Ha - didn't know anyone else saw these. My mate who I used to play poker with does this stuff for a living and he posted these on Twitter over the weekend. I commented that it just shows we're massively under-achieving.
I really dont get the expected goals thing.
The other week it said Bright had more expected goals than Chaplin in a game where Chaplin missed a couple of sitters and Bright didn't get any chances.
Is it half term in Sheffield ?If you into this (and it DOES tell a story of our season so far) this set of graphs really quite an interesting read from https://experimental361.com/2019/02/04/scatter-graphics-league-1-4-feb-2019/
Shot dominance
First of all, here is how the number of shots taken by each club compares with those they face in return. The average number of shots taken per match is on the horizontal and the average number faced is on the vertical, so bottom right (take plenty, allow few in return) is good while top left (take few, allow plenty) is bad. The stripes are like contours: the greener the stripe, the better the performance (and vice versa for red).
Attacking effectiveness
Now let’s look at attacking alone. The horizontal axis stays the same as in the graphic above, but now the vertical shows the average number of shots needed to score each league goal. Therefore bottom right is good (taking lots of shots and needing fewer efforts to convert) and top left is bad:
Defensive effectiveness
Next let’s look at the defensive situation – basically take the above chart and replace the word “taken” for “faced” on both axes. Now top left is good – facing fewer shots and able to soak up more per goal conceded – and bottom right is bad:
Expected goals
Finally here’s an attempt at correcting the first graphic for the quality of chances created and allowed, using the same “expected goals” values that power my shot timelines (explained here). The reason for doing this is that the results tend to correlate more strongly with performance than when we treat all shots equally:
Opinions?
Sent from my iPhone using Tapatalk Pro
Decent footballer too, had dodgy knees and so started playing in goal (made it to decent Non league level at Bury St Edmunds). James Scowcroft's cousin too if you remember him from his stint with us (ex Ipswich)Your mate a talented bloke Rob. His website genius and where football is growing massively.
Sent from my iPhone using Tapatalk Pro
Is it half term in Sheffield ?
A simple way to think about the Expected Goals value is as the number of goals a team “deserved” to score, if we lived in a world where all players were equally skilful and we had perfect information about the shot.
These are obviously both massive over-simplifications. While skill doesn’t vary massively between players in the same division (particularly in lower league where the better players are likely to be promoted or poached) there isn’t any data on defensive positioning or how the ball was delivered at this level. However I have to work with what I can get and to me this still feels intuitively superior to merely counting shots.
The differences between the actual and expected goals scored and conceded by a club in a given match can be explained by three non-exclusive reasons:
They were lucky (or unlucky)
They were playing against an unusually weak (or strong) opponent
They genuinely performed more (or less) effectively than the average club.
Sent from my iPhone using Tapatalk Pro
Yeah but in reality what is it based on?
For example in a match why would Bright have a higher expected goal than Chaplin after Chaplin had a couple of chances one on one with the keeper? What about if our most skillfull player only got 10 minutes and had no chances but a less skillfull player missed 10 sitters over 90 minutes. Would that mean the more skillfull player has a higher expected goals?
I can understand guessing before the game but afterwards when it has actually happened surely it's easier to be more accurate?
Yeah but in reality what is it based on?
For example in a match why would Bright have a higher expected goal than Chaplin after Chaplin had a couple of chances one on one with the keeper? What about if our most skillfull player only got 10 minutes and had no chances but a less skillfull player missed 10 sitters over 90 minutes. Would that mean the more skillfull player has a higher expected goals?
I can understand guessing before the game but afterwards when it has actually happened surely it's easier to be more accurate?
FOOKING HELL its like a really shit Spot the Ball Competition entry!If you into this (and it DOES tell a story of our season so far) this set of graphs really quite an interesting read from https://experimental361.com/2019/02/04/scatter-graphics-league-1-4-feb-2019/
Shot dominance
First of all, here is how the number of shots taken by each club compares with those they face in return. The average number of shots taken per match is on the horizontal and the average number faced is on the vertical, so bottom right (take plenty, allow few in return) is good while top left (take few, allow plenty) is bad. The stripes are like contours: the greener the stripe, the better the performance (and vice versa for red).
Attacking effectiveness
Now let’s look at attacking alone. The horizontal axis stays the same as in the graphic above, but now the vertical shows the average number of shots needed to score each league goal. Therefore bottom right is good (taking lots of shots and needing fewer efforts to convert) and top left is bad:
Defensive effectiveness
Next let’s look at the defensive situation – basically take the above chart and replace the word “taken” for “faced” on both axes. Now top left is good – facing fewer shots and able to soak up more per goal conceded – and bottom right is bad:
Expected goals
Finally here’s an attempt at correcting the first graphic for the quality of chances created and allowed, using the same “expected goals” values that power my shot timelines (explained here). The reason for doing this is that the results tend to correlate more strongly with performance than when we treat all shots equally:
Opinions?
Sent from my iPhone using Tapatalk Pro
In general before a player takes a shot, the situation is analysed and compared to hundreds of other situations in the past of football played at similar level. So if a tap in with no one nearby is scored 99.3% of the time so the xG for that chance is 0.993. A difficult 1-on-1 with a defender putting pressure may be converted 70% of the time, so the xG for that situation is 0.7. A long range attempt from a particular spot maybe works 5% of the time, so the xG is 0.05
So in theory Chaplin has scored from harder chances so therefore his expected goals will be higher.
Clear as mud?
Sent from my iPhone using Tapatalk Pro
So it works?I agree with Nick.
I can understand how before a game xG could be useful in a bookies type situation but after a game it's irrelvant - the expected number of goals in that game after the match is how many you ended up scoring.
For me, xG is only useful if you plot it against actual goals - because then it shows how many you SHOULD be scoring and how many you ARE scoring. Looking at the xG graph that just emphasises the point of how poor we are at converting chances because our actual goal tally is much, much lower than expected.
Overall, it looks like we've got an average defence and a wasteful attack. Which is how I think most people on here would describe us.
So it works?
So it works?
Some of those graphs do, because they use actual not expected data.
The last one using expected goals for and expected goals against has us performing much better than we actually are. It only has any worth if you then compare it against actual goals for/against.
1.19 actual goals against per game (1.38ish xGA)
1.03 actual goals for per game (1.53ish xG)
Which shows that the defence is actually performing slightly better than expected, but the attack is way short of what it should be doing.
So does this indicate whether we are not clinical enough or if it’s that we’re not creating enough clear cut chances? (If that makes sense) Or is it a bit of both?
What I’m asking is does the data tell us where we’re going wrong.
Sent from my iPhone using Tapatalk
It tells us exactly what we already know, creating a lot, and not scoring enough. Predicts that we should be scoring a lot more than we are (given our chances created).
What’s the average error? Is everyone skewed the same way?
So does this indicate whether we are not clinical enough or if it’s that we’re not creating enough clear cut chances? (If that makes sense) Or is it a bit of both?
What I’m asking is does the data tell us where we’re going wrong.
Sent from my iPhone using Tapatalk
I’d say a bit of both. We’ve certainly had the chances to have better results than we’ve got. Tactically, however, I would say we are generally quite cautious and mainly concede from mistakes.
We use cookies and similar technologies for the following purposes:
Do you accept cookies and these technologies?
We use cookies and similar technologies for the following purposes:
Do you accept cookies and these technologies?