Ask anyone to measure the efficiency of a striker today and invariably you will receive a reply which will include a metric that is linked to shots on target.  Of course any such measure will be more accurate than those which are based solely on the number of shots that a player took, however that doesn’t mean that is the best measure by which to measure the effectiveness of strikers.

That opening thought was prompted by the following tweet from Squawka Football that I read this morning

Squawka

So Adebayor had the best shot accuracy, yet he only scored a handful of goals and I didn’t recall anyone last season remarking on well the big forward was striking the ball.  Although 70% of Adebayor’s shots found the target (which is the traditional method for defining accuracy) he was missing some va-va-voom which would have allowed more of those shots to turn into goals.

At this, I dusted down my trusty database and had a look at the shots that Adebayor took last season:

Adebayor Shots

Adebayor Shots

We are looking at the goal from the view of the striker, with the two black rectangles representing the frame of the goal.  The red balls signify shots that resulted in goals, and the white balls are shots that either missed the target or were saved.

Now, we can begin to understand why despite having such a high proportion of shots on target that Adebayor didn’t overly work the scoreboard operators across the country last season.
Virtually all of his on target shots were hit straight up the central channel of the goal.  In fact I only count five on target shots which were directed towards the corners, of which two were goals.

If you read my piece from yesterday you will recall that, all other things being equal, a player can only expect to score from shots which are hit straight up the middle about 12% of the time.  That compares to 50% or more when the shot is tucked right into the corners.
I would contend that the metric of shots on target as the measure of choice for determining player accuracy is flawed such is the rarity of centrally struck shots becoming goals.  I believe it’s time to start thinking of centrally struck shots as relatively bad things, not good things; or at the very least that not all shots on target are equal.

Rafael van der Vaart and Darío Cvitanich

Here are the stats for last season for two players, one of whom is much better known than the other.  We have the Hamburg and former Tottenham Dutch man Rafael van der Vaart and Dario Cvitanich a player without any international caps who plays for Nice in Ligue 1.

VDV Comp

The two players had virtually identical numbers in terms of shots and shots on target, yet we can see that Cvitanich had a conversion record for his unblocked shots which is worthy of another planet.  His total of 19 goals was enough to see him finish as joint runner up to Ibra in the French Ligue 1 goal standings table last season.

Although the two players played in different positions, and thus I have no doubt that the types of shots that each player had were not comparable, both players would have identical accuracy ratings per the current accuracy metric.
Based on this bare statistic, that both had an “accuracy rating” of  approximately 60%, we are left to wonder why one player scored almost four times as many goals as the other.
We might be forgiven for thinking that it was due to variance in that the Van der Vaart faced a series of keepers who had the games of their lives, but that thought is short lived when we actually see the location of where the shots were targeted.

Van der Vaart Shots

Van der Vaart

It’s no surprise to me to see that all five of VDV’s goals were scored with shots
that were aimed towards the sides of the goal and that the whole cluster of shots that
would have hit the goalkeeper were effectively wasted, as not one single goal
came from them.

Cvitanich Shots

Cvitanich Shots

Now isn’t that a pretty picture?

Despite having a similar volume of non blocked shots as Van der Vaart look at how few of Cvitanich’s 47 shots were hit straight up the middle.
Again, I have no doubt that the types of shots taken by the Nice striker meant that it was easier for him to hit the sides of the goals than Van der Vaart.
But this article isn’t written as a downer on the Dutch maestro, instead it is intended to highlight why simply using a shots on target measure to determine accuracy is flawed, especially give the level of data and statistics that are available in today’s football world.

The above is not an isolated case either.

Pog Comparison

As before we can see that the number of unblocked shots and shots on target are similar for Reading’s Russian forward Pavel Pogrebnyak and another relatively unknown Ligue 1 player, Toulouse’s Israeli international Eden Ben Basat.

However, the two players post vastly different goal scoring records from last season, and again the shot placement images for them clearly reveals the reason for this.

Pogrebnyak Shots

Pog Shots

Ben Basat Shots

Ben Basat Shots

Pretty much all of Pogrebnyak’s shots were struck towards the central portion of the goal, whereas Ben Basat’s shot placements seem almost remarkable.  The central portion of the goal, the area that leads to easy saves for the goalkeepers has a noticeable absence of footballs – there were simply no shots aimed there.

eden-ben-basat
Either Ben Basat was lucky in where the shots were targeted or else the Israeli has a serious football brain on him.  A clear pattern can even be seen in the shots that he struck off target, nearly all of them are arrowed towards the bottom corners.

Of course, by attempting to seek out the corners of the net he is going to have his share of off target shots, but with those shots which do hit the target attracting a success rate of five times that of the centrally struck shots so favoured by Pogrebnyak I know which player is likely to turn more of his shots into goals.

Parting Thoughts

In this piece I have set out why I feel that the traditional shots on target metric has its shortfalls in assessing player accuracy, I’m just not sure it’s measuring the correct thing.
Perhaps in a future piece I will have a look at creating a metric that I feel better reflects the expected goals value of shots that a player took, which after all  is what the shooting accuracy metric is supposed to do in my opinion.

  • Rob

    Could you use a scoring method using your previous article on shot placement, to measure the efficiency of shots on target to therefore rate strikers on their shooting efficiency?

    • Colin Trainor

      Rob that could be a possibility, and it’s something that I intend to get back to in the near future.

  • http://blogs.columbian.com/portland-timbers/ Chris Gluck

    Interesting article and I’d like to speak with you more for a number of reasons – to get things started here’s some initial thoughts I have I what is not dicussed here…
    1. Quality of the ball delivered by the teammate and consistency of that quality.
    2. The level of expertise of the goal keeper in knowing and understanding the danger area at the time of pressure – better keepers in better leagues read the game better therefore getting it past the keeper is more relative to accuracy than getting it to the keeper.
    3. Time on the ball to execute the shot – in better leagues better defenders will give the striker less time to put a shot on target and therefore a shot that scores a goal.
    4. Comparing a striker from one league to another league is mixing apples and oranges becuase the competition of the opponents is playing at an overall different skill level.
    5. Without getting into hyper-detail the simplified approach remains reasonable and is an equal measuring stick.
    6. At almost every level of football every striker is always told to shoot for the far post to maximize rebound and opportunity for deflection if the ball doesn’t go in – getting the time to put the foot to the ball has much more to do with the final location of the shot as anything.
    7. I would offer that Van der vaart may be working with subpar service or is himself in a league now where he can’t perform at his best because his opponents are better.
    8. For me and my research this year scoring a goal has more correlation relative to a composite score of team activities than just merely putting a shot on goal.
    9. Look forward to discussing this as the opportunity presents itself – you can read up on my “Possession with Purpose” Attacking Efficiency by visiting my website and you can see my diagram by joining the Possession with Purpose – Portland Timbers Face-book group…

    • Colin Trainor

      Chris,
      First of all thanks for taking the time to write such a detailed comment. It’s great to see that someone is so interested that they would spend so much time giving their thoughts and ideas. Cheers for that.

      You can contact me on gutties@hotmail.com if you wish.

      Most of the points that you raise relate to reasons why players might have better strike rate / efficiency rates than others.
      I totally agree with what you are saying, and (as stated within the article), the purpose of including the Van der Vaart and Cvitanich examples was not to draw conclusions about which player was a better striker or which had the more difficult chances. Instead they were included to simply illustrate how ineffective shots on target is as a measure of accuracy.

      I am all too aware that my analysis is not totally comprehensive. I did not, and cannot, distinguish which shots which were hit centrally where placed there after the keeper had been rounded or had beaten the keeper from a narrow angle at his near post. Those two examples would denote good placing of shots, but my data set (despite me being immensely proud of it) does simply not allow for that level of analysis so I have no option but to make broad assumptions.

      However broad those assumptions are though, I think it is pretty clear that shots on target as a measure of shooting accuracy is flawed.

      And as for Facebook, I’m afraid that’s one vice I don’t participate in so we’ll have to come up with another way.

  • Toshack

    Very interesting Colin!

    Regardless of Chris’ in many ways well thought out “obstacles”, I would still like to see some new metrics be created, like SoTL = Shots on Target Left etc, or whatever is relevant – if it can be done?
    I don’t believe we will ever get a perfect metric, but I’m convinced there can be a better one than “just” Shot on Target.

    And I’d like to see Ted’s comment on this, as he uses Shots on Target a lot in his analysis.

    • Colin Trainor

      Toshack, those are exactly my thoughts too.
      I know that I’ll not come up with a perfect metric, but we shouldn’t just accept that Shots on Target is the best metric for determining accuracy.

    • http://blogs.columbian.com/portland-timbers/ Chris Gluck

      Interesting Name to go by 🙂 Toshack has pedigree in the UK obviously but not sure you are aware that Mike Toshack (related to John) is the goal keeper coach for Portland Timbers…

      Perhaps I did a bad job of not reinforcing that I fully support how ineffective ‘shots on target’ is and the search for better statistics to get a better sense of who’s more effective than someone else… it’s that twist that got me working on the ‘team’ statistics concept and the potential concept to be able to measure ‘chemistry’ and the effect of one single player on the overall production for a team…

      For example, I did an analysis on how effective the Timbers were with and without Mikael Silverstre – but not by using his individual statistics – my approach looked at the ‘big picture’ statistics for the team in combination with the end result – a win, draw or loss… I am a firm believer that more precise development of individual statistics will further support the intent to be more precise on team statistics… for me a striker is just one of ten on the pitch relative to attack – and at times one of ten on the pitch relative to defense…

      I sense we all would agree that if the team (can’t) penetrate the attacking third and create reasonable goal scoring opportunities then it doesn’t matter how good the striker is; no service = no goals.
      That doesn’t mean the striker is not a good striker – therefore it’s my belief that using just individual statistics (like shots on target) is just part of the picture… especially if you wager coin on the game. 🙂 For me I don’t measure it at all relative to the team statistics for each player – but I will measure it collectively for all the players who do take shots that are either on target or not…
      And part of the success or failure of a striker is directly related to part of the success or failure of the defenders… so measuring the effectiveness of a striker must include the ineptness or greater of a defender… again – for me a collective representative of the opponents defense as a whole – for example if a fullback is part of an attack – and therefore has overlapped and penetrated deep into the opponents half – and then a turnover occurs – there is a defensive weakness – and with that weakness your striker you are measuring get’s ‘more time’ on the ball to get off a better shot – he shoots he scores… on the other hand if the opponent doesn’t play overlapping attacks and channels play towards the middle than that winger who attempts to strike the ball has less time – therefore less chance… he shoots – he misses — he doesn’t miss because he’s a bad striker he misses because it’s a better defense — my thoughts anyway and one of the primary reasons why I have been working on the possession with purpose approach …

      So yes, Toshack, you are right; those initial thoughts are obstacles – but not obstacles intended to get in the way – more like obstacles, in my view, that need to be passed in order to maximize value.

      Bottom line here is I’m not a naysayer on individual statistics but I do feel that the thirst for something like moneyball statistics in soccer is fundamentally flawed – the game itself if played in parallel and series at the same time with 25 variables on the pitch – 2 AR’s one R, 11 players for each team – whereas baseball is strictly a game played in series where one event can be seen to directly impact a next event – soccer has multiple events occuring at the same time where the prediction on where the ball my find it’s way can be influenced by anyone – for me it’s the cohesive ability of a team that works effectively together that maximizes opportunity for individual success.

      I’ve ranted quite a bit for an early morning here in Portland and appreciate your patience – I may write as if I have a closed view on this – not true – I’m just a poor writer – I would just offer I’m very passionate about it and simply love this bloody game :)..
      This comment may appear disjointed – it’s early for me – not yet had my coffee… appreciate your patience Toshack 🙂

      • Toshack

        Thanks Chris,
        I actually didn’t think of the pedigree, nor his brother :-), when I chose the name. I’m a Liverpool supporter since 1971 and old John just happened to be in his pomp then. Actually I liked Steve Heighway more, but when I started to make some comments on the TTT site (best Liverpool fan site there is IMO) for some reason I chose Toshack. To declare my age? Who knows? To make things worse (?) on Spotify my alias is toshackandkeegan :-).

        Back to business. We agree on the need for better metrics. I would love to see team metrics in what ever form they may be. By the same token it’s natural there will be metrics on individuals, same way as in baseball or basketball or NHL for that matter. Perhaps soccer is more complex since a lot of players is interfering with “everything”, but still each one is important individually as well as part of a group.
        So we do not disagree here, I believe.

        There’s just one passage in your piece that I’m not really sure I agree with? It could be that I interpret you wrong, but it seems that you say that the striker’s skill matter less than the circumstances for his scoring chance, i.e. that the goal is more a result of the other teams mistake/defensive failure than a result of the strikers skill?
        If I’m understanding you correctly, then I’l have to say that there are enough men, and strikers, that have carried their team almost single-handedly – regardless of the opposition doing this or that. Suarez in one way last year. A young Torres in Atletico Madrid. and there are more.
        But in the end I guess we just want better metrics all of us. Which is a good thing, right? So maybe better to focus on the things we agree on, than the other way around?

        Ps. At the TTT forum a guy asked if he could call me John (with a smiley)? I said: You can call me Al :-).
        But since we are Internet friends now, you can call me Peter (since that is my name).

        • http://blogs.columbian.com/portland-timbers/ Chris Gluck

          Peter,
          I actually think we agree on all accounts and it’s down more to me not communicating it very well – I am not saying a strikers skill matters less than the circumstances associated with his scoring chance – what I am saying however is that when looking at “points in the league table” (not goals scored for a striker) the circumstances associated with the scoring chance have more to do with teammates and opponents than one individual striker scoring a goal.
          Whether or not there is value in trying to capture that statistically I’m not sure but I would sense over a period of time the basic metric of shots on target would outweigh all the other teammate and opponent factors… but it doesn’t address the fact that teams hire strikers to win games and get points in the league table.
          So, for me, I nonviolently believe that the value of a striker scoring goals only has relevance to getting points in the league table – therefore I would offer that goals scored by strikers need to go into three categories before all else…
          1. Goal scored – team wins
          2. Goal scored – team drew
          3. Goal scored – team lost
          When you do that then I sense an appropriate filter has been put in place to rank the ‘team value’ of that goal based upon the team… players who score lots of goals for teams that lose should be less effective ‘strikers’ than players who score lots of goals for teams who win…….

          • Toshack

            Hi Chris (again),
            Early morning in Portland and coffee taken care of?
            As you say, basically we agree. Just looking for better metrics with different approaches. And btw it’s really Colin’s work to start with – i’m just a interested soccer fan.

            Anyway, the metric Stevie proposes could be one improvement, but to get back to the three steps you reflect upon I just today read a piece on the TTT site where a guy called Neil Mundy had broken down Liverpool’s last season in “decisive” goals and sort of rewritten the goal scoring statistics. Neil defined a “decisive” goal as:
            “Note: I have termed a ‘decisive’ goal as any strike which has made a contribution towards the team either winning or drawing a match. For example: the first three goals in the 5-2 win over Norwich last season are classed as decisive, because the opposition scored two goals, therefore we needed to score at least three in order to win the match…”

            Now since TTT is a subscription site I’m not at liberty to copy in the entire piece, buti can tell you that Suarez, Gerrard and Sturridge stood for >50% of Liverpool’s decisive goals, with Suarez alone for more than one third (and more than half of his total goals were decisive).

          • http://blogs.columbian.com/portland-timbers/ Chris Gluck

            Aye. Peter no coffee today — green tea and proper lemonade this morning – 🙂 bit scary but sometimes a cool drink in the morning on a hot day goes down well.
            I’m kinda in a brainstorming mode on this so throwing stuff out that may or may not stick on the walls 🙂 I love the intent to quantify top strikers and think the location of the strike has value – but I, personally, have trouble dissecting the location of a strike without dissecting the influence of other teammates in creating the opportunity for that strike. Perhaps that is a hurdle I’ll not get past 🙂

            In the bigger picture I’m really pleased to land on this site and the conversations so far – especially since I’m doing research myself – I will defer to Colin on offering up my email address here to chat abut other stuff 🙂
            All the best, Chris

          • Toshack

            Sounds good Chris!
            I hope to see your brainstorming activities translate into a metric or a theory eventually! 🙂
            And don’t forget to read Ben Pugsley’s new piece that came on the web some 15 minutes ago. I even introduced your thinking a bit in my comment (I hope you don’t mind and I hope a did not misrepresent you)
            Take care,
            Peter

          • http://blogs.columbian.com/portland-timbers/ Chris Gluck

            HI Peter,

            I have already written and published the theory on my blog site 🙂 – am haflway through the season and the simplified data supports the theory – I won’t offer up a link to the original article as it’s not appropriate but you can do a google search on my name and add portland timbers’ after it and you should reach my blog site.

            I saw the article just a wee bit ago and did offer some initial thoughts 🙂 Appreciate the shout-out!
            All the best,
            Chris

          • Toshack

            Thanks Chris,
            Will google you! .-)
            In Dubai at the moment, so bedtime for me.
            Talk to you later and take care,
            Peter

  • sid

    So you ended up here too!!! Fantastic!

    A slight suggestion if i may. The links in the article could be made more prominent. As of now, i know that if i mouse over the text, the colour changes and i know that a link exists. That may not be the case for others.

    As for the post, it is great!

    As fans, who squabble over internet forums, throwing stats here and there to prove our points, it is a bit difficult to get access/develop our own database to get such statistics.

    I hope this site takes off in a big way, and you might be inclined to share the figures for players for all of us to refer.

    Thanks again!

  • Stevie

    What about a metric called QSoT, Quality shots on target. You can just use those shot charts and then overlay the goal grid for successful shots in various zones. Then use the percentages from each zone multiplied by the number of shots in that area of the goal. Shots that miss are given a 0. Wouldn’t that give you QSoT? It wouldn’t address the quality of service or the defense, but if you have enough shots over the course of a season, those variables would diminish.

    Stevie

  • Leron

    For simplistic sake, wouldn’t a better measure be: the amount of goals vs unblocked shots?
    Over a wide range of unblocked shots, the placement in the corners will result in more goals..

    • Colin Trainor

      Only issue with that Leron is that this metric would be influenced by goalkeepers (which doesn’t sit easy with me).
      However, it’s certainly another option.

  • http://blogs.columbian.com/portland-timbers/ Chris Gluck

    Colin,
    I did a dig on some research last year before embarking on my other effort and this is what I came up with respect to productivity and strikers…

    I originally took shots taken divided by minutes played to offer a simplified way to account for team productivity in creating opportunities for the striker – for Robbie Keane that figure was .037302; I then took goals scored (16) and divided that into shots taken (rough prooductivity), for Robbie Keane that was .1702…
    in the end I didn’t create an Index as it wasn’t my intent – my intent was to discount (statistically) the value and relevance of the Castrol Index… – a simple approach would be to add those two ratios together to create an Index on striker productivity… for Keane that was .2075… I did the same thing for Wondowloski (.2581), Henry (.2118), Saborio (.2314), Cooper (.2465), Boyd (.1442), Sapong (.1756), Bruin (.1852), Espindola (.1601), Cummings (.1170) and Rolfe (.1710).

    The real reason I started this was that the Castrol Index indicated Kris Boyd was the 15th more productive player in all of MLS last year — complete bollocks it was and by my stat rating he didn’t come out anywhere near that but anyhow; when I matched up the overall figures the top player and goal scorer who aslo was on the team that won the Supporters Shield was Chris Wondolowski… and in comparing the final rankings I looked at it and it made sense.

    I did not use the 90 minutes per game approach since some players are more effective in their first 75 minutes (read Kuper and his story about Bergkamp in Soccernomics) – so the better representative figure to me was total minutes played, total shots and goals scored – shots on goal for me was bringing the keeper into the rating and keepers have their own save percentage – this was strictly shots taken versus goals scored… it’s been awhile since I did that research and completely forgot about it until thinking some more about your article.
    Not sure if that helps but I have an article on that analysis as well… let me know Colin…
    All the best,
    Chris

  • Pingback: Another football blog: why the shot-on-target is too harsh | Polidigwerkschdaeddle()