An Analysis of late goals in the Premier League

An Analysis of late goals in the Premier League


In this article I am looking at the late goals in the EPL as historically more goals are scored 76-FT then any other time range. Currently there have been 124 goals scored 76-FT making up 20.8% of total goals and 2011 this was 22.3% and 2010 it was 23.2% and 2009 it was 23.6% and 2008 it was 25.1% so we should see a few more late goals in 2013 in the EPL to get close  to an average of 23.55% for the last four years.

We should examine if there is any correlation in terms of the strength of a team as I  wish to understand if Man City and Man United score more late goals against weaker opposition.

Goals scored for and against 76-FT

 [table id=144 /]

If we look at the table below we see that Man City, Man United and Arsenal do score more late goals than the other teams in the EPL and there is correlation between the strength of a side and scoring late goals. It is even more interesting to see that there is also a bias towards scoring late goals away with Man United, and you would be right to think that being at Home would result in more late goals and also the famous “Fergie Time” which was shown that when Man United are losing a further 79 seconds have been played on average .

As expectation of a goal is dependent on the current score it is probably a good idea to look at Man United in terms of what the score was at the time of the goal 76-FT. Were the goals expected or a random event?  Do stronger teams score because weaker sides have fatigue or negative behaviour or a mixture of both? I have watched all the games and can report on each individual game.

Example 1: Manchester United Vs Wigan

Against Wigan, Van Persie was introduced into the game on 71 minutes and Wigan tried to push forward for a consolation goal but Nani won the ball in midfield and drove forward to be fouled just outside the area with the referee playing the advantage and McCarthy lost total concentration and played the ball into the path of Hernadez who layed the Ball back to Van Persie who scored with a clinical finish from the edge of the area. Is there a correlation between the fourth goal and the Van Persie substitution and the mindset of Mccarthy when 3-0 down to give away the ball in such a fashion.

I have written about the effect of the red card which has an effect on goal expectation. In gerneral if the red card is on or before 60 minutes then the expectation of goals to the team with 11 men is accelerated.

Example 2: Liverpool Vs Manchester United

Liverpool have Shelvey sent off on 39 minutes and the score is 0-0 and so immediate expectation of United goal(S). On 75 minutes Valencia drives into the box and is pulled down  and Van Persie scores from the spot. Valencia won the ball after Suso plays a cross field pass to Agger who is closed down by van Persie and loses the ball which Valencia runs onto. What started as an attack in the United final third ended with a United goal.

Example 3: Manchester City Vs Manchester United

Man United were winning 2-0 at Half Time v Man City and before that game the record this season when 0-2 HT was 10 away wins and one draw which was when Swansea managed to come back from 2-0 down and draw 2-2 v Reading. My argument is that the second City goal accelerated expectation of a third United  goal. In the EPL this season when 0-2 HT the away side has averaged exactly one goal second half and the home side have averaged 1.06 goals second half. If you still have no belief in goal expectation and the thought that games can mirror each other, then look at Man United v Spurs which was 0-2 at HT (2-3 FT) with a FT shot on target data of 7 for Man United and 3 for Spurs and in the City game it was 9 for City and 3 for  Man United (2-3 FT)  so each time both the home teams who lost had many more shots on target and if you had plotted the shots strength of the games then you may see very similar clusters of pressure in the games. My research is to show that games can be profiled to fit a game in the past and combine that with goal expectation so knowing all this if you accept what I have said above, the first thing you would have done at 2-2 is to back Man United to win. At 2-2 expectation of a United goal given the current score and no expectation of a further City goal.

City score on 86 minutes (2-2) and then City have a chance on 90 and the board goes up to announce four minutes of extra time. Tevez who has been booked a minute early for petulance and looks frustrated brings down Rafael in a  dangerous area which is central to the goal and just outside the box and Van Persie takes the direct free kick which gets a deflection off the wall and ends up in the far corner off the post.

Example 4: Aston Villa Vs Manchester United

Villa were 1-0 up at HT and I have also researched the scenario when the Home team are winning by +1 goal at HT and concluded a historical trend where the away team will very rarely win FT. This season before the Villa game not one team had achieved that and United were the first to win away when losing 1-0 HT 2012-2013 breaking a run of  22 in a row not winning away.

This game certainly does not fit the variables that I have spent two years looking at and I accept that some games I will predict expectation of goals and they will not arrive and other games I will predict that goals will be impeded and they will arrive in a cluster.

In this game Man United when losing 1-0 at HT at Villa should have had their goal expectation impeded and they scored three second half goals and only had 5 shots on target and interestingly all 5 were in the second half. In this game this gave Man United a goal to shot on target of 3/5 so 0.6 which means they outperformed expectation as their average goal to shot on target ratio is 0.428 and at Villa they produced a well above average 0.6.

Example 5: Manchester United Vs Newcastle United

Against Newcastle, when the score was 3-3 on 71 minutes, Man United had two further chances to score on 79 and 80 . In the first Van Persie had a ball played into the box which he hit first time and was unchallenged and just failed to hit it cleanly enough to hit the target and a minute later Valencia sent over a very inviting cross where Hernadez was unchallenged in the centre of the box and headed just wide. I had expectation of a late United goal so it was no surprise when on 90 minutes the ball was played into an unmarked Hernadez who beat the offside trap and had the simple task of guiding the ball into the net.

Example 6: Manchester United Vs West Bromwich Albion

I watched the WBA game live and keyed the shot strength for the game and I also provide goal expectation for EPL games before KO and for the WBA game I gave Man United 0.38*6.06 which means I have identified that I think that Man United will have 6 shots on target and have a goal to shot on target ratio in this game of 0.38 giving goal expectation of 2.30 for Man United which bookmakers also use to price the goals market before a game starts .

In the game Man United had 7 shots on target which was very close to my estimate and this is because shot on target data is very consistent. As I have advised before Man United shot on target data at home reads 7 5 7 7 6 7 7 5 7 7  and there was no expectation that against WBA that this would deviate in any away from the norm. The pressure in the game did build near the end and the second United goal came as no surprise.

Example 7: Southampton Vs Manchester United

If we look at Southampton v Man United this is another game of great interest in terms of goal expectation as I have looked at the effect of 1-1 at HT in terms of goal expectation. In general terms the fight back goal in the first half impedes further goals but if a goal is scored in the second half then it should trigger a “fightback” which is why there have been 59% draws this season when 1-1 HT in the EPL.

At 2-2 in the game I had no expectation of a late Man United winner, but at 2-1 I had massive expectation of a United second.

Example 8: Wigan Vs Manchester United

Looking at the second Wigan game with United 7 shots on target and 4 goals so a goal to shot on target ratio of 0.57 which is very close to the away game against Villa.

Our behaviour is learned which is a major problem. I was the child at school who used to get in trouble for questioning the belief of the teacher. If you are reading this and ready to hit the OK button (What do you think of this post rating below) then consider that some games fit the profile of a predictive model and some do not and it is possible to predict the expectation of goals as I have done it for two seasons now with very promising results.

Anyone can go to a website, will hill football betting, to look at late goal data but very few understand how to utilize the data.

  1. Learn about shot on target data and goal to  shot on target  ratio  and build your own predictive model
  2. Understand that expectation of a goal is dependent on the current score and work out if further goals are impeded or accelerated
  3. Key the shot strength during the game to see the clusters of pressure
  4. Accept that not all games will fit a predictive model or an In running and do not waste time trying to.
  5. Consider that understanding who plays Left Back for WBA is of less value then predicting goal expectation.

The biggest criticism of football research is that the sample is too small but I think this is a red herring as when I predict the expectation of a goal in a game, although I do profile a game at HT to see if it fits a particular full time score via shot strength, I am really focusing on  the game I am watching as past performance is no indicator of future performance. A team could have a goal to shot on target of 0.3 one week and 0.55 the next game.

Going back to learned behaviour as some of you will be getting ready to “trigger” the OK button it is accepted as part of survival analysis in terms of keeping “staying alive” within the norm.

I have always been someone who has challenged the status quo and my mindset has helped me to develop my research on goal expectation.

I conclude that there is expectation of a late goal if the goal before has accelerated the process and it could be down to a mixture of fatigue and mindset.

I have also looked at the expectation of a goal when it is 0-0 on 80 minutes because the longer the games goes without a goal, the more likely it is to end 0-0 FT with that caveat that shot strength build up will accelerate expectation of a goal. In the EPL the average shots on target when 0-0 FT for the home side is just 2.94 and for the away side it is 2.82 so if you are watching a game and there are very few shots on target it may end 0-0.

The most number of shots on target for the home team was 9 to Norwich and away it was Liverpool 9 and 0-0 FT.  We could say there was expectation of a goal and both the games but we need to look at the shot strength before making any assumptions.

Below you will also see a season review of the games that were 0-0 on 80 minutes .  This is also very important with regard to expectation of a goal and shot strength. It is possible to identify with accuracy which 0-0 80 minute goals should have an expected late goal. Fulham went 32 games not scoring 80-FT when 0-0 after 80 minutes simply because they did not try and score but had the mindset to try and consolidate a point rather then pushing forward for a late winner. Against Villa, Fulham had 2 shots on target 80-FT and scored with one of them. Spurs remain the only team this season to score the first goal of the game 76-FT and lose to an Everton “fightback”.

EPL 2012-2013 0-0 after 80 minutes

 [table id=145 /]