Summer Transfer Window 2020

[The Lusophone]

3 hours ago, TCO said:

Is he even really that good? Obviously average is an upgrade over Kepa, but I haven't heard much about Mendy.

Incoming :stats: which I'm sure will make everyone happy.

He's a massive upgrade on Kepa in the areas where Kepa does really really badly. Since August 2018 Kepa has had a save success rate of just 60.6%. Last season his save success rate was even lower - 53.9%. Since August 2018 Mendy's save success rate is 73.5%, which is very good for a team that isn't amongst the best in their league.

You can always use stats and analysis to make a player look better or worse than they are, so I'd remiss if I didn't mention that Kepa has faced a higher quality of shot than Mendy has. Last season Kepa's PSxG of 0.35 is by a ways some distance above Mendy's 0.22 - basically the shots Kepa aren't saving are better than the ones Mendy isn't. At the other end of the same scale though - Kepa expected goals conceded is much much worse than it should be - last season he conceded almost 10 goals more than expected, far and away the worst in the entire Premier League.

WhoScored did a really interesting comparison between Kepa, Mendy, and a number of possible alternatives, including players Chelsea had been linked to, like Ajax's Andre Onana and Mike Maignon of Lille, and ones they hadn't but could perhaps have been available, like Predrag Rajkovic of Reims and Nice stopped Walker Benitez.

They took a look at four metrics - saves per 90 minutes, goals conceded per 90 minutes, save success rate and the average number of saves each keeper has made in matches in which they have conceded one goal or less. That last one is included in an attempt to mitigate the effect a bad defence has on keeper stats - plenty of goalkeepers are made to look better or worse by the poor defence in front of them.

Basically, Kepa is terrible. 1.67 saves per 90, 53.9 per cent save success and just 1.17 saves when conceding one goal or less are all figures you wouldn't want your starting goalkeeper to have, let alone one that cost so much money. He had a disaster of a season.

As for Mendy? He's better than Kepa in each of the four metrics WhoScored looked at, but he's not class leading in any of them - his saves per 90 is 2.54, conceded per 90 is 0.79, save success rate is 75.3% and in that last metric with the long name he provides a figure of 2.3. Those are good, the success rate especially, but someone like Predrag Rajkovic or Cagliari's Alessio Cragni is better than Mendy in all of those.

So why Mendy? He's a bit older than all the other keepers looked at, except Nick Pope, at 28 and in a summer where Chelsea have, Thiago Silva aside, focussed on younger players, might suggest that they still hold out some hope for Kepa to eventually come good, like how David De Gea improved after Fergie dropped him.

Alternatively it could just be the likes of Onana and Maignon were too expensive to buy this summer, with other targets in other positions being prioritised.

tl;dr: Kepa bad, Mendy better, but he wasn't the best choice, stats wise at least.

[Baddar]

Leeds have signed Diego Llorente. 

[The Lusophone]

Good signing that.

[FLiam]

Sky Sports are saying Arsenal will turn their attention to Jorginho if we fail with our attempts to sign Partey and/or Aouar. 

[Lineker]

 

10 hours ago, Jimmy said:

Levy's lost the plot. What's going on? Is he okay?

Such a shame we never backed Poch like this. 

Absolutely - Poch and Mitchell never got this sort of backing! But this Skriniar deal certainly appears to have legs to it:

I think there's appetites on all sides for this to happen.

[METALMAN]

11 hours ago, MadJack said:

Incoming :stats: which I'm sure will make everyone happy.

He's a massive upgrade on Kepa in the areas where Kepa does really really badly. Since August 2018 Kepa has had a save success rate of just 60.6%. Last season his save success rate was even lower - 53.9%. Since August 2018 Mendy's save success rate is 73.5%, which is very good for a team that isn't amongst the best in their league.

You can always use stats and analysis to make a player look better or worse than they are, so I'd remiss if I didn't mention that Kepa has faced a higher quality of shot than Mendy has. Last season Kepa's PSxG of 0.35 is by a ways some distance above Mendy's 0.22 - basically the shots Kepa aren't saving are better than the ones Mendy isn't. At the other end of the same scale though - Kepa expected goals conceded is much much worse than it should be - last season he conceded almost 10 goals more than expected, far and away the worst in the entire Premier League.

WhoScored did a really interesting comparison between Kepa, Mendy, and a number of possible alternatives, including players Chelsea had been linked to, like Ajax's Andre Onana and Mike Maignon of Lille, and ones they hadn't but could perhaps have been available, like Predrag Rajkovic of Reims and Nice stopped Walker Benitez.

They took a look at four metrics - saves per 90 minutes, goals conceded per 90 minutes, save success rate and the average number of saves each keeper has made in matches in which they have conceded one goal or less. That last one is included in an attempt to mitigate the effect a bad defence has on keeper stats - plenty of goalkeepers are made to look better or worse by the poor defence in front of them.

Basically, Kepa is terrible. 1.67 saves per 90, 53.9 per cent save success and just 1.17 saves when conceding one goal or less are all figures you wouldn't want your starting goalkeeper to have, let alone one that cost so much money. He had a disaster of a season.

As for Mendy? He's better than Kepa in each of the four metrics WhoScored looked at, but he's not class leading in any of them - his saves per 90 is 2.54, conceded per 90 is 0.79, save success rate is 75.3% and in that last metric with the long name he provides a figure of 2.3. Those are good, the success rate especially, but someone like Predrag Rajkovic or Cagliari's Alessio Cragni is better than Mendy in all of those.

So why Mendy? He's a bit older than all the other keepers looked at, except Nick Pope, at 28 and in a summer where Chelsea have, Thiago Silva aside, focussed on younger players, might suggest that they still hold out some hope for Kepa to eventually come good, like how David De Gea improved after Fergie dropped him.

Alternatively it could just be the likes of Onana and Maignon were too expensive to buy this summer, with other targets in other positions being prioritised.

tl;dr: Kepa bad, Mendy better, but he wasn't the best choice, stats wise at least.

you are what the tabloid media would call.a boffin. Pretty sure we voted you people out in June 2016.

[Roster Depth Sicko No. 2]

FM Wonderkid Florentino Luis has joined Monaco

And Mbwanna Samatta has left Aston Villa for Fenerbache.

 

[Rich]

23 hours ago, MadJack said:

Incoming :stats: which I'm sure will make everyone happy.

He's a massive upgrade on Kepa in the areas where Kepa does really really badly. Since August 2018 Kepa has had a save success rate of just 60.6%. Last season his save success rate was even lower - 53.9%. Since August 2018 Mendy's save success rate is 73.5%, which is very good for a team that isn't amongst the best in their league.

You can always use stats and analysis to make a player look better or worse than they are, so I'd remiss if I didn't mention that Kepa has faced a higher quality of shot than Mendy has. Last season Kepa's PSxG of 0.35 is by a ways some distance above Mendy's 0.22 - basically the shots Kepa aren't saving are better than the ones Mendy isn't. At the other end of the same scale though - Kepa expected goals conceded is much much worse than it should be - last season he conceded almost 10 goals more than expected, far and away the worst in the entire Premier League.

WhoScored did a really interesting comparison between Kepa, Mendy, and a number of possible alternatives, including players Chelsea had been linked to, like Ajax's Andre Onana and Mike Maignon of Lille, and ones they hadn't but could perhaps have been available, like Predrag Rajkovic of Reims and Nice stopped Walker Benitez.

They took a look at four metrics - saves per 90 minutes, goals conceded per 90 minutes, save success rate and the average number of saves each keeper has made in matches in which they have conceded one goal or less. That last one is included in an attempt to mitigate the effect a bad defence has on keeper stats - plenty of goalkeepers are made to look better or worse by the poor defence in front of them.

Basically, Kepa is terrible. 1.67 saves per 90, 53.9 per cent save success and just 1.17 saves when conceding one goal or less are all figures you wouldn't want your starting goalkeeper to have, let alone one that cost so much money. He had a disaster of a season.

As for Mendy? He's better than Kepa in each of the four metrics WhoScored looked at, but he's not class leading in any of them - his saves per 90 is 2.54, conceded per 90 is 0.79, save success rate is 75.3% and in that last metric with the long name he provides a figure of 2.3. Those are good, the success rate especially, but someone like Predrag Rajkovic or Cagliari's Alessio Cragni is better than Mendy in all of those.

So why Mendy? He's a bit older than all the other keepers looked at, except Nick Pope, at 28 and in a summer where Chelsea have, Thiago Silva aside, focussed on younger players, might suggest that they still hold out some hope for Kepa to eventually come good, like how David De Gea improved after Fergie dropped him.

Alternatively it could just be the likes of Onana and Maignon were too expensive to buy this summer, with other targets in other positions being prioritised.

tl;dr: Kepa bad, Mendy better, but he wasn't the best choice, stats wise at least.

Talking to my TIFO loving heart here

[Colly]

I know I keep banging on about this, but is there any detail anywhere on how xG and all the other various x stats are inputted and collected? I analyse data for a living and the first thing you have to consider is the quality of what's going in in the first place, is there basically an Opta person watching the telly and estimating the various factors that comprise xG?

It's great that you can take that and knock up xG conceded and all the weird permutations of that, but if that initial data collection is completely subjective it all collapses at the first hurdle.

[DavidMarrio]

Don't talk about goals talk about XG is the new Don't talk about spend talk about netspend 

[Gazz]

24 minutes ago, Colly said:

I know I keep banging on about this, but is there any detail anywhere on how xG and all the other various x stats are inputted and collected? I analyse data for a living and the first thing you have to consider is the quality of what's going in in the first place, is there basically an Opta person watching the telly and estimating the various factors that comprise xG?

It's great that you can take that and knock up xG conceded and all the weird permutations of that, but if that initial data collection is completely subjective it all collapses at the first hurdle.

xG values were calculated by analysing thousands of shots and noting whether they resulted in a goal (1) or not (0). Then the conversion ratio was calculated based on the percentage of goals versus no goals. Every place on the football pitch was then assigned a value based on this calculation. So it's literally noting where the shot came from and assigning it a value based on a matrix.

It's not shown on that map as it's classed as a special occurrence but, for example, the dataset suggested that on average, for every 1000 penalties taken, 760 resulted in a direct goal, therefore 760/1000 = .76, which is why every penalty opportunity awards a team or individual player .76 xG.

Problem is, xG doesn't account for a variety of factors, defensive pressure (a 30 yard full bloot volley has the same xG as a 30 yard roll-in in the 94th minute when the keeper is still in the opposition box), type of shot, where the ball is being struck from (ground, knee height, head height etc.), weather, and it doesn't factor in for who is taking the shot. According to xG, I have the same probability of scoring a penalty as Shola Ameobi.

It has it's uses (teams that create a lot of shots in and around the six yard box will generally score more goals) but it's not the be all and end all of football statistics.

[Colly]

I genuinely didn't realise it was that blunt an instrument and now hate it even more than previously.

[gunnar hendershow]

I think with analytics, especially in a sport with so many likely unaccounted variables, you can't use just one as an end-all "this player is better than this player". It's not like baseball, where a lot of sports analytics began, where you can quantify it easier because of the individual performance nature of the sport. It's like in basketball when PER (player efficiency rating) came into vogue it was based on a number of different statistics. Looking at just xG won't give you anywhere close to the whole story. It's still nifty when looking at otherwise identical or nearly identical players though.

[Roster Depth Sicko No. 2]

 

[Rich]

1 hour ago, Colly said:

I genuinely didn't realise it was that blunt an instrument and now hate it even more than previously.

xG is a better metric than shots or even goals, which are even more blunt, because it is all about teams and players putting themselves in the best position to score.

It's why possession and pass completion are not great stats either.

If you take a lot of stats at face value for instance Jack Grealish last year was the best English midfielder, if you then took his stats as a per 90, he was still good, but was massively outshone in all reasonable metrics by Phil Foden. but even then you lose context of Foden being at City adn Grealish at Villa, and what the opposition they faced are.  Very few metrics are perfect other than the final score lol.

[Lineker]

 

[METALMAN]

I think data can be used in a number of very interesting and very useful ways but I think when we talk about it or use it to make decisions without understanding how it was collected and analysed - or knowing how to interpret it - we have problems. (Well not really problems because it’s football and nobody dies but you know what I mean.)

I didn’t really know much about how xG works (thanks gazz for very good explainer) but I am innately sceptical about any combined metric that attempts to force a number variables into one number. My background is political science and we get these all the time. A lot of my students will say “country X scored 3/10 on the Freedom House index so that means it’s authoritarian” and I’m like “okay but how did they get that information, what does 3/10 mean, how did they calculate it, what factors do they consider, what don’t they consider”. 
 

And ultimately a country getting 3/10 on Freedom House will tell you little more beyond the fact that they entered the data into their algorithm and it spat out 3/10. But that doesn’t tell you anything about the many complicated factors that can determine how authoritarian a country is.

I’d always had an innate scepticism of xG for similar reasons. I’m not saying it’s useless, but I’d question how reliable it is, and how reflective that number is to reality. Use with caution, basically. Or use with data from another source to back up the point you’re making, which is what I’d tell my students.

[Colly]

It's a good way to flog books to gullible idiots though.

https://www.amazon.co.uk/dp/1089883188/ref=cm_sw_r_cp_apa_i_M6IBFb921X3WZ

Some of the claims in the description are fanciful to say the least.

[METALMAN]

Yeah, the way it’s portrayed as the answer to everything in that kind of snake oily way has made me even more sceptical than I usually would be.

[DavidMarrio]

I just get nightmares of when we tried "moneyball" and the results were Downing and Charlie Adam...