Pape Sarr

  • The Fighting Cock is a forum for fans of Tottenham Hotspur Football Club. Here you can discuss Spurs latest matches, our squad, tactics and any transfer news surrounding the club. Registration gives you access to all our forums (including 'Off Topic' discussion) and removes most of the adverts (you can remove them all via an account upgrade). You're here now, you might as well...

    Get involved!

Latest Spurs videos from Sky Sports

what I was hinting at without really explaining is that the 2 stats that chart is looking at are xA and key passes per 90. Both of those are cumulative stats with a lot of overlap (meaning 1 key pass could also increase xA) and they don’t measure efficiency at all.

So any risk taker, like Bruno, is going to show strongly even if he’s far less efficient than someone that shows lower on this graph. these 2 pieces of data reward players for taking risks but don’t penalize them at all for failing (giving the ball away). And when a player loses the ball, he’s taking away an opportunity for his teammates to make a play.

Basketball is a sport that is centered around efficiency when it comes to data. Possession is important because both teams get a relatively equal number of possessions. And football is not too different in that possession is important but it’s a little less equally distributed.

Point being, let’s take a team like man city. I guarantee their analysts are looking at creative players but making sure they are efficient. Because pep knows you only get so many possessions in a match to score, and if you have several players that give the ball away too frequently, you won’t score as much, even if they have great stats like this.

Consider the Alexis Sanchez transfer to man United several years ago. A lot of people were saying that was a big signing because of his assists, chances created and all sorts of cumulative stats. But it was pretty clear at the time that he was inefficient and that’s why he failed at United.

Now United have another high volume creator like Bruno. Sure he gets his stats but no one else can put up stats because either Bruno plays the key pass or the ball is given away.

I don't disagree with your points but no one is claiming the graph shows the "best overall playmaker/most efficient passer" etc.
It's XA and key passes per 90. Does what it says on the tin.

I follow the NBA too (as an Englishman) and I'm aware of volume shooters vs efficient shooters.
 
It's a poor visualization, IMO, which has led to some consternation. By charting xA on the Y and key passes on the X, high-right is basically an indicator of volume. The trend line is the prime indicator.

Re: bonaspur23 's post, this visualization shows Maddison creates more xA/key pass than Bruno which indicates he's more efficient - his passes are better at creating goal scoring situations and not just advancing attacking play.

You really need to plot touches, or passes, to get an understanding of how volume factors in. I.e., is it as simple as getting the ball to Maddison more increasing both his key passes and xA numbers. Or is it more heavily impacted by the team around him, i.e. Maddison is getting similar numbers of touches in similar situations but is less often able to find a threatening pass (even if his passing is more effective). This could mean he's either missing opportunities, or that his teammates are failing to present him opportunities.

Stats in football are uneasy bedfellows. They're absolutely useful, but the game is exceedingly complex and the metrics most easily translated into figures aren't always the most valuable.
There are all sorts of stats like stats to do with what a player can do when he receives the ball within 5-10 yards of an opposition defender, forward pass %, pass completion etc. Various combinations can paint different pictures and each of those have their own unique context. I don't think this means we can't enjoy a chart/graph based on basic stats - they're still fun and/or useful.

Re: the bold how so?
 
There are all sorts of stats like stats to do with what a player can do when he receives the ball within 5-10 yards of an opposition defender, forward pass %, pass completion etc. Various combinations can paint different pictures and each of those have their own unique context. I don't think this means we can't enjoy a chart/graph based on basic stats - they're still fun and/or useful.

Re: the bold how so?
The graph plots xA vs. KP, so if you draw a line between the origin and Bruno's spot on the chart, anyone with a dot above that line is creating more xA/KP and anyone below that line is creating fewer xA/KP.

It's also a very messy dataset with no discernible trendline. That is likely due to a relatively small dataset at this point, but also indicates that the two variables may not be well correlated. Football is a difficult sport to quantify numerically, everything is fluid and subjective with the main objective, scoring goals, especially rare.
 
He’s is amazing. I go to see him every home game, comes out with some great new chants
Yeah they should play some over the speakers after we score our customary last min winner and put the words on the screens so people can learn them...need to get them chanted in the matches somehow and seems we only manage two or 3 songs all game.
 
The graph plots xA vs. KP, so if you draw a line between the origin and Bruno's spot on the chart, anyone with a dot above that line is creating more xA/KP and anyone below that line is creating fewer xA/KP.

It's also a very messy dataset with no discernible trendline. That is likely due to a relatively small dataset at this point, but also indicates that the two variables may not be well correlated. Football is a difficult sport to quantify numerically, everything is fluid and subjective with the main objective, scoring goals, especially rare.
Do you mean to connect Bruno's dot to where the dotted lines meet? With anything above that line being better efficiency?

I don't understand my own stat :ange-facepalm2:
 
Do you mean to connect Bruno's dot to where the dotted lines meet? With anything above that line being better efficiency?

I don't understand my own stat :ange-facepalm2:
Looks like the dotted lines represent an average of all the players in the dataset, but I'm not sure. If that's the case, then an argument could be made that Bruno is just an average player with a lot of opportunity, as the line from the origin to his dot runs basically through the "average" intersection.

Here's the graph with the line I meant drawn on it.

View: https://imgur.com/a/F9GehHs
 
Looks like the dotted lines represent an average of all the players in the dataset, but I'm not sure. If that's the case, then an argument could be made that Bruno is just an average player with a lot of opportunity, as the line from the origin to his dot runs basically through the "average" intersection.

Here's the graph with the line I meant drawn on it.

View: https://imgur.com/a/F9GehHs

Ehh, that's not how I read that graph. Both axes are independent metrics, with vertical being 0 to Exceptional in expected Assists, and horizontal being 0 to Exceptional in Key Passes. I've never been a big fan of B. Fernandes, but as I read the graph he is Exceptional in both categories, not average. He outperforms Maddison in both categories.

If there's a distinctly "average" player in that graph, it looks to me to be Odegaard.
 
Ehh, that's not how I read that graph. Both axes are independent metrics, with vertical being 0 to Exceptional in expected Assists, and horizontal being 0 to Exceptional in Key Passes. I've never been a big fan of B. Fernandes, but as I read the graph he is Exceptional in both categories, not average. He outperforms Maddison in both categories.

If there's a distinctly "average" player in that graph, it looks to me to be Odegaard.
They aren’t independent stats though. A key pass is a pass that creates a shot (im finding various definitions here… some say creates a goal scoring opportunity but I’m not sure what determines that if not a shot ). A shot creates xA for the player that made the pass. These stats have extreme overlap.

So top right is an indicator of volume. Whoever creates the most will end up in the top right.

Being in the “more xA / Kp” section just means the average chance that you create is a higher % chance. Because key pass will always increase by 1 for each key pass, and each key pass will guarantee an increase in xA but the increase will vary depending on the quality of the chance. It’s better to be in this section vs the “fewer xA / kp” section but not at the expense of significantly less key passes (or less efficiency, which isn’t measured here).

So this graph basically measures quality of chance per chance created if you measure which side of the y=x line drawn by MemphisWill MemphisWill as well as volume of chances created if you look at the top right vs bottom left.

To me, these stats are too related to get much value. The only thing I really get out of it is the top right is the place to be IF you are just as efficient as the rest. It would mean you create more given the same opportunity. But there’s no efficiency measure in either stat, so it doesn’t tell me that.
 
Last edited:
They aren’t independent stats though. A key pass is a pass that creates a shot (im finding various definitions here… some say creates a goal scoring opportunity but I’m not sure what determines that if not a shot ). A shot creates xA for the player that made the pass. These stats have extreme overlap.

So top right is an indicator of volume. Whoever creates the most will end up in the top right.

Being in the “more xA / Kp” section just means the average chance that you create is a higher % chance. Because key pass will always increase by 1 for each key pass, and each key pass will guarantee an increase in xA but the increase will vary depending on the quality of the chance.

So this graph basically measures quality of chance per chance created if you measure which side of the y=x line drawn by MemphisWill MemphisWill as well as volume of chances created if you look at the top right vs bottom left.

To me, these stats are too related to get much value.
I see your point. Frankly I've never put much attention into tracking Key Passes and "Expected" results. Okay, they overlap heavily but volume of shots is still a great indicator. At the end of the day, the old adage still holds true that you miss 100% of the shots you don't take.
 
Okay so

Key pass = pass to teammate that leads to shot (but not a goal)
Expected Assist = likelihood a pass leads to a goal

So a key pass can have a high xA value or a low xA value

thus

key passes measure reality, only passes made in games leading to shots

expected assists measures the % chance that type of pass gets converted based on averages

So there is overlap but not really. There graph doesn't show volume imo cos you can't really make that deduction based on just these 2 values.
 
Ehh, that's not how I read that graph. Both axes are independent metrics, with vertical being 0 to Exceptional in expected Assists, and horizontal being 0 to Exceptional in Key Passes. I've never been a big fan of B. Fernandes, but as I read the graph he is Exceptional in both categories, not average. He outperforms Maddison in both categories.

If there's a distinctly "average" player in that graph, it looks to me to be Odegaard.
Yeah, obviously he's not average as that's the intersection of the dotted lines. But, summarizing the point bonaspur23 already made, Bruno's efficiency is approximately average effectiveness extrapolated out to far above average volume. Part of that extrapolation is above average technical ability and vision, some is above average opportunity thanks to the quality of his teammates.

In any event Maddison > Bruno is the point.
 
Okay so

Key pass = pass to teammate that leads to shot (but not a goal)
Expected Assist = likelihood a pass leads to a goal

So a key pass can have a high xA value or a low xA value

thus

key passes measure reality, only passes made in games leading to shots

expected assists measures the % chance that type of pass gets converted based on averages

So there is overlap but not really. There graph doesn't show volume imo cos you can't really make that deduction based on just these 2 values.
Assists don't count as key passes though so if you created a tap in for your striker with a 90% chance of your teammate scoring you would get 0.9 xA but only get a key pass if they missed
 
Assists don't count as key passes though so if you created a tap in for your striker with a 90% chance of your teammate scoring you would get 0.9 xA but only get a key pass if they missed
Ah true.

Now I think about it MemphisWill MemphisWill is right

A higher xA to key pass ratio whilst being generally upper right suggests - an overall elite tier but a better quality of pass from my understanding. Making Madders more efficient arguably?
 
So Bruno is the best player in the world?

Until you look at his possession stats and see how many times he gives the ball away.
And this is why it’s bollocks. Bruno is a great footballer but he’s a beaky nosed shirker when it’s backs against the wall. I’d have any of our midfield players over that Ronaldo wannabe.
 
Last edited:
MemphisWill MemphisWill - great value in this thread. The scales seem somewhat arbitrary to me? Chosen to maximise the spread over the viewable area.

The y=x line would not be at 45 degrees given the scales right?

I’m too tired to think through the implications of that. One thing that leaps out is that a true y=x line would have no one over it as nobody can have more xA than key passes assuming that an assist is always a key pass.
 
Looks like the dotted lines represent an average of all the players in the dataset, but I'm not sure. If that's the case, then an argument could be made that Bruno is just an average player with a lot of opportunity, as the line from the origin to his dot runs basically through the "average" intersection.

Here's the graph with the line I meant drawn on it.

View: https://imgur.com/a/F9GehHs


Don’t fully understand why the XA axis starts before 0. Can you have a -xA?

Also, obviously key passes lead to xA so the correlation is always going to be strong. You’d expect the rule to be more key passes = more xA and the chart shows that.

To me that makes it more interesting to look at the outliers rather than the ones on the trend line.

Wirz and Bruno are making more volume of key passes so you’d expect them to be higher up the the xA, Maddison jumps up the xA without going further on key passes. Asks the question if his key passes are of higher quality in terms of being likely to get an assist.

Not all key passes are alike.
 
Football is a difficult sport to quantify numerically, everything is fluid and subjective with the main objective, scoring goals, especially rare.
This this this. I've been banging on about it for years. Loads of people trying make football more data driven but it's not simple.

Even famously complex games like go are much easier to model than football.
 
Back
Top Bottom