Can Plucky Underdogs Norwich Upset the Odds at Preston?

[nutty nigel 7,509]

13 minutes ago, king canary said:

When people talk about expected goals or 'XG' it is a stat based on the % of a chance becoming a goal. Generally this number will have been settled on by looking at a large number of similar chances in terms of position, distance, covering defenders etc etc (all quite complex). 

So the goal first Pukki scored against Ipswich would have a pretty low XG rating as most of the time you'd expect it wouldn't go in- he hit it perfectly into the one part of the goal the keeper couldn't reach with pressure from a covering defender. Alternatively the chance he scored v Leeds would have a very high XG- basically an open goal from 6 yards, you'd expect him to bury that most times.

So who decides the chances from which we should expect a goal? 

[It's Character Forming 1,160]

OK we'd have to agree to disagree on that.  When you say he had to take the chance on Sunday "first time", the ball had been laid in by Buendia so perfectly that he didn't need to take a controlling touch and he was able to run next to it for several strides while he picked his spot, which is a lot easier than I'd normally think of when you have to hit a ball first time eg when a ball is crossed.  I can't think of a chance that good last night.  Does XG take account of how long the striker has to take the chance ?

[Bethnal Yellow and Green 1,557]

52 minutes ago, nutty nigel said:

So who decides the chances from which we should expect a goal? 

The name ‘expected goals’ is probably the biggest problem with this stat. People hear it and make a judgement and then disregard it.

It is simply a measure of how many times out of 100 similar shots have been scored. There are numerous different models which take different considerations into there workings - I gave an extensive list earlier in this thread. 

I don’t get why people get so defensive about numbers. 

[king canary 7,578]

1 hour ago, nutty nigel said:

So who decides the chances from which we should expect a goal? 

As Bethnal says- it isn't about 'you should expect a goal from this chance.' It is about how high the likelihood is of each chance being scored and a cumulative total. So against Ipswich one site has our XG at 1.6 which suggested from the chances we had we should have expected between one and two goals. The fact we scored three suggests that our finishing was good and we made the most of the chances we had.

[nutty nigel 7,509]

23 minutes ago, king canary said:

As Bethnal says- it isn't about 'you should expect a goal from this chance.' It is about how high the likelihood is of each chance being scored and a cumulative total. So against Ipswich one site has our XG at 1.6 which suggested from the chances we had we should have expected between one and two goals. The fact we scored three suggests that our finishing was good and we made the most of the chances we had.

It's even more confusing to me. What was Pukki's expected goals from his chances last night?

[king canary 7,578]

20 minutes ago, nutty nigel said:

It's even more confusing to me. What was Pukki's expected goals from his chances last night?

I don't know, I haven't seen the stats- not many places post Championship XG so it is difficult to say. Apparently it was 0.8 v Ipswich though.

It certainly isn't perfect and to be honest it took ages to get my head around it but once I had it properly explained to me it made much more sense. 

[nutty nigel 7,509]

48 minutes ago, Bethnal Yellow and Green said:

The name ‘expected goals’ is probably the biggest problem with this stat. People hear it and make a judgement and then disregard it.

It is simply a measure of how many times out of 100 similar shots have been scored. There are numerous different models which take different considerations into there workings - I gave an extensive list earlier in this thread. 

I don’t get why people get so defensive about numbers. 

Why do you think I'm defensive about numbers Beth? You're barking up the wrong tree buddy. I do however believe that for the numbers to be useful they have to have a solid base. As I understand it the basis for these figures is the same for every team in every season. In my experience they may build a trend with a large sample over a long period of time. However to transfer that to such a small sample over a short period of time would not have the same results and other factors such as the form of players and individual systems make these stats pretty much redundant.

[Bethnal Yellow and Green 1,557]

10 minutes ago, nutty nigel said:

Why do you think I'm defensive about numbers Beth? You're barking up the wrong tree buddy. I do however believe that for the numbers to be useful they have to have a solid base. As I understand it the basis for these figures is the same for every team in every season. In my experience they may build a trend with a large sample over a long period of time. However to transfer that to such a small sample over a short period of time would not have the same results and other factors such as the form of players and individual systems make these stats pretty much redundant.

Sorry, that was a more in general point and not aimed directly at you nutty. 

What you say is actually what I have been saying on here also. These are good stats to see trends but are not going to tell you what will or won’t happen tomorrow. 

For me, the trends suggests that Pukki (and Norwich) are overachieving in finishing. When that stops is anyone’s guess - but I do believe in ‘reversion to the mean’. 

[BigFish 1,986]

1 hour ago, Bethnal Yellow and Green said:

Sorry, that was a more in general point and not aimed directly at you nutty. 

What you say is actually what I have been saying on here also. These are good stats to see trends but are not going to tell you what will or won’t happen tomorrow. 

For me, the trends suggests that Pukki (and Norwich) are overachieving in finishing. When that stops is anyone’s guess - but I do believe in ‘reversion to the mean’. 

Certainly this season looks like one of those where the fates conspire to create a good season that exceeds expectations and the stats are interesting as a tool to analyse what has happened. I also agree with your point earlier (I think), that football is too chaotic and low scoring to predict accurately. I note the quotes you use around reversion to the mean so I think I can agree that this current group of players are unlikely to see another season like this even if the term isn't scientific. However, I don't think anyone could stand up a stat that would subjectively back this up.

 

 

 

[westcoastcanary 173]

3 hours ago, king canary said:

This is why XG exists though- to actually try and put some form of qualifier on what is actually a 'good chance.' For me I completely disagree about the first goal v Ipswich- he has to take it first time from the edge of the area, under some pressure with basically one corner of the goal to aim at. 

 

 

I agree. I think it helps to ask what margin of error there was with that first goal against Ipswich. Several factors made it a quite wonderful goal to score -- e.g. the distance (the further the ball travels, the more a slight misdirection becomes magnified), and the narrowness of the scoring "corridor" formed by Bialkowski's fingertips and the inside of the post (Bialkowski was within centimetres of getting a touch, and the ball within centimetres of maybe bouncing off the post). I'd liken it to a golfer striking a hole in one; the best golfers in the world have a margin of error of several feet round the pin on a par three hole, which means that a perfectly struck shot still has far more chance of missing the hole than going in. I'd describe Pukki's strike for that first goal as perfect, and his strike for the second one not far off that either (and I already said on the Ipswich match thread that Emi's through balls were perfect in themselves). 

Nutty, it's not a matter of what can reasonably be expected of Pukki. The point is that he has been exceeding reasonable expectation for most of the season and that has been a crucial contributor to our points total. Thus against Ipswich, that 3:0 scoreline came about only by dint of one "perfect" and one "very near perfect" Pukki strike. On expectation alone, a 1:0 win would have been the "right" result (or going by xG, 1:1 draw).

[Jim Smith 2,314]

5 hours ago, king canary said:

I think you're misunderstanding my point. I'm not saying I expect him to score every chance or even that he should have scored more yesterday. What I'm saying is that Pukki has been exceptional at taking low % chances and turning them into goals (his first v Ipswich is a great example), last night those similar low % chances didn't go in, which is entirely to be expected.

I'd say the opposite. Most of Pukki's goals have been from very high quality chances or as they are often referred to "tap ins" because his movement and positional sense are so good. I can;t really think of too many "difficult" chances he's taken. perhaps his goal against derby is one example but a most of his goals are straight forward finishes which he takes very well as he is so composed. 

[westcoastcanary 173]

50 minutes ago, Jim Smith said:

Most of Pukki's goals have been from very high quality chances

What does "most" mean? 51%? 95%?  It can be perfectly true that "most" of TP's goals have come from "high quality chances", yet also true that "he has been exceptional at taking low % chances and turning them into goals". The stat that started this debate off (Bethnal's post from Norwich City Stats) showed that our GF total exceeds our xG by 12.5. In percentage terms, that means that 20% of the 60 goals we had scored before the Preston game have been goals scored "against the odds" as measured by xG. Of course, not all those goals will have been scored by Teemu, but it's pretty safe to say that he has been responsible for the bulk of them. 

Edited February 14, 2019 by westcoastcanary
Typos

[Jim Smith 2,314]

9 minutes ago, westcoastcanary said:

What does "most" mean? 51%? 95%?  It can be perfectly true that "most" of TP's goals have come from "high quality chances", yet also true that "he has been exceptional at taking low % chances and turning them into goals". The stat that started this debate off (Bethnal's post from Norwich City Stats) showed that our GF total exceeds our xG by 12.5. In percentage terms, that means that 20% of the 60 goals we had scored before the Preston game have been goals scored "against the odds" as measured by xG. Of course, not all those goals will have been scored by Teemu, but it's pretty safe to say that he has been responsible for the bulk of them. 

What it means is that all bar about 2 of Pukki's goals this season have in my view been from chances that I would regard as high % chances. That's not a criticism of him, it happens because he gets in such great positions so regularly, makes good runs and has great anticipation. I would not agree that his chance for the first against Ipswich was a low % chance (although his quality meant he made it look easy) but again one thing I am uncertain of is do they take into account the level of the game in terms of assessing whether a chance is high % or low %. For example, for a league 1 standard striker I would say that I would expect them to score that chance less frequently than a better striker so is the xG analysis undertaken based purely on previous championship chances? Pukki essentially had two one on ones with the keeper against ipswich and indeed had a third reasonable chance on his right foot (where he took a bit too long) even if you ignore the Stiepermann 2 on 1 opportunity we had. I am therefore struggling to see how xG concludes we had a xG of 1.8 for that match when in my view we had at least 4 very good goal scoring chances (for any championship striker). Similarly I think in previous games this season we have  created and missed lots of good chances so again I am dubious as to the accuracy of these xG stats and how they are arrived at.

[It's Character Forming 1,160]

For me talking about the first Pukki goal, yes it does require a high level of accuracy from the striker to put the ball in the corner, between the keeper and the post.

 

Mostly in top level football that is the case to score. Typically you will have a keeper guarding the goal and having a wider area to aim at than Pukki had is, I would say, relatively unusual. The angle could be better but as these things go it wasn’t bad.

 

For Pukki’s first goal he had time to prepare for the shot, he was perfectly balanced, he was under no real pressure from a defender (the one next to him was just matching pace with him) and the ball was running parallel to him at a comfortable speed. All in all I’d say that shot was equivalent to a tennis player putting a ball down the side of the court near the baseline, which pro players do all the time.

 

compare the first preston goal yesterday. The player has to jump and is in physical contact with at least one player so he is not balanced with numerous others around him, the ball is coming across at pace, he need to change it’s direction very precisely with the header while in mid air, and like Pukki he’s got to get it in the corner to prevent it being saved.  So a much more difficult chance, but IMO much more typical of a  chance in football at this level.

 

I’m not saying Pukki’s first goal was easy. But looking at the factors which would make a goal harder or easier to score at this level, I would say for me it is easier than the average.

 

On question - do sides at the top of the table outperform their XG stats normally ? Because I’d expect that if the stats predict an average conversion rate at Prem leve, Man City surely will convert more of their chances than that  average team in the Prem? Because their strikers are better than average at finishing, that’s for sure. Is this the case? 

 

 

[westcoastcanary 173]

8 hours ago, Jim Smith said:

...........all bar about 2 of Pukki's goals this season have in my view been from chances that I would regard as high % chances.

What exactly do you mean by a "high % chance"? The xG of a penalty is 0.76, i.e. on average, 3 in every 4 penalties are converted, 1 is missed, or 76% of penalties are converted, 24% are missed. Are you saying that all but 2 of TP's goals this season have been as easy, or even easier, to score as a penalty? Are you saying that his two goals against Ipswich were in that category? 

[It's Character Forming 1,160]

Is it possible to find out the XG % chance that was calculated for each of Pukki’s goals on Sunday?

[king canary 7,578]

This graph gives an idea...

Pukki's first (where the second yellow dot is) looks like it is given an XG of about 0.2 or 0.3. You can see from how our XG rating leaps after the first goal that was clearly considered a much higher % chance.

Edited February 15, 2019 by king canary

[lake district canary 4,531]

13 hours ago, It's Character Forming said:

For Pukki’s first goal he had time to prepare for the shot, he was perfectly balanced, he was under no real pressure from a defender (the one next to him was just matching pace with him) and the ball was running parallel to him at a comfortable speed. All in all I’d say that shot was equivalent to a tennis player putting a ball down the side of the court near the baseline, which pro players do all the time.

compare the first preston goal yesterday. The player has to jump and is in physical contact with at least one player so he is not balanced with numerous others around him, the ball is coming across at pace, he need to change it’s direction very precisely with the header while in mid air, and like Pukki he’s got to get it in the corner to prevent it being saved.  So a much more difficult chance, but IMO much more typical of a  chance in football at this level.

I’m not saying Pukki’s first goal was easy. But looking at the factors which would make a goal harder or easier to score at this level, I would say for me it is easier than the average.

 

Sorry, but I find that a tad ridiculous.  A ball fired in from a free kick is already heading towards a very dangerous area where the slightest flick can cause it to go in the net. 

To say that Pukki's goal was "easier than average" is to take away the level of skill, awareness and confidence of the player involved.  The level of accuracy of his shot was superb, the ball as it was running away to the left. You rightly mention his balance - but that was because he is a great striker and had the confidence and composure to let the ball run to allow himself to be balanced, where other strikers would have taken a touch. In other words he made the goal look easy, but easy it certainly wasn't.  A terrific strike from a terrific player......

Anyone can jump up at a ball coming towards them and get a connection with their head and send it goalwards.....

 

 

 

[lake district canary 4,531]

17 hours ago, Jim Smith said:

I'd say the opposite. Most of Pukki's goals have been from very high quality chances or as they are often referred to "tap ins" because his movement and positional sense are so good. I can;t really think of too many "difficult" chances he's taken. perhaps his goal against derby is one example but a most of his goals are straight forward finishes which he takes very well as he is so composed. 

Pukki's goals are varied, many due to his positional sense - but most due to the fact he has a great touch on the ball (witness his assists which are quality assists) and he has the ability to put the ball in the net with whatever he needs to (the one off his thigh comes to mind - no luck, no fluke, but totally intentional). 

The goals only look easy because he makes them look easy and to say he is playing above expectations I find hard to find credibility in.  He is a quality player as can be seen from everything he does on the ball and off the ball.

He is not playing above expectations, he is playing to expectations.

 

 

[Thirsty Lizard 3,155]

3 hours ago, king canary said:

This graph gives an idea...

Pukki's first (where the second yellow dot is) looks like it is given an XG of about 0.2 or 0.3. You can see from how our XG rating leaps after the first goal that was clearly considered a much higher % chance.

When you say that Pukki's first goal is given an XG of 0.2 or 0.3 King C is that just a measure of how far the line on the graph jumps vertically at that point? So Hernandez's goal has an XG of 0.5 because the graph line jumps vertically from 0.0 to 0.5?  And what does Chambers 0.3 represent? Is that just a chance he had at some point which he didn't score with? (Or does he get given a 0.3 for grabbing Godfrey's nuts?). 

[king canary 7,578]

1 minute ago, Thirsty Lizard said:

When you say that Pukki's first goal is given an XG of 0.2 or 0.3 King C is that just a measure of how far the line on the graph jumps vertically at that point? So Hernandez's goal has an XG of 0.5 because the graph line jumps vertically from 0.0 to 0.5?

Exactly that.

1 minute ago, Thirsty Lizard said:

And what does Chambers 0.3 represent? Is that just a chance he had at some point which he didn't score with? (Or does he get given a 0.3 for grabbing Godfrey's nuts?). 

That is his cumulative total throughout the game- so it could mean he had 3 chances that were each worth 0.1 or one chance worth 0.3 etc etc. The fact he's Ipswich's highest XG rating suggests they never created any notable chances.

[It's Character Forming 1,160]

OK so what I take from these xG stats is...

 

With average quality strikers, 1p5wich would on average have scored 1.2 goals on Sunday.  Obviously they scored nil, and this is consistent with their strikers being pretty **** this season, which is a large part of why they are stranded at the bottom of the table.  So there's no surprise that they are scoring fewer than would be expected from the chances they had, that's pretty much the story of their season and why they are likely to get relegated.

 

City meanwhile, with average quality strikers would have been expected to score 1.6 goals.  Obviously we scored 3, which is consistent with our striker (Pukki), having a great season, for various reasons, which is a large part of why we're at the top end of the table.  So it's no surprise that we're scoring more than would be expected from the chances we had, that's a big part of the story of our season and why we are likely to get promoted.

[It's Character Forming 1,160]

In fact thinking about it, I would find it surprising if the stats showed otherwise... If the bottom side is converting it's chances as well as average, that would be a bit strange, and ditto if the top side is not above average in its finishing, that would be a surprise.  The sides are top and bottom for a reason.

[Jim Smith 2,314]

10 hours ago, westcoastcanary said:

What exactly do you mean by a "high % chance"? The xG of a penalty is 0.76, i.e. on average, 3 in every 4 penalties are converted, 1 is missed, or 76% of penalties are converted, 24% are missed. Are you saying that all but 2 of TP's goals this season have been as easy, or even easier, to score as a penalty? Are you saying that his two goals against Ipswich were in that category? 

The second one wasn;t harder than a penalty certainly. With our record on penalties there is clearly a potential joke here somewhere but yes I would say that at least half the goals he's scored this season were as easy or perhaps even easier than a penalty.

[Jim Smith 2,314]

2 hours ago, lake district canary said:

Pukki's goals are varied, many due to his positional sense - but most due to the fact he has a great touch on the ball (witness his assists which are quality assists) and he has the ability to put the ball in the net with whatever he needs to (the one off his thigh comes to mind - no luck, no fluke, but totally intentional). 

The goals only look easy because he makes them look easy and to say he is playing above expectations I find hard to find credibility in.  He is a quality player as can be seen from everything he does on the ball and off the ball.

He is not playing above expectations, he is playing to expectations.

 

 

I'm not saying he's playing above expectations. I agree with you. He's scoring chances i would expect him and other similar quality strikers to score on a regular basis, albeit he on occasions makes them look very easy due to his quality. I am saying I think that the xG stats for us this season have to be taken with a pinch of salt if they are concluding we are scoring way more goals than we should be based on the chances created because I think we create very good quality chances on a regular basis.

[Jim Smith 2,314]

1 hour ago, king canary said:

Exactly that.

That is his cumulative total throughout the game- so it could mean he had 3 chances that were each worth 0.1 or one chance worth 0.3 etc etc. The fact he's Ipswich's highest XG rating suggests they never created any notable chances.

The fact Pukki's cumulative total is 0.8 when he scored from two one on ones and also had a third chance through on goal where he failed to get a shot away illustrates clearly to me that xG downplays the quality of chances.

[ZLF 271]

I think the xG for a penalty is about .74.  Which was the % of penalties scored in the prem last year.

 

Apply that uo each scenario and you get a team xG per game.

 

Do that for attacking your xGf and defence you have xGa.   

 

 

[It's Character Forming 1,160]

So is the xG for the Championship lower that for the Prem ? 

 

I.e. it is based on what an average player at that level would be expected to score.  An average Champs striker would be expected to convert fewer chances than a Prem striker.

 

 

[westcoastcanary 173]

This thread is a marvellous illustration of the value of an objective measure of quality of chance, which is what xG is designed to provide. No-one claims it is a perfect measure, and the methodology underpinning it is constantly being refined and improved, but the concept of it is sound and the idea that the values provided by leaders in the analytics field such as Opta are somehow as wildly wide of the mark as Ipswich's shooting simply doesn't hold up. 

ICF's and LDC's disagreement about how hard or easy it was for Davies to score that first Preston goal is a classic illustration, and Ben Mayhew's graphic for the match clearly shows that ICF's is a more accurate assessment: 

Contrary to LDC's apparent belief, headed goals in general have lower xGs than conventional strikes with the foot.

[westcoastcanary 173]

3 hours ago, It's Character Forming said:

So is the xG for the Championship lower that for the Prem ? I.e. it is based on what an average player at that level would be expected to score.  An average Champs striker would be expected to convert fewer chances than a Prem striker.

As Bethnal posted earlier, xG values are arrived at by analysing data from multiple leagues over multiple seasons. As far as I am aware, there is no such thing as an "xG for the Championship". Championship strikers play against Championship defenders in teams with Championship goalkeepers and chance creaters. Ditto Premiership strikers. xG discounts those sorts of differences of level by averaging across multiple leagues.