it-improve-your-understanding-of-soccer/

Last night I had a drink with a young and enthusiastic soccer and analytics fan who told me he had some modeling for predicting the most undervalued players in the MLS and wanted me to see his work.  Common in statistics he didn't want to show me much of his work and he basically had a list of players with rankings out of 10 including decimals. Aware of the players through seeing them play and my own work I inquired more how he came up with these evaluation. Basically he said he was a fan of the site whoscored.com but he disagreed with their rating system but used the stat "shots per 90" and "key passes per 90" without any raw data so it didn't surprise me a couple of the players on his list were players who only played a few minutes last season.

I don't use any OPTA work in my material which is the source of all whoscored statistics although since its available and OPTA is regarded as the official stat source. people have been drawn to the data and are using it in their own analysis. Although in comparing the data with my own it seems that "shots per 90" are shots on target + shots off target over 90 minutes of play well "key passes per 90" are the assist that lead to the shot (on & off) target over 90 minutes of play.

I don't calculate key passes in any of my work although I know in conservation with people in Spain that stat has been around for a number of years although there. they counted that as an assist and didn't count what we would normally regard as an assist. Personally its hard for me to get my head around awarding an assist on a failed offensive possession since it's not done in any North American sport but I can see some value if you were very specific on which play it should be counted.

Although awarding it on all shots off or on target makes no sense to me because your giving a player credit when more than half of these opportunities lead to a player missing the net and the opposing goalkeeper getting control of the ball. To further emphasize my point on how a miscue stat shots are to goals I went through the top leagues in the world and looked at the current top 20 in goals not including penalties, shots in total, and shots on target all based on the current data although the MLS is the complete 2013 regular season.

English Premier League
Players	Club	Goals	Rank	Shots (on & off)	Rank	Shots (on)	Rank
Suarez	Lvp	23	1st	118	1st	56	1st
Sturridge	Lvp	16	2nd	62	11th	32	3rd
Aguero	Mnc	15	3rd	70	4th	35	2nd
Remy	New	11	4th	69	5th	26	7th
Hazard	Chl	10	5th	53	19th	28	4th
Rodriguez	Sth	10	5th	69	5th	25	9th
Coutinho	Lvp	3	NA	71	3rd	23	13th
Barkley	Eve	3	NA	62	11th	13	NA
Paulinho	Tot	4	NA	55	17th	16	NA
German Bundesliga
Players	Club	Goals	Rank	Shots (on & off)	Rank	Shots (on)	Rank
Aubameyang	Dor	13	1st	49	14th	30	6th
Mandzukic	Bay	12	2nd	46	18th	31	4th
Ramos	HB	12	2nd	62	3rd	33	2nd
Losogga	Hsv	10	4th	53	7th	31	4th
Lewandowski	Dor	10	4th	78	1st	46	1st
Reus	Dor	5	NA	71	2nd	33	2nd
Kruse	Bay	5	NA	53	7th	26	8th
Hunt	WB	1	NA	46	18th	19	NA
Italian Serie A
Players	Club	Goals	Rank	Shots (on & off)	Rank	Shots (on)	Rank
Immobile	Tor	13	1st	54	16th	29	7th
Tevez	Juv	12	2nd	86	2nd	41	2nd
Palacio	Int	11	3rd	58	13th	28	10th
Toni	Ver	11	3rd	72	6th	31	5th
Llorente	Juv	10	5th	39	NA	23	17th
Balotelli	ACM	8	15th	101	1st	42	1st
Diamanti	Bol	2	NA	82	3rd	27	12th
Candreva	Laz	3	NA	74	5th	23	17th
Spanish La Liga
Players	Club	Goals	Rank	Shots (on & off)	Rank	Shots (on)	Rank
Ronaldo	RM	17	1st	160	1st	72	1st
Costa	AM	17	1st	76	3rd	43	2nd
Sanchez	Bar	15	3rd	46	16th	25	9th
Griezmann	Soc	14	4th	69	4th	38	4th
Benzema	RM	14	4th	58	6th	26	9th
Messi	Bar	11	6th	83	2nd	40	3rd
Rakitic	Sev	7	20th	62	5th	21	19th
Di Maria	RM	3	NA	48	14th	27	7th
Dutch Eredivisie
Players	Club	Goals	Rank	Shots (on & off)	Rank	Shots (on)	Rank
Pelle	Fey	18	1st	84	2nd	50	1st
Castaignos	Twe	13	2nd	66	4th	39	3rd
Finnbogason	Hen	13	2nd	71	3rd	37	4th
Promes	Twe	10	4th	49	10th	31	6th
Johannsson	AZ	10	4th	37	NA	24	18th
Depay	PSV	6	NA	95	1st	42	2nd
Tadic	Twe	4	NA	61	6th	29	8th
Chery	Grn	4	NA	49	10th	25	12th
Major League Soccer
Players	Club	Goals	Rank	Shots (on & off)	Rank	Shots (on)	Rank
Di Vaio	MTL	20	1st	118	2nd	53	2nd
Magee	Chi	17	2nd	114	3rd	50	3rd
Camilo	Van	17	2nd	123	1st	55	1st
Oduro	Clm	13	4th	92	8th	36	5th
Fagundez	NE	13	4th	55	NA	29	10th
Bruin	Hou	8	NA	98	5th	30	7th
Higuain	Clm	5	NA	86	10th	31	6th
Zardes	LA	5	NA	78	12th	23	6th

From my view there is no case where shots in general equals goals better than shots on target does as well based on the examples there is many cases where players have a large amount of shots but scored very small amount of goals. And in almost every case a player in the top 20 of scoring ranked higher or equal in shots on target than he did on total shots.

This is true in team data as well as Juventus was the top shooter in the Champions League with 101 shots but only scored 9 goals, Tottenham and Newcastle have 40 more shots than Arsenal in EPL this season. Braunschweig has more shots than Schalke in the Bundesliga yet scored 26 less goals and in La Liga last place Real Betis has more shots than league leaders Atletico Madrid.

This is a trend I been seeing since I started tabulating data for soccer including the most extreme story in Japan J-League in 2001 when Brazilian Wanger Lopes played for relegated side Avispa Fukuoka and had 186 shots in only 34 game season although only 41 were on target and seeing other players with shots in the 30 but only one or two on target I decided to stop recording shots and concentrate only shots on target and in my analysis going over a number of seasons I'm confident with that choice.

In truth some of this new analytics have been around in football for many years with people involved with clubs telling me they were looking at play per 90. shooting analysis, and even expected goals since the early 2000's although I'm encouraged by the enthusiasm by people especially younger ones to look at statistics and analytics. This being said I think there is a fear of group mentality where everyone jumps on idea and that stat becomes a new trend and even in North American sports there have been ratings and other forms of data analysis that were loved at one point but lost its value when it really made no difference on the field. So I love the sharing of ideas but I feel proof of concept is where the value is.

My acquaintance and I at the end of the night agreed to disagree his view was their was a lot more work online proving his concept over my concerns and was dishearten by my initial response as he thought I could introduce the work to MLS clubs in which I told him his best option was to approach OPTA since he was using their work to produce his analysis.  

Although when I got home I felt it unfair just to dismiss a concept especially one that is popular so hopefully this sample of data shows my point and I can include many more samples if wished. Although my footballing view on the matter is - Do I want a player who is willing take a shot? The answer is no, what I want is a player who can score and taking shots in my view doesn't mean your going to score especially if a large percentage are off target or easy saves for the keeper.

Creating a Flowchart of a Goal in Soccer/Football

Creating a Flowchart of a Goal in Soccer/Football

By Aaron Nielsen

After the recent OPTA pro forum, I posted some comments including my proposed paper - Opta Pro Research Paper I got a decent amount of back and forth on my twitter account @ENBSports including links to other work. What I noticed from others is that the MCFC data released in 2011-2012 by OPTA is having an influence with writers trying to take advantage of the 211 columns of data. Personally the only thing I really looked at when it came out were free kick shot opportunities - Direct Free Kicks as data for the la liga was also available through Marca magazine so I thought it created a good comparison piece of analysis.

My main concern with the detail set is the amount of cost/time to tabulate this depth of data and what can be gained from it. So in tabulating my own data I broke down over time what I thought were the most valuable statistics so I'm now able to cover over 60 leagues of data as have been tabulating data for over 20 years. Although now with more detailed statistics becoming a bigger part of the everyday conversation including most recently Statsbomb.com - Expected Goals, I decided to reopen my MCFC excel sheet and reexamine the data from the 2011-2012 English Premier League season and see if I can find specific information from it that I think could be used in understanding the game better through statistical data.

I thought my first attempt was using the data if I could work out a proper flowchart to a goal. One reason for this is that one of my biggest issue in soccer is that most people including statistician see shots (including missed and blocked shots) as a positive stat. Its true almost all goals come from a shot via a foot or head although they also come from a shot on target (on average around 30% of all shots) well most shots off target lead to a restart from a goalkeeper and an end to the offensive possession. So in recording data of less detail I've decided to only include shots on target as shots and ignore other shot attempts as I feel shots in total gives a poor representation of the players success assumed by this stat. Although since OPTA makes it available I've decided to use it as part of this analysis in help breaking down each offensive possession.

I start in looking at the flowchart of a goal with offensive possession which can either come from Open Play, Penalty, Free-Kick (Direct or In Direct), Corner, or a Throw-In. All goals including own goals fit under each applicable category and an opponent’s defensive turnover leads to an open play on offensive possession even if the offensive position only includes a shot. OPTA doesn't record each individual offensive possession in the stats but does show 1025 goals, 10891 total shots which led to 10496 Goalkeeper Distribution and 22555 clearances that stats do not include other forms where we can assume lost of possession but one could estimate in total there were about 40,000 offensive possession or a goal in every 40 or so attempts at possessions.

OPTA does a great job in analyzing penalties, free kicks and corners so I will start there.  First in terms of free kicks OPTA judged that there were 1884 Free Kicks in a dangerous area not including penalties from these free kick opportunities there were 854 that lead to a shot including 553 direct attempts on net that led to 29 direct goals or one in every 19.1 attempts. OPTA did record shots information but in terms of a flow chart it’s difficult to analyze the play because we don't know of the play conclusion on a direct attempt that did not lead to a goal. Of the 1291 free kicks passed instead of direct shot 301 lead to shots via a key pass and 50 goals or a goal in every 25.8 attempt although again we have no information on plays that didn't lead to a goal. Meanwhile there were 100 penalties with 72 converted, 23 being saved and 5 were off target.

In terms of Corners, OPTA has a total 4321 corners that lead to 129 goals inside the box or a goal in every 33.5 attempt. Of these Corners 3496 were attempted in the box with 1163 being "successful" which I assume means lead to a shot and 377 of those shots were on target. There were 663 short corners although OPTA gives no detail on any further actions from these short corners. Meanwhile throw-ins lead to 20 goals from inside the box, 67 shots on target and 184 shots off target although we don't have information on how many throw-in were attempted in that zone or again what happen to the play on corners and throw-ins outside of ones that lead to goals and shots.

Which leads to open play opportunities, unfortunately with some of the data missing above we can't assume all the open play data. The data shows there were 725 goals not counted above so we can assume they were from open play and of these goals 577 goals were inside the box and 148 from outside. We can't breakdown the inside shots due to missing data but from the outside minus direct free kicks 1164 were on target, 1395 were blocked and 1819 missed the net. Overall there were 3522 shots in total on target, 2902 were blocked, and 4467 missed the target and we can also assume most of the shots that missed the target lead to an opposition restart.

The problem with the flow chart concept is we’re missing key stats related to creating data that OPTA does record for example of the Shots on Target we don't know how many of them were controlled by the goalkeeper to change over possession. So I would suggest that OPTA starts recording the stat rebound and save leading to corner so we have a better sense of shots on target that do not count as a goal. We also don't have any information regarding an offensive player mistake in the offensive zone for example a foul, wayward pass, or loss of possession out of bounds so we can only assume when the offensive team losses possession by looking at clearances and goalkeeper distribution which we assume the offensive possession is over since the keeper has control of the ball.

There is also an issue in my view with the stat clearances and as a defensive statistic the assumption like shots is that this is a positive play although we have no further information regarding the clearances which affect the flow chart concept and also like total shots a misleading stat. I'm assuming clearance includes kicking a ball out of bounds and giving the opponent the ball as a throw-in or corner.  If so well looking at the goal per attempt stat this could be problematic because with only the data in front of us the ratio in terms of how many attempts per goal is lower for a corner or a throw-in then in open play. So analytically a clearance could actually increase his opponent’s chance of scoring depending on the clearance.

Ultimately for a flow chart you would want the play and then the lead in to the next possession be a goal, the offense getting another offensive possession via a rebound, clearance, or set play opportunity or the defense controlling the ball and putting an end to the offensive possession. Based on this we can start looking at general views in terms of possession but also specific for example if a particular player tends to kick the ball out for a corner well under pressure a team can use that to their advantage when breaking down opposition analysis. Or alternatively if a team likes to shoot maybe allow it because you know a chance to score is poor and a shot off target leads to a loss in possession.

I'm impressed with OPTA work and with a background in most sports it amazes me the effort they put into collecting it especially when traditionally there is a long history of stats recorded for North American sports in far less in detail. Although I must say that when breaking down a sport I find much more useful, traditional North American sports data. And in doing my own soccer stats is the same because I know what I want to get out of it before I start to tabulate the data. I fight with teams and leagues all the time over the value of stats in soccer and to dismiss my work they show me huge books of animated drawings in what they believe breaks down every play in soccer. Although the reality of what breaks down every play is actually what really happens over a good number of samples which statistics can do, all we need to do now is prove it through our work.