On the face of it, asking if Daniel Sturridge is more important to Liverpool football club than Luis Suarez appears a pretty silly question.

How could any other Liverpool player be deemed more important than the brilliant Uruguayan who has bagged 29 goals, 11 assists and looks destined to be crowned as the best player in the Premier League this season?

Everyone knows that Sturridge has been an integral part of their excellent relationship, and his 20 Premier League goals and 7 assists is a tremendous return in its own right.  These 27 goal involvements from his 2,127 Premier League minutes represent a very healthy figure of 1.14 goals and assists per 90 minutes played.  However, this value still falls short of Suarez’s impressive output of 1.35 goals and assists per90.

Aguero

Incidentally, the only player in the Premier League to have a higher combined non-penalty goal and assist figure than the twin Liverpool attackers is Sergio Aguero who has posted 1.43 goals and assists per90.
We can envisage the drop off in performance Liverpool would suffer if they were without either Suarez or Sturridge. That is effectively the hand that Man City has been dealt this season as Aguero has only been able to start 15 league games due to his continual injury problems.  It is my personal opinion that had Aguero played as many minutes as either half of the SAS combination that the title run in wouldn’t be anywhere near as nail-biting as it currently is.  But I digress.

Why Sturridge?

So on what basis can I suggest that Sturridge is more important to Liverpool’s chances of winning the Premier League title than Suarez?
Very simply,  the Englishman has been involved (defined as scoring or assisting) in a higher number of important goals than his pal.  Did you know that of Suarez’s 29 league goals this season, only 1 has been scored when Liverpool were behind?
The goal in question was his first goal in the victory over Cardiff on 22nd March.  According to my calculations this solitary goal is the total return from Suarez playing 358 minutes this season when Liverpool have been trailing in a game.

On the other hand, Sturridge has been instrumental in lifting Liverpool by their bootstraps when required.  He has scored 5 goals this season when Liverpool have been behind, and all of those have come from his just 256 losing pitch minutes.

I understand that these examples are from a relatively low number of minutes and concentrate solely on goals, so I have broadened my analysis and looked at a combined attacking output of goals and assists across the different Game States.
The viz below shows the combined non-penalty goals and assists per 90 minutes for each of the 3 Game States (losing, tieing and leading) for games played upto and including Sunday 6th April.  I have included 34 players which cover all players that have a combined non-penalty goal and assist (NPG&A) total of at least 10 goals this season.

2013-14 EPL NPG&A per90 Per GameState

Flat Track Bully?

Whisper it quietly, but the information in the above image suggests that Suarez may be something of a Flat Track Bully.  If Liverpool were to fall behind to Man City this Sunday, recent history would suggest that Daniel Sturridge is much more likely to be the catalyst for bringing Liverpool back into the game than Suarez is.  Sturridge’s 2.10 NPG&A per90 when Liverpool is losing is remarkable, and puts Suarez’s combined goal contribution of just 0.5 NPG&A from the Game State firmly in the shade.

We know that Suarez has been involved in 40 NPG&As this season, and we can now see that he is excellent at boosting his total output when Liverpool are already leading.  His 1.75 NPG&A per90 is, by some margin, the best in the league.  I want to be clear that I am not belittling Suarez’s excellent personal tally this season.  He has shown sublime skills on countless occasions and he has been a joy to watch but the point of this article is to make people aware that he has produced proportionally more of his output when Liverpool are already in the ascendancy in games.

Perhaps this might be because he is at his best against a high line with the opposing team putting more emphasis on attack.  His speed, both of feet and thought, ensures that he is excellent at converting and creating such goal opportunities.  When the game is tighter and Liverpool are searching for an equaliser why is it that his NPG&A output is less than virtually all the other top strikers in the Premier League?  It is possible that it’s just the samall sample size at work here but his NPG&A per90 figure is not what you would expect to see from a player that is running away with the Golden Boot trophy.

Arsenal’s problems

One other noticeable pattern in the 2013/14 viz is just how toothless Arsenal are when they have been behind in a game.  Giroud, Ramsey and Ozil all have extremely low NPG&A per90 values when Arsenal have been behind.  The data would suggest that the current Arsenal team is lacking game changers; someone with the ability to mix things up or to provide a spark when the opposition invariably pack the defence and frustrate the Gunners.  At this point it’s worth noting that Theo Walcott has been a huge miss for Arsenal.  With only a combined output of 9 NPG&As this season he didn’t make the above table but he has posted a very impressive 1.99 NPG&A per90 from the limited minutes that he has played this season.  On this measure, he stands head and shoulders above his Arsenal teammates.

Over the last 2 seasons Arsenal has taken an average of 59 minutes to score a goal when they have been behind.  The only one of the other “Big 7” teams that has taken longer on average to score when behind in a game during the last 2 seasons is Spurs at 64 minutes.
These numbers compare unfavourably with Man United (38), Man City (40), Chelsea (43), Everton (45) and Liverpool (49).

Once again, I’m conscious of drawing conclusions from a very small dataset but this data does seem to support the notion that Arsenal have to do something different this Summer if they wish to challenge for the title next season. Arsenal need someone that can stand up and be counted – they appear thin on the ground at the Emirates right now.

The supposed redemption of Danny Welbeck

Danny Welbeck’s stats suggested that he has improved as a player this season, however this data shows that a significant portion of his NPG&A output has originated from times when Man United have been in the lead this season.  His output at losing and tied Game States still leaves a lot to be desired, and if he were a racehorse perhaps he would earn a Timeform squiggle for being unreliable when the final effort was required.

 

2012/13 EPL Season

As stated above, I was concerned at the relatively low minutes that some of the players were on the field due to splitting all the data (both goals and assists and minutes played by Game State), so I replicated my analysis for last season – the 2012/13 Premier League.
This time I included all players that had contributed at least 13 NPG&As.

The Sturridge / Suarez data for last season is revealing as it is a pretty close replica of the current season, and reinforces the suggestion that Suarez’s output when Liverpool needs a goal falls short of Daniel Sturridge’s.

I wonder will people begin to speak of the possibility that Suarez, for all his brilliance, has his output significantly improved by the fact that he is most effective when his team are in the lead and that if Liverpool need a goal that Daniel Sturridge appears to be more likely to be involved.

Man United

As noted by many commentators, including in a recent post by Omar Chauduri, Man United excelled last season at recovering quickly from falling behind in games.  During the 2012/13 United took an average of just 23 minutes to score a goal when trailing and the above viz captures this phenomenon with the very high NPG&A per90 values posted by Rooney and van Persie when Man United were behind (red bars).

The above analysis is only a snapshot of what can be attempted when we look at actions by players across a range of GameStates.  As is the case with virtually all the analysis that is currently being carried out in football we have to work with relatively small datasets but I would hope that with another few seasons under our belt we will be able to draw some conclusions around the repeatability (or lack of) of players’ actions across the different GameStates.
It does seem intuitive that some players will have a greater sense of responsibility than others or the mental fortitude that sees them willing to take the initiative when others would shrink into the backkground.  This type of analysis should assist us in arriving at such judgements in the near future.

  • toshack

    Good one Colin!
    Would be interesting to see a deeper analysis of any potential difference in Liverpool’s tactical set-up when ahead vs. when behind (to maybe understand if Suárez and/or Sturridge are played diffferently or not under the different circumstances). Your theory sounds viable, but is it true? One impression I’ve got is that on the few (!) occassions Liverpool are behind, Suárez seems to work even harder, thus maybe creating space for Sturridge (even if Suárez is not directly involved in the subsequent goal). Just a thought.

  • http://vishruthsanalysis.wordpress.com Vishruth

    Colin,
    Simple analysis yet has potential to reveals a lot. Would love to have you elaborate on Man U. Maybe if you had this analysis for more seasons you could really tell if the Van Persie buy really did cover Man U’s inadequacies and get them the title.

    Vishruth

  • ballsmcgee

    You do give the caveat, but I do worry about any conclusions whatsoever from a small dataset for some players. From your numbers, I’m fairly sure you can conclude that Suarez has 1 goal and 1 assist from those 358 minutes while trailing. From memory, Coutinho and Sterling each missed clear cut chances from Suarez passes in the City game while trailing where you’d expect a >50% chance of a goal. One of the two gets put away and he goes up to ~.75NPG&Ap90 or both and it’s over 1. His actual assist could’ve been flimsy so I see the counterpoint. I do watch Liverpool with an emotional tint for sure, but I don’t notice flat track bully characteristics from Suarez.

    Flat track bully is somewhat of a weird term as far as what you are measuring anyways. Just going on recency, his recent goal against Spurs when the game was 1-0 falls into the 3rd category of measurement and is lumped in with goals while the team was up 2+ goals against the likes of WBA at home and many others. Dammit, I sound like a homer. Just don’t know that I love the characterizations of the results of your findings. Piece made me think a lot so cheers. 🙂

    • Colin Trainor

      That’s probably a fair comment overall.

      There was a little poetic licence with the narrative. I wanted to introduce these NPG&A per90s metrics.

  • tactistician

    Very, very interesting.

    Game state could become an essential part of football stat analytics. It is amazing that even in the small sample there are some players observable who recorded high values while losing both in 12/13 and 13/14 (Sturridge, Rooney, van Persie).The only risk factor is sample size.

    One question: Did you calculate with minutes played in total or did you separate for each game state (e.g. time of player on field while the team was losing).

    • Colin Trainor

      Thanks for the comments.

      I did the calculations properly. By that I mean I split the minutes that each player played over the various gamestates. So for each GS I had the number of G&A as well as the time spent at each GS. Getting the data into this format was the most challenging part of this process.

  • Vignesh

    Colin, well written as always.

    How relevant do you think Game State analysis is to individual players and their output?
    Somehow, intuitively I feel that insights drawn from Game States are more applicable to team dynamics, output etc. – Shot Ratios, PDO et al.
    I’m not entirely comfortable with applying GS insights to a single player’s performances. After all, Game State itself is a result of the two teams’ performances rather than any individual’s.
    You do caveat the entire piece with ‘the small data set’ line – in terms of minutes. Considering goals are rare occurrences, a better picture might be obtained if shots taken and chances created are analyzed at different GS instead of the resulting G+A. (I understand collecting this data set will be way more difficult)

    Maybe I’ve not been very articulate – that’s because I’m not entirely sure of why I feel that GS isn’t too relevant to individual players. Thoughts?

    • Colin Trainor

      As this was my first attempt at GameState analysis for players I wanted to keep it simple and so just used the factual numbers of Goals and Assists. If I was doing this analysis properly for myself I would use ExpG and ExpA values for the players in question. This way we aren’t simply looking at goals scored.
      However, the reason I didn’t go down that path in this article is that I fear that choice of metric may have been the focus of the article and thus I may have distracted from the realy point of the article. For example, Suarez has more goals than ExpG. Had I used ExpG numbers then some readers may have said that I should be using actual goals because there may be something wrong with the way I calculated ExpG.

      Hopefully the above make sense, and confirms your line of thinking.

      Of course GSs are very (more) relevant to teams, but I wanted to see if patters would emerge when looking at individual players. But as always seems to be the case over a season, we run out of data before we can arrive at solid conclusions.

  • Jonny

    Excellent article Colin. The betting syndicates who key chance creation to give expected goal totals to do not consider the simple metrics such as game state but they are the most affective.

    I can help you out with Man United and fighting back quickly ( 2012-2013 ) which is as a result of the early away goal metric. In 4 of the games United at home conceded an early away goal which resulted in acceleration of home goal production .

  • Michael

    Any chance you could extend this data (for Suarez and Sturridge) to include goals scored as either openers or match-winning goals that break a deadlock?

  • Egil Ytrehus

    A really interesting analysis! I hope you will do a similar one using ExpG aswell! It would be interesting to see who is underperforming and overperforming in the different gamestates.

    This article lead me to thinking about goalkeepers, could you do something similar with them? Could that potentially give some interesting results?
    I mean, are there flat track bullies among keepers?
    Of the pieces I have read regarding statistical analysis of keepers, they (and yours?) all compare the overall average performance of keepers, but it would be interesting aswell to see what keepers are big game performers with game changing saves (even if they on average dont compare that high, Cech?).
    Additionally, could there be a general trend? That keepers are performing better, or worse, when their their team is trending?

  • Steve

    How often do Liverpool play from behind? Their current run of play suggest that they deliberately come out fast and hard to get a lead of 2 or 3 goals and then put it into cruise control. Remember that Suarez wasn’t available during the first 5 games of the season, when Sturridge was the main go to guy. During that time LFC weren’t playing well at all. It seems to me if you are going to use game state statistics then you have to account for the amount of time that a team plays in that particular state. Of course, much the same could be said about Man City playing from behind. But with Aguerro, when a player can’t stay on the pitch the per 90 stats just don’t hold any water.

    Steve

Improve Performance and Productivity in Your Club:
State-of-the-art Football Analytics