Alongside the EFL cup final on Saturday, the Premier League returns after a break for the FA cup last weekend. For this round of matches, I’m going to be exploring the likely game state paths that a selection of this weekend’s fixtures could take based on scoring performances. To do this, I’ve again queried our database for all English Premier League data from the start of this season to the 22nd of February 2017. Using StrataData, I’ve calculated how frequently each team score the next goal in a given game state, as well as their expected goals (xGoals) in a given game state.
Game State (GS) simply refers to the difference of goals between two teams. For instance, if a team is up by 1 goal, then they are in a +1 GS, whilst the losing team will be in a -1 GS. If a team is up by 2 goals, then they are in +2 GS, whilst the losing team are in -2 GS (and so on). At kick-off, the GS is always 0 (as no team has yet scored), but if it is a score draw (e.g. 1-1, 2-2, 3-3 etc.), then the GS is 0+ (indicating a difference of 0 goals when there have been goals scored by both teams.)
With this in mind, we’ve calculated each team’s goal ratio, which is the number of goals scored / (goals scored + goals conceded) whilst in each game state. The bigger the percentage, the better the goal scoring performance. For instance, Chelsea have a goal ratio of 83% from the 12 instances that they’ve been in a 0-0 game state. This is calculated by counting the number of times they’ve scored first (10) from a total of 12 that they’ve been involved in (10/12 = 0.83). If they had only scored the first goal 3 times out of the 12 they’ve been involved in, their goal ratio would be 3/12 (or 25%.) This can be further explained using a smaller sample – Chelsea’s last 5 home games at 0-0. Here, Chelsea have scored the first goal in all of their last 5 home games*, so their goal ratio is 5 goals scored/5 goals total, which is 100%.
To visualise this, I’ve created some tree diagrams to demonstrate the likely path I think a game will take based on a team’s goal ratio at a given GS. I’ve created branches coming off each GS “node”. The likelihood of a team scoring the next goal based on the current GS is written on the branch, and the outcome is written at the end of the branch. I’ve then extended the tree diagram to further goal possibilities as the match develops. It’s worth noting here that the goal ratios are calculated for each team over the course of the season, and that the outcomes are not interactions between the two playing teams.
Alongside this tree diagram, I’ve created xGoals heat maps coloured green (indicating good performance) to red (indicating poor performance). As this metric is calculated per GS, I’ve calculated xGoals per 90mins based on the time each team has spent in a GS.
So now that we’re up to speed, lets jump straight into a selection of this weekend’s fixtures:
Chelsea vs. Swansea City (Saturday 25th February 2017 @ 3pm)
Chelsea entertain Paul Clement’s Swansea City on Saturday afternoon and our analysis only sees this going one way. At a 0 GS (i.e. 0-0), Chelsea have scored the first goal at home 83% of the time, whereas Swansea have scored the first goal away from home 31% of the time, so Chelsea taking the lead (1-0) would not be at all unprecedented.
When Chelsea have been in a +1 GS at home (14 instances), they’ve gone onto score the next goal 53% of the time, whereas Swansea have only gone onto equalise 14% of the time; so I’m going to back Chelsea to double their advantage here. This is corroborated when we analyse each team’s xGoals; as when in a +2 game state, Chelsea are expected to score 1.36 goals per 90, whilst Swansea are expected to concede 1.72 goals per 90 when they are in a -2 GS.
From the 9 times that Chelsea have been in a +2 game state, they’ve gone on to score the next goal 60% of time, and whilst Swansea have won 3 of their last 5 games, Chelsea going 3 up before the final whistle does not seem to be out of the question. This is further coupled by both teams’ xGoals, as when Chelsea are +3, they’re expected to score 1.71 goals per 90, whilst Swansea are expected to concede 1.52 goals per 90 when they are 3 down away from home. Based on this, I’m going for a comfortable home win for Chelsea.
Final Score: 3-0
Watford vs. West Ham United (Saturday 25th February 2017, 5:30pm)
This game is tough to call and here’s why; both teams have scored the first goal 42% of the time, and both teams’ xGoals pattern appears to cancel each other out. Looking at the xGoals heat map at the bottom of the graphic, at 0-0, West Ham are expected to concede almost 2 goals per 90 (having spent ~211mins in this game state), but Watford are only expected to score around 0.5 goals. On the flip side, West Ham are expected to score 1.26 goals at 0-0, but Watford are only expected to concede 0.39 goals at home. Given this xGoals balance, I’m going to stick my neck out and go with Watford to strike first.
Now that we’ve banked on Watford to score first, the rather diminished goal ratio for both teams (36% vs. 30%) makes me wonder whether there are going to be any further goals. Turning to the xGoals calculations, Watford are expected to score 0.65 goals when a goal up at home, with West Ham appearing to concede 0.73 goals. On the flip side, when West Ham are a goal down, they are expected to score 0.81 goals. That, alongside their recent strong scoring form away from home and the knowledge that Watford concede an average of 0.75 goals when they are a goal up, suggests that West Ham will pull a goal back.
In a score draw GS, neither team appears to have a very strong propensity to score the next one (although you could argue with West Ham’s goal ratio). However, turning to their respective xGoals gives us more granularity. Watford are expected to score 1.03 goals per 90 whilst West Ham are expected to concede just over a goal (1.06 goals) per 90. The inverse stats also favour Watford considerably, so I’m going to bank on Watford getting a winner and taking all 3 points.
Final Score: 2-1
Tottenham Hotspur vs. Stoke City (Sunday 26th February 2017, 1:30pm)
Tottenham currently sit in third place and will be hoping to continue their strong form when Stoke City visit North London on Sunday afternoon. I’ve backed Tottenham to take the lead here, as they’ve scored the first goal of the game 75% of the time playing at home. They are also expected to score 1.18 goals per 90 min in a 0 GS, whilst Stoke are expected to concede 1.30 goals per 90 in 0 GS.
Both teams will feel in familiar territory in the ±1 GS, as both teams have occupied this space 11 times this season respectively, and both have scored the next goal 45% of the time. Tottenham’s goal scoring strength is inferred by their 1.18 expected goals when one up, whilst Stoke are expected to conceded 0.74 goals when one down. On the flip side, Stoke are expected to score 0.81 goals when in a -1 GS, but Tottenham are only expected to concede just over a half a goal (0.62) when they are 1 up. This, coupled with Tottenham’s home advantage makes me think they will have enough to double their lead.
At a -2 GS, I’m giving Stoke no chance of scoring a goal. From the 4 times they’ve been in a -2 GS, they’ve never scored next. Furthermore, they are only expected to score 0.15 goals in this GS, and Tottenham are only expected to concede 0.35 goals when they’re 2 goals to the good. Having said that, I’m banking on Stoke to keep their xGoals conceded below 1, so I’ll settle on this as the final score!
Final Score: 2-0
Leicester City vs. Liverpool (Monday 27th February 2017, 8:00pm)
Champions Leicester City will host Jurgen Klopp’s Liverpool on Monday evening and will be hoping to win 3 points to push them away from the relegation zone. I’m not sure whether Ranieri’s departure will have any instant impact against Liverpool so putting the psychology aspect aside, let’s explore the numbers. Liverpool have been in a 0-0 GS 13 times this season so far, and have scored the first goal 62% of the time. Comparing this with Leicester’s 42% goal ratio from their 12 instances in a 0-0 GS, I’m going to go with Liverpool to take an away lead. Liverpool’s xGoals corroborates this, as they’re expected to score 1.30 goals whilst in the 0-0 GS.
Liverpool are most familiar with being a goal up this season, as they’ve been in this game state 14 times already. From these 14 instances, they’ve managed to score the next goal 35% of the time. Leicester on the other hand have been a goal down 7 times at home this season, scoring the next goal 38% of the time. Whilst I appreciate that Liverpool have maintained a goal ratio similar to Leicester’s even after being in +1 game state significantly more frequently, I’m going to explore the xGoals table to help us decide the next scoring team. Leicester have spent approximately 149 minutes in a -1 GS this season, so are expected to score 2.29 goals per 90 when in down 1. This, coupled with Liverpool’s propensity to concede (xGoals = 1.43 when in +1 GS) suggests that Leicester City may pull a goal back.
In this “goal draw” GS, I can only see Liverpool going on to score next. Leicester have been in this GS 3 times this season and have never scored the next goal, whilst Liverpool have a goal ratio of 56%. Furthermore, in the short time that Liverpool have ‘lived’ in this GS (~154mins), their xGoals is 2.38 per 90mins. That attacking dominance from Klopp’s men makes me believe that Liverpool will snatch the win.
Final Score: 1-2
In this analysis, we’ve combined two metrics, namely goal ratio when scoring the next goal and xGoals to help predict the likely paths a selection of games will take in this Premier League round. We believe Chelsea will have a comfortable time hosting Swansea, and whilst Watford vs. West Ham appears to be a more difficult tie to predict, I believe Watford will have the slight edge. I’m hoping Tottenham’s Europa League fixture on Thursday doesn’t affect the statistics too much, as I’ve predicted their strong domestic run to continue against Stoke. Leicester on the other may have parted company with their league-winning manager, but I don’t think this will change Liverpool taking all 3 points at the King Power.
Until next time,
Sagar Jilka (@DrSagarJilka)
* Here are Chelsea’s last five home games to breakdown the example:
4th Feb 2017: Chelsea vs. Arsenal, Marcos Alonso scored the first goal of the game (13’)
22nd Jan 2017: Chelsea vs. Hull City, Diego Costa scored the first goal of the game (45’)
31st December 2016: Chelsea vs. Stoke City, Gary Cahill scored the first goal of the game (34’)
26th December 2016: Chelsea vs. Bournemouth, Pedro scored the first goal of the game (24’)
11th December 2016: Chelsea vs. West Brom, Diego scored the first (and only) goal of the game (76’)