xG .... explained


As requested by some followers to Celtic Trends from Twitter and Facebook etc I have given some explanations below to help understand the mechanics behind some headline data measures.


Apologies to those of you who are fully conversant with the data terminology.

Included are;

  1. xG (Expected Goals)

  2. PS xG (Post Shot Expected Goals) or xGOT (Expected Goals On Target)

  3. xA (Expected Assists)

  4. NP xG (Non Penalty Expected Goals)

  5. Per 90 (Per 90 minutes)

  6. Some examples of the above are listed.

  7. Celtic in 2020/2021 season


xG

xG is a measurement based on the quality and quantity of chances a team has during a match.


It is a good guide to see how dominant a team were and if the result really reflected what happened during the game. It also reflects how good the chances they created were. IE A team can say they were unlucky not to win and they created a lot of chances etc however those chances may have been low quality ones and therefore they may not have been unlucky not to win.


Each shot is given an xG measure irrespective of whether or not the shot was on target or not as that action (on target) is post shot. xG is a pre-shot measure and is based on the likelihood of scoring based on the shot opportunity. A shot from outside the box is less likely to be a goal compared to a shot inside the 6 yard box. However a header from 16 yards is far less likely to score than a shot from 16 yards.


Details that are taken into account when assigning an xG value to an opportunity are;

  • - The distance from goal

  • - The angle

  • - Was it a shot or header and if a shot was it with the stronger or weaker foot.

  • - Was it from a cross, long ball, short pass and was it in front of them or high etc

  • - Were they one on one or were they being closed down by defenders etc

A shot from outside the box could have as low a value as 0.01, 0.02 or 0.03 etc while a shot in the 6 yard box could be 0.4, 0.5 or upwards. For info a shot with an xG of 0.02 has a 2% chance of getting a goal. 1 shot in 50 would result in a goal whereas a shot with an xG of 0.4 has a 40% chance and 2 in 5 would go in.


A penalty is given an xG of 0.79 as a penalty has a 79% chance of a goal. (Some models differ and award 0.76 for example.


These measures are calculated from thousands of previous shots from the data company’s database that were similar in situation and calculating the % of them that resulted in a goal.

Therefore an average xG is calculated and hence the term Expected Goals.

The companies that produce xG data are Wyscout, Instat, Statsbomb and OPTA amongst others.


Those players who are more clinical may outperform their xG however it is worth bearing in mind that their excellent movement and getting themselves into good scoring positions could create a very high xG for them to be compared against. Some players may outperform their xG due to just luck as we know luck does play a big part in football.


After each shot’s xG is calculated you then get a player’s xG for the match and a team’s overall xG for the match for comparison. Looking at one game in isolation doesn’t tell you too much but looking at them over more games (ie a 6 game rolling average or for a full season etc) you can start to get a better measurement on how a player or team are creating chances and then converting those chances. (Please see below Celtic's xG chart from last season).


xGA is Expected Goals Against and therefore the xG of the opponents and xGD is Expected Goals Difference which is the difference between xG and xGA.


Post Shot xG

Another metric that is worth reviewing to supplement the above information is Post Shot xG or xGOT. This measures only shots on target and measures the likelihood of a goal being scored based on the quality of the strike.


This is a good measure for player shooting accuracy and is also a good measure for measuring goalkeeper performance. How does the goalkeeper compare with goals conceded against expected goals conceded based on the quality of the shot?


A shot in the middle of the goal would be given a lower Post Shot xG while one in the top corner would be given a higher Post Shot xG.


xA

xA is calculated on the same formula as xG however in this instance the xG value is given to the player assisting in the form of an xA value. IE If a player passes to another player outside the box and they shoot from an xG 0.03 position then the player passing to the shooter will be given an xA of 0.03.


The total xA for a team in a match will always be either equal to or lower than the xG value. It can be lower due to an xG chance that has been created without an assist. IE a rebound or a player dribbling through by themselves etc.


This is my favourite measure as it shows the players who are best at producing assists and doesn’t rely on whether the shooter scores or not (which is outwith the control of the player assisting).


Therefore a player’s xA is their Expected Assists based on the chances they have created for team-mates and the average probability of their team-mates then scoring.


NP xG

NP xG is non penalty xG and is just xG minus 0.79 for any penalties. As penalties can give a player xG when they weren’t necessarily involved and can artificially inflate a team’s chances created some find it a more even measure to use NP xG instead of xG.


Per 90

Per 90 is a phrase you will see some use to balance out the minutes played when comparing players. Calculating the minutes they have played to an average of 90.


IE Player A has played 3 games (all 90mins) and scored 3 goals, Player B has played 2 games (all 90mins) and scored 3 goals and Player C has played 45mins and scored 1 goal. Player A looks the best as he has scored 3 goals but on goals per 90 Player B has scored 1.5 versus Player A’s 1. Player C has scored 2 goals per 90 as their 1 goal was scored in 45mins.


Examples (for all the above);

- Kyogo Furuhashi’s goals on Saturday were all high xG as he got himself into good shooting positions. His first goal had an xG of 0.61, his second 0.81 and his third 0.54 with his misses having an xG of 0.41 and 0.39.


- Ryan Christie therefore getting an xA of 0.81 and 0.54 for his 2 assists for Kyogo Furuhashi


- The Post Shot xG for Kyogo Furuhashi’s 3 goals were 0.93, 0.92 and 0.94.


- Rogic’s xG was 0.32, Ralston was 0.07 and Edouard’s was 0.79 as it was a penalty. The Post Shot xG for these goals were 0.45, 0.17 and 0.98.


- Against Midtjylland away McGregor’s xG was 0.03 and his Post Shot xG was 0.28.


- For further comparison Turnbull had 3 shots against Dundee and all 3 shots had an xG of 0.03 (similar to McGregor’s against Midtjylland). Turnbull’s Post Shot xG was 0.23 and 0.31 and 0 for the 3rd one as it was off target.


- With Christie getting xA of 0.03 for each of Turnbull’s chances as he was the player who passed to Turnbull before he shot.


- Our xG for the game against Dundee was an incredible 6.6 and our NP xG was 5.81


- Kyogo Furuhashi is leading the way with xG this season on 3.94 and he has outperformed that scoring 4 goals. His xG per 90 minutes is 2.45.


Celtic in season 2020/2021

The below chart is a 6 game rolling average of Celtic's NP xG for last season. The black line is NP xG and the red line is NP xGA with the dotted lines the trendlines.


As you can see last year there were peaks and troughs last year for us with our highest 6 game rolling average being an NP xG of 2.38 on the 2nd last game of the season (ie the averge for the 6 games of Dundee Utd (A), Rangers (H), Livingston (H), Aberdeen (A), Rangers (A) and St Johnstone (H)).


2nd best was the 6 game run ending on Hamilton on Boxing Day with NP xG of 2.31.


Our lowest was the 3-3 Aberdeen game in October where we were at 1.32. (NP xG of 0.48 v Rangers and 1.29 v Aberdeen were big contributors to that low average).


Our worst run for NP xGA was 1.29 for the run ending with Hibernian 2-2 draw in November. At that time our NP xGD average was only 0.32. Our worst run for NP xGD was the one that ended with Kilmarnock away (4-0) in February as NP xGD was as low as 0.29.


Our best NP xGD was 1.575 and we achieved that twice. The very first 6 games of the season ending against St Mirren and the December run ending in the game at Ibrox.


For comparison Rangers NP xG chart is below.

They actually ended the season on a negative NP xG of -0.23 with NP xGA at it's highest and NP xG at it's lowest however before we get too excited that was a natural decline when the league title was already one. Their decline from the start of the season to mid-season is worth noting though but they did start from a very high base with an NP xG of 2.63 and an NP xGD of 2.39.


For those of you who requested the above I hope you found it helpful.



Celtic Trends 🍀📈⚽️


If you would like to read more content similar to the above then please follow Celtic Trends on Twitter @Tony_McLaughlin and enable notifications for all Celtic Trends tweets.


If you like the content then please like, retweet and share with friends please.



www.CelticTrends.com

https://twitter.com/Tony_McLaughlin

https://www.facebook.com/CelticTrends

Tony@CelticTrends.com

785 views0 comments

Recent Posts

See All