Turnbull's move to celtic could be off

Status
Not open for further replies.
I get where you're going, but you're looking at it the wrong way IMO.

Instead of looking at his statistical output in terms of raw numbers (goals, assists) and basing your assessment of his potential based on that, you should base your assessment of his potential on the metrics.

The headline is that Celtic are paying £3.25m for a midfielder from Motherwell who scored 15 goals last season. On the face of it, 15 goals last season is a very impressive figure.

BUT

Three of those goals were penalties. While penalties still have to be scored, they're not really a good measure of your worth as a player, because they're high percentage scoring chances and they don't come from open play. And he is unlikely to be taking penalties for Celtic.

Therefore, Turnbull scored 12 goals from open play. Again, still an impressive number, but not as impressive as 15.

Digging into those twelve goals, using an xG model, a significant amount of them were either long range shots that were deflected, or shots that the keeper should have saved. His xG total means that from the chances Turnbull had, you would expect him to score only 4 goals. So he got very, very lucky to score 12 goals from those opportunities. Even if he was a very good finisher, you would only expect him to put away 4/5/6 goals. 12 is a stastical outlier that simply has to regress to the mean.

So the headline is that Celtic are buying a 15 goal-per-season midfielder. But if Turnbull hadn't got lucky, Celtic would be buying a goal-scoring midfielder who scored 4 goals and 3 penalties. For £3.25m. From Motherwell.

And FWIW, Turnbull is a goal scoring midfielder, because the metrics also show:

His defensive stats are poor compared to other SPFL midfielders (doesn't win the ball back, doesn't make interceptions, average at closing down)
His playmaking stats are league average at best (average for xA, average at dribbling, average for giving the ball away).

The source:
Good post mate. Makes you wonder what Celtic are thinking as they’ll be looking at exactly the same raw data.
 
They've had a bid accepted for this French defender and now they cant afford Turnbulls fee.
"The banter years" patter is embarrassing btw.
At least wait till we actually win something.

Don’t see that bid for the French CD being accepted in the media Teddy. Where you seeing that ? I only see an interest in what I’m reading.
 
Surely they are way to far down the line to pull the plug based on performance stats??
Celtic will have had this information well in advance of making an offer.

Whether they actually incorporate advanced metrics into their decision making process, I don't know.
 
Good post mate. Makes you wonder what Celtic are thinking as they’ll be looking at exactly the same raw data.
They may not consider it to be very important. They also may use a different model which assesses xG differently and therefore have his xG as higher.

But even if they do, it's worth pointing out that StatsBomb are as close to industry leaders as you get for this kind of thing. They're really clever guys who do a lot of work for other clubs.
 
It's based on a number of factors (and Statsbomb won't tell you exactly how it's broken down in their model, for obvious reasons) but broadly speaking the factors for xG:

The location of the shot
Type of shot (e.g. if it's a header, a volley etc.)
How the player received the ball

There will be some others mixed in, but here's Statsbomb's own analysis on the strength and weakness of xG:

https://statsbomb.com/2018/05/the-dual-life-of-expected-goals-part-1/

Ok, that makes a bit more sense.

Whatever way you look at it, £3.25m for this kid is crazy. It makes you wonder if their scouts looked at him in depth or someone higher up is sanctioning the signing because of the wee bit of buzz that's built up around the player.

The goals aren't really the part that stick out to me, it's that he seems to be fairly shît at everything else. He barely had any open play assists and even then, some of the ones he did have were unintentional. His passing/build-up play/break-up play/control of the midfield seems pretty non-existent.
 
They may not consider it to be very important. They also may use a different model which assesses xG differently and therefore have his xG as higher.

But even if they do, it's worth pointing out that StatsBomb are as close to industry leaders as you get for this kind of thing. They're really clever guys who do a lot of work for other clubs.

Hope you & they are right mate. Interesting new science ;)
 
Ok, that makes a bit more sense.

Whatever way you look at it, £3.25m for this kid is crazy. It makes you wonder if their scouts looked at him in depth or someone higher up is sanctioning the signing because of the wee bit of buzz that's built up around the player.

The goals aren't really the part that stick out to me, it's that he seems to be fairly shît at everything else. He barely had any open play assists and even then, some of the ones he did have were unintentional. His passing/build-up play/break-up play/control of the midfield seems pretty non-existent.
FWIW, I think analytics like this are most useful for two things:
1) They show you when a player is getting lucky, and is due for regression
2) You can use them to set thresholds for potential signings, and eliminate your targets on this basis. E.g. if we wanted a playmaker, we might say "I only want a playmaker with an xA per game of 0.6. Everyone lower than this doesn't make the cut".

As far as I know, this is the way that most modern football clubs use analytics for their recruitment.

So it does make you wonder what thresholds Celtic set for their Ntcham replacement, if they even bothered to look at the data at all.
 
I'm unable to question the content of these type of statistics - I hated this subject when I was at college - no doubt because I couldn't understand it

So on a dummy level - I have difficulty seeing just how significant any conclusions arrived at are

On a simple level - surely someone in a struggling team will end up running greater distances trying to help out exposed areas & / or teamates on the park & depending on the opposition - will be 'forced' to pass more or less if they zonally mark your own players ?

On that basis statistics gathered from one match to another are imo likely to tell you more about the work rate of your own teamates or the opposition than the player being studied

Sure the figures might well give you a guideline - but basing much credence, making conclusions & decisions on this data seems absurd to me
 
I'm unable to question these type of statistics - I hated this subject when I was at college - no doubt because I couldn't understand it

So on a dummy level - I have difficulty seeing just how significant any conclusions arrived at are

On a simple level - surely someone in a struggling team will end up running greater distances trying to help out exposed areas & / or teamates on the park & depending on the opposition - will be 'forced' to pass more or less if they zonally mark your own players ?

On that basis statistics gathered from one match to another are imo likely to tell you more about the work rate of your own teamates or the opposition than the player being studied

Sure the figures might well give you a guideline - but basing much credence, making conclusions & decisions on this data seems absurd to me
You're right - so the model takes that into account.

For example, a team who sees less of the ball and is lower down the league will play more defensively than a team higher up the league. St Mirren see less of the ball than Rangers do.

Sticking with defensive statistics, if you look at Turnbulls metrics again, you'll notice that some of the statistics are "PAdj Tackles" and "PAdj Interceptions" (About ten / eleven o' clock on the radar). That stands for possession-adjusted tackles, and possession-adjusted interceptions. So in laymans terms, those statistics take into account that St Mirren are under a lot more pressure defensively than Rangers are, so the number of tackles that St Mirren players make are "weighted"accordingly.

To use a non-footballing example, if you ever look into political polling, you'll see that they "weight" the polls to account for the fact that they're using a very small snapshot of people. So they "weight" the polls according to the demographics of the country.
 
Last edited:
They may not consider it to be very important. They also may use a different model which assesses xG differently and therefore have his xG as higher.

But even if they do, it's worth pointing out that StatsBomb are as close to industry leaders as you get for this kind of thing. They're really clever guys who do a lot of work for other clubs.
Yeah I use them and other models a lot for (whisper it) fantasy football. It’s amazing how accurate xG and xA is over the course of the season!
 
Yeah I use them and other models a lot for (whisper it) fantasy football. It’s amazing how accurate xG and xA is over the course of the season!
They're really useful predictive tools. In America, you will notice that a lot of the fantasy sports community are also heavily into analytics because they're so predictive.
 
You're right - so the model takes that into account.

For example, a team who sees less of the ball and is lower down the league will play more defensively than a team higher up the league. St Mirren see less of the ball than Rangers do.

Sticking with defensive statistics, if you look at Turnbulls metrics again, you'll notice that some of the statistics are "PAdj Tackles" and "PAdj Interceptions" (About ten / eleven o' clock on the radar). That stands for pressure-adjusted tackles, and pressure-adjusted interceptions. So in laymans terms, those statistics take into account that St Mirren are under a lot more pressure defensively than Rangers are, so the number of tackles that St Mirren players make are "weighted"accordingly.

To use a non-footballing example, if you ever look into political polling, you'll see that they "weight" the polls to account for the fact that they're using a very small snapshot of people. So they "weight" the polls according to the demographics of the country.

Thanks - I did say I'd difficulty understanding this - but your explanation does help

I'm still unsure just what conclusions I'd be willing to make using that data though
 
Is the guy who negotiated our kit contract with Ashley now handling the transfers at the Knew Camp?
 
Thanks - I did say I'd difficulty understanding this - but your explanation does help

I'm still unsure just what conclusions I'd be willing to make using that data though
It can be a difficult thing to wrap your head around, but analytics has proven to be quite good at predicting regression across a number of sports, not just football.

If you're interested in it I'd recommend looking it up some more. It's been the single biggest change across sports in the last decade IMO.
 
This is a nice mental picture, but factually incorrect. Johnston was paraded by them a few days before the 1990 Cup Final versus us. There followed the close season where the deal wasn't completed till him signing for us some 5/6 weeks later.
He signed in 89
 
It can be a difficult thing to wrap your head around, but analytics has proven to be quite good at predicting regression across a number of sports, not just football.

If you're interested in it I'd recommend looking it up some more. It's been the single biggest change across sports in the last decade IMO.

So does this mean that because Messi has never again scored 95 goals in a season as he did in 2010/11 (or 11/12?) he's been in regression ever since?
 
So does this mean that because Messi has never again scored 95 goals in a season as he did in 2010/11 (or 11/12?) he's been in regression ever since?
No. Because that is statistical output rather than anything the metrics are showing.

Metrics continue to show that Messi is an incredible player and isn't showing any signs of regression at all. His xG remains remarkably solid year to year, and you won't be surprised to hear he outperforms it significantly.
 
No. Because that is statistical output rather than anything the metrics are showing.

Metrics continue to show that Messi is an incredible player and isn't showing any signs of regression at all. His xG remains remarkably solid year to year, and you won't be surprised to hear he outperforms it significantly.

Whilst I appreciate Messi could be described as a phenomenon, surely if he is regularly outperforming his xG and xA it calls their validity into question? At a lower level surely any other player could regularly outperform their own xG and xA.

I love these stats things by the way - fascinates me.:)
 
He signed in 89
3 days after my daughter was born. The Tim who was in visiting his wife in the bed across from my good lady was a absolutely shattered by the news. I had to add to his misery with an overstated "check oot who weve just signed " with some laughter thrown in :))
 
3 days after my daughter was born. The Tim who was in visiting his wife in the bed across from my good lady was a absolutely shattered by the news. I had to add to his misery with an overstated "check oot who weve just signed " with some laughter thrown in :))
Rules FFS
 
No. Because that is statistical output rather than anything the metrics are showing.

Metrics continue to show that Messi is an incredible player and isn't showing any signs of regression at all. His xG remains remarkably solid year to year, and you won't be surprised to hear he outperforms it significantly.

I don't even have an Xbox to fall back on, I'll stick to books and the old maxim that there are lies, damned lies and statistics.

I love Tav offensively, I don't defensively. Are my eyes deceiving me when I think that a helluva lot of threats to our goal come down our right/opposition left wing?

Or are there stats proving otherwise.

Maybe I'm even more of a dinosaur than I think, because in my opinion we should play and buy players on actual skill, not on what a graph tells us.
 
I get where you're going, but you're looking at it the wrong way IMO.

Instead of looking at his statistical output in terms of raw numbers (goals, assists) and basing your assessment of his potential based on that, you should base your assessment of his potential on the metrics.

The headline is that Celtic are paying £3.25m for a midfielder from Motherwell who scored 15 goals last season. On the face of it, 15 goals last season is a very impressive figure.

BUT

Three of those goals were penalties. While penalties still have to be scored, they're not really a good measure of your worth as a player, because they're high percentage scoring chances and they don't come from open play. And he is unlikely to be taking penalties for Celtic.

Therefore, Turnbull scored 12 goals from open play. Again, still an impressive number, but not as impressive as 15.

Digging into those twelve goals, using an xG model, a significant amount of them were either long range shots that were deflected, or shots that the keeper should have saved. His xG total means that from the chances Turnbull had, you would expect him to score only 4 goals. So he got very, very lucky to score 12 goals from those opportunities. Even if he was a very good finisher, you would only expect him to put away 4/5/6 goals. 12 is a stastical outlier that simply has to regress to the mean.

So the headline is that Celtic are buying a 15 goal-per-season midfielder. But if Turnbull hadn't got lucky, Celtic would be buying a goal-scoring midfielder who scored 4 goals and 3 penalties. For £3.25m. From Motherwell.

And FWIW, Turnbull is a goal scoring midfielder, because the metrics also show:

His defensive stats are poor compared to other SPFL midfielders (doesn't win the ball back, doesn't make interceptions, average at closing down)
His playmaking stats are league average at best (average for xA, average at dribbling, average for giving the ball away).

The source:

Excellent post, there's information waiting to be parsed out all over football that means you can make far fewer howlers in the transfer market, if you know what you're looking for.

I'd said that we'd probably looked at his game tape and saw something completely different to what the Tims saw and we also didn't have the pressure of expectation after the business leaked out.

If their transfer sheet was true, then it alluded to a real lack of insight into what they're researching as I could produce a far better summary on players in one line than they were doing.
 
Whilst I appreciate Messi could be described as a phenomenon, surely if he is regularly outperforming his xG and xA it calls their validity into question? At a lower level surely any other player could regularly outperform their own xG and xA.

I love these stats things by the way - fascinates me.:)
Yes, and that absolutely was the case in the early days of xG. The metric needed to be refined, as you'd expect - it's nearly impossible to produce a statistical model and get it right first time.

However, statistical models get better the more data that you feed into them. Because the more data you get, the more you can correct obvious errors. For example, if Fernando Llorente's xG is lower than his goals, and he scores most of his goals through headers, then maybe they undervalued the likelihood of a header to be a goal. As time has gone on and thousands of games have been played each season, the models have been refined to the extent that they are now pretty solid. But you're absolutely right that in the early days, that could be an indicator that the model was wrong.

Nowadays, a player consistently outperforming his xG is generally taken to mean that the player is actually just a really good finisher.

But here's why that (probably) doesn't apply to Turnbull:
1. His sample size is small
2. Even if he was as good a finisher as Messi, his scoring rate is still unsustainable.

Messi had an xG of 19.02 last season and scored 32 goals from 166 shots. So he outperformed his xG by 12.98, or 68%.

Turnbull had an xG of 4.45 and scored 12 goals from 57 shots. So he outperformed his xG by 7.55, or 169%.

The other point: Messi isn't a good example to use, because his streak of outperforming xG is so incredible that no other player comes close to it. Mbappe, for example, had an xG of 24.08 and scored 31 goals - a difference of 6.92, or 28%.

The stats really show just how incredible Messi is.
 
Unless its career threatening the fee shouldnt change and mwell should stick to their original demands. If im wrong Kieran wunderboy must be worth heehaw.
 
I get where you're going, but you're looking at it the wrong way IMO.

Instead of looking at his statistical output in terms of raw numbers (goals, assists) and basing your assessment of his potential based on that, you should base your assessment of his potential on the metrics.

The headline is that Celtic are paying £3.25m for a midfielder from Motherwell who scored 15 goals last season. On the face of it, 15 goals last season is a very impressive figure.

BUT

Three of those goals were penalties. While penalties still have to be scored, they're not really a good measure of your worth as a player, because they're high percentage scoring chances and they don't come from open play. And he is unlikely to be taking penalties for Celtic.

Therefore, Turnbull scored 12 goals from open play. Again, still an impressive number, but not as impressive as 15.

Digging into those twelve goals, using an xG model, a significant amount of them were either long range shots that were deflected, or shots that the keeper should have saved. His xG total means that from the chances Turnbull had, you would expect him to score only 4 goals. So he got very, very lucky to score 12 goals from those opportunities. Even if he was a very good finisher, you would only expect him to put away 4/5/6 goals. 12 is a stastical outlier that simply has to regress to the mean.

So the headline is that Celtic are buying a 15 goal-per-season midfielder. But if Turnbull hadn't got lucky, Celtic would be buying a goal-scoring midfielder who scored 4 goals and 3 penalties. For £3.25m. From Motherwell.

And FWIW, Turnbull is a goal scoring midfielder, because the metrics also show:

His defensive stats are poor compared to other SPFL midfielders (doesn't win the ball back, doesn't make interceptions, average at closing down)
His playmaking stats are league average at best (average for xA, average at dribbling, average for giving the ball away).

The source:

Think I'll stick with the tried and trusted method of using the old mince pies.

Incidentally, how do these metrics stack up on Tavernier?
 
I don't even have an Xbox to fall back on, I'll stick to books and the old maxim that there are lies, damned lies and statistics.

I love Tav offensively, I don't defensively. Are my eyes deceiving me when I think that a helluva lot of threats to our goal come down our right/opposition left wing?

Or are there stats proving otherwise.

Maybe I'm even more of a dinosaur than I think, because in my opinion we should play and buy players on actual skill, not on what a graph tells us.
There are a lot of analytics truthers out there, and that's okay. The truth is that analytics can't tell you everything, and that watching a player to determine his skills is still every bit as important in the scouting process.

Andy Scoulding said himself that data analysis and getting eyes on the player were both crucial components of Rangers recruitment strategy.

The Tav question is interesting, and it's one I cannot answer because I don't have the data unfortunately.
 
You're right - so the model takes that into account.

For example, a team who sees less of the ball and is lower down the league will play more defensively than a team higher up the league. St Mirren see less of the ball than Rangers do.

Sticking with defensive statistics, if you look at Turnbulls metrics again, you'll notice that some of the statistics are "PAdj Tackles" and "PAdj Interceptions" (About ten / eleven o' clock on the radar). That stands for pressure-adjusted tackles, and pressure-adjusted interceptions. So in laymans terms, those statistics take into account that St Mirren are under a lot more pressure defensively than Rangers are, so the number of tackles that St Mirren players make are "weighted"accordingly.

To use a non-footballing example, if you ever look into political polling, you'll see that they "weight" the polls to account for the fact that they're using a very small snapshot of people. So they "weight" the polls according to the demographics of the country.
....and they've been spectacularly wrong in recent years.
 
Excellent post, there's information waiting to be parsed out all over football that means you can make far fewer howlers in the transfer market, if you know what you're looking for.

I'd said that we'd probably looked at his game tape and saw something completely different to what the Tims saw and we also didn't have the pressure of expectation after the business leaked out.

If their transfer sheet was true, then it alluded to a real lack of insight into what they're researching as I could produce a far better summary on players in one line than they were doing.
That's the ultimate goal of analytics that executives across all sports are now realising: analytics can't always identify the next Lionel Messi for you, but it can absolutely stop you from buying Rafael Scheidt.
 
Tam McManus 12:28

David Turnbull has a small problem with his knee that requires an op (tidy up). Be out for 10-12 weeks. That’s it. Some of the shite I have heard about him on here (twitter) is poor & embarrassing. Hurtful rumours about the kid. The deal will 100% still go through.
 
Yes, and that absolutely was the case in the early days of xG. The metric needed to be refined, as you'd expect - it's nearly impossible to produce a statistical model and get it right first time.

However, statistical models get better the more data that you feed into them. Because the more data you get, the more you can correct obvious errors. For example, if Fernando Llorente's xG is lower than his goals, and he scores most of his goals through headers, then maybe they undervalued the likelihood of a header to be a goal. As time has gone on and thousands of games have been played each season, the models have been refined to the extent that they are now pretty solid. But you're absolutely right that in the early days, that could be an indicator that the model was wrong.

Nowadays, a player consistently outperforming his xG is generally taken to mean that the player is actually just a really good finisher.

But here's why that (probably) doesn't apply to Turnbull:
1. His sample size is small
2. Even if he was as good a finisher as Messi, his scoring rate is still unsustainable.

Messi had an xG of 19.02 last season and scored 32 goals from 166 shots. So he outperformed his xG by 12.98, or 68%.

Turnbull had an xG of 4.45 and scored 12 goals from 57 shots. So he outperformed his xG by 7.55, or 169%.

The other point: Messi isn't a good example to use, because his streak of outperforming xG is so incredible that no other player comes close to it. Mbappe, for example, had an xG of 24.08 and scored 31 goals - a difference of 6.92, or 28%.

The stats really show just how incredible Messi is.

I hear what you say, and appreciate the response. However, just in the examples you quote we have three players from last season, Messi, Turnbull and Mbappe (surely the last time DT will be getting a mention in that sort of company!) outperforming their xG by 68%, 169% and 28% respectively. On that basis alone I'd say there's a way to go in refining the statistical model.

As I said, I love this stuff and it undoubtedly has its place - but that's all. Its should be just one of many factors considered in looking at a player (and, in truth, probably is). FWIW, I've seen a bit of Turnbull and I have a mate who's been watching Motherwell for more than 50 years who tells me he's the best prospect they've ever had. He COULD be a great player - but he could just as easily disappear. At £3.5m, at this stage of his career, the Dhims have taken a massive gamble.
 
Last edited:
Think I'll stick with the tried and trusted method of using the old mince pies.

Incidentally, how do these metrics stack up on Tavernier?
Metrics pretty much confirm the narrative for Tavernier in terms of his attacking contribution - he's one of the best players in the league and an incredible attacking force.

Nobody seems to have done much modelling on defensive players in the SPFL - probably because that's less sexy to write about, in all honesty.
 
....and they've been spectacularly wrong in recent years.
Polls are slightly different because:
1) The pollsters probably failed to adjust their weighting in accordance with the rapid change in the British electorate for the 2017 GE
2) For the 2016 US Election, polls did have Hillary favourite, but Trump wasn't a total write off. 538 gave him a 28.6% chance of winning: https://projects.fivethirtyeight.com/2016-election-forecast/

I was trying to make the comparison between the principle of weighting in both polls and advanced metrics.
 
Yes, and that absolutely was the case in the early days of xG. The metric needed to be refined, as you'd expect - it's nearly impossible to produce a statistical model and get it right first time.

However, statistical models get better the more data that you feed into them. Because the more data you get, the more you can correct obvious errors. For example, if Fernando Llorente's xG is lower than his goals, and he scores most of his goals through headers, then maybe they undervalued the likelihood of a header to be a goal. As time has gone on and thousands of games have been played each season, the models have been refined to the extent that they are now pretty solid. But you're absolutely right that in the early days, that could be an indicator that the model was wrong.

Nowadays, a player consistently outperforming his xG is generally taken to mean that the player is actually just a really good finisher.

But here's why that (probably) doesn't apply to Turnbull:
1. His sample size is small
2. Even if he was as good a finisher as Messi, his scoring rate is still unsustainable.

Messi had an xG of 19.02 last season and scored 32 goals from 166 shots. So he outperformed his xG by 12.98, or 68%.

Turnbull had an xG of 4.45 and scored 12 goals from 57 shots. So he outperformed his xG by 7.55, or 169%.

The other point: Messi isn't a good example to use, because his streak of outperforming xG is so incredible that no other player comes close to it. Mbappe, for example, had an xG of 24.08 and scored 31 goals - a difference of 6.92, or 28%.

The stats really show just how incredible Messi is.
There are a lot of analytics truthers out there, and that's okay. The truth is that analytics can't tell you everything, and that watching a player to determine his skills is still every bit as important in the scouting process.

Andy Scoulding said himself that data analysis and getting eyes on the player were both crucial components of Rangers recruitment strategy.

The Tav question is interesting, and it's one I cannot answer because I don't have the data unfortunately.

Cheers for that, but admittedly I did ask as a self confessed dinosaur, so it does take a bit of getting used to.

It makes you wonder on what the stats on the likes of Gerd Muller would have been back in the day?

Cryuff? He'd have been in the same rarefied atmosphere as Messi.
 
I get where you're going, but you're looking at it the wrong way IMO.

Instead of looking at his statistical output in terms of raw numbers (goals, assists) and basing your assessment of his potential based on that, you should base your assessment of his potential on the metrics.

The headline is that Celtic are paying £3.25m for a midfielder from Motherwell who scored 15 goals last season. On the face of it, 15 goals last season is a very impressive figure.

BUT

Three of those goals were penalties. While penalties still have to be scored, they're not really a good measure of your worth as a player, because they're high percentage scoring chances and they don't come from open play. And he is unlikely to be taking penalties for Celtic.

Therefore, Turnbull scored 12 goals from open play. Again, still an impressive number, but not as impressive as 15.

Digging into those twelve goals, using an xG model, a significant amount of them were either long range shots that were deflected, or shots that the keeper should have saved. His xG total means that from the chances Turnbull had, you would expect him to score only 4 goals. So he got very, very lucky to score 12 goals from those opportunities. Even if he was a very good finisher, you would only expect him to put away 4/5/6 goals. 12 is a stastical outlier that simply has to regress to the mean.

So the headline is that Celtic are buying a 15 goal-per-season midfielder. But if Turnbull hadn't got lucky, Celtic would be buying a goal-scoring midfielder who scored 4 goals and 3 penalties. For £3.25m. From Motherwell.

And FWIW, Turnbull is a goal scoring midfielder, because the metrics also show:

His defensive stats are poor compared to other SPFL midfielders (doesn't win the ball back, doesn't make interceptions, average at closing down)
His playmaking stats are league average at best (average for xA, average at dribbling, average for giving the ball away).

The source:

Would you not expect If/when Turnbull signs for the mentally challenged that as they will be attacking more/winning more games/getting into better positions + any attacking mid for them will have less defensive duties than if he were playing for motherwell?

Taking this into consideration will his stats not improve from playing with a bottom 6 side to a team winning things?

Genuine question as statistical based ratings like this fascinate me.
 
Would you not expect If/when Turnbull signs for the mentally challenged that as they will be attacking more/winning more games/getting into better positions + any attacking mid for them will have less defensive duties than if he were playing for motherwell?

Taking this into consideration will his stats not improve from playing with a bottom 6 side to a team winning things?

Genuine question as statistical based ratings like this fascinate me.

I think that's covered in post #463 mate.
 
I hear what you say, and appreciate the response. However, just in the examples you quote we have three players from last season, Messi, Turnbull and Mbappe (surely the last time DT will be getting a mention in that sort of company!) outperforming their xG by 68%, 169% and 28% respectively. On that basis alone I'd say there's a way to go in refining the statistical model.

As I said, I love this stuff and it undoubtedly has its place - but that's all. Its should be just one of many factors considered in looking at a player (and, in truth, probably is). FWIW, I've seen a bit of Turnbull and I have a mate who's had a season ticket at Motherwell for more than 50 years who tells me he's the best prospect they've ever had. He COULD be a great player - but he could just as easily disappear. At £3.5m, at this stage of his career, the Dhims have taken a massive gamble.
First off, you're absolutely right. No one should use analytics as the sole criteria for recruitment. It should be part of a process. Turnbull absolutely passes the eye test, but he fails the mathematical one. It's up to each individual club how much stock they put in either test.

Wrt your other point, firstly I'd note that Messi is the outlier that breaks pretty much every statistical model. He really is genuinely incredible. What analytics have found this season is that he's taking more shots from outside the box when compared to last season (therefore lowering his xG, as long rangers tend to go in less) but his goals scored has remained consistent. You don't need to believe in a model to know that is incredible.

Secondly, the model only predicts the probability of a shot going in regardless of who takes it - xG is fundamentally Expected Goals, i.e. if you hit this exact shot 100 times, how many times does it go in? That doesn't change depending on who the player is - if Umar Sadiq has the exact same opportunity as Kylian Mbappe, the xG will be the same. But Kylian Mbappe's actual goals will be higher, because he his a better finisher, so therefore he will outperform his xG. Umar Sadiq is a poor finisher, so he will probably score less than his xG.

The point as it pertains to Turnbull is that he would have to be a finisher like we have never seen before, and will never see again, to maintain his level of statistical output. Could that happen? Sure. But it's so incredibly unlikely that we're safe in assuming that he is going to regress to the mean, and will go through a dry patch very quickly.
 
I get where you're going, but you're looking at it the wrong way IMO.

Instead of looking at his statistical output in terms of raw numbers (goals, assists) and basing your assessment of his potential based on that, you should base your assessment of his potential on the metrics.

The headline is that Celtic are paying £3.25m for a midfielder from Motherwell who scored 15 goals last season. On the face of it, 15 goals last season is a very impressive figure.

BUT

Three of those goals were penalties. While penalties still have to be scored, they're not really a good measure of your worth as a player, because they're high percentage scoring chances and they don't come from open play. And he is unlikely to be taking penalties for Celtic.

Therefore, Turnbull scored 12 goals from open play. Again, still an impressive number, but not as impressive as 15.

Digging into those twelve goals, using an xG model, a significant amount of them were either long range shots that were deflected, or shots that the keeper should have saved. His xG total means that from the chances Turnbull had, you would expect him to score only 4 goals. So he got very, very lucky to score 12 goals from those opportunities. Even if he was a very good finisher, you would only expect him to put away 4/5/6 goals. 12 is a stastical outlier that simply has to regress to the mean.

So the headline is that Celtic are buying a 15 goal-per-season midfielder. But if Turnbull hadn't got lucky, Celtic would be buying a goal-scoring midfielder who scored 4 goals and 3 penalties. For £3.25m. From Motherwell.

And FWIW, Turnbull is a goal scoring midfielder, because the metrics also show:

His defensive stats are poor compared to other SPFL midfielders (doesn't win the ball back, doesn't make interceptions, average at closing down)
His playmaking stats are league average at best (average for xA, average at dribbling, average for giving the ball away).

The source:

Very interesting piece.

I think that what’s happened is the scum have rushed into this transfer before actually doing any due dilligenece on the boy. They’ve then had an in-depth look at him and thought, WOW, WTF have we done!

They are shameless and would have no qualms in destroying the boy if that suited their agenda
 
Can't help but feel sorry for the lad, one minute he's chasing £20k per week, next he's got his name bandied about the footballing community that he is a crock and facing a future back at Murderwell on what? 1.5k per week (if he is lucky).

One minute on a lads holiday having a celebration, world at his feet, hopefully not indulging in the usual lads holiday excesses then he is off back into the dark sepulchre that is the Celtic medical unit and his dreams have been dashed by an evil empire intent on getting their own back for his public negotiation with them.

Who knows though, he may still sign later today and he has something that can be easily remedied.

Its further embarrassment for the scumbags though, which is nice.
 
Good post mate. Makes you wonder what Celtic are thinking as they’ll be looking at exactly the same raw data.

Lennon has apparently got rid of all the analysis guys they had as well as nutritionists & those in sports science etc.
 
Status
Not open for further replies.
Back
Top