Football Analytics

[bosnian_red]

Thanks

Also they posted this, a detailed explanation of their reports!

https://fbref.com/en/about/scouting-reports-explained

Just reading through, they plan on doing possession adjusted stats too so that'll help a lot in removing bias from certain play styles.

 

[Grande]

The stats are per 90 already!
Dean Henderson shot stopping wise:

• 19/20 prem - +0.14 (77th percentile)
• Last 365 days big 5 leagues and European comps - -0.04 (41st) ... sample size for this one would be really small and hurt by a few mistakes he made in the small sample

His only 2 underlying stats that are 90+ are the "launch %" (percentage of kicks longer than 40 yards) and the average pass length... so for hoofing.
Comparing De Gea and him for crosses stopped %, De Gea last season was at 5.3% (27th percentile), Henderson at 7.9% (79th percentile). De Gea this season down in the 7th percentile though, which makes sense as it's been awful this year.

Now there isn't a stat for what counts as a stoppable cross, so could be bad luck/low sample size... but I doubt it. Every player has the same bias here so it wouldn't affect it that much.

Also a thing to note - comparing individual seasons is league specific, comparing past 365 days makes it compare all top 5 leagues. So play styles across competitions would influence it a lot more, significantly harder to stop crosses in the premier league probably with the more physical style (and as we've seen so many goalkeepers struggle in that aspect when they come here).

Thanx!

 

[Teja]

I was about to create this exact same thread with the Rashford data because I saw it floating around on Twitter. I haven't found anything super interesting that's United related yet, but the Statsbomb data / analytical process is pretty incredible. Here's one on Cavani:

Tweet

— Twitter API (@user) date

I've been binging on videos from a conference they did in 2019. My favorite is xT (expected threat). It's a bit technical, but I highly recommend if you're an engineer / math person / data analyst. It's also super super new, so data isn't available yet on sites like fbref to play around with.

 

[Classical Mechanic]

I was about to create this exact same thread with the Rashford data because I saw it floating around on Twitter. I haven't found anything super interesting that's United related yet, but the Statsbomb data / analytical process is pretty incredible. Here's one on Cavani:

Tweet

— Twitter API (@user) date

Its a good time for English strikers too!

 

[Stack]

Its a good time for statisticians, so many new stats to invent to fool the masses with.

 

[duffer]

Who records this sort of data?

Can't be people watching TV coverage as replays are often shown when the game is ongoing so passes, tackles, etc would be missed. Do they have a different broadcast feed or are the numbers just wrong?

 

[bp19992]

As an actuary I work a lot with predictive modelling (predicting fair risk premiums for motor insurance based on several attributes of the insurance holder for example), so this new stats obsession in football is quite exciting for me.
If anyone wants to go a little more into the machine learning details this is exciting:
https://medium.com/analytics-vidhya...science-with-my-soccer-obsession-f81d721432c7

However people should be careful with interpreting the results and without watching the matches and just looking at the numbers you can't judge a player's performance well.
Best example was Michael Carrick, he usually didn't excel at most of the usual metrics used to describe a player's performance. He was excellent though in closing spaces between players, so that opponent's key passes won't even be played, you won't find that in any stats or models.
My experience is that when you've watched (and played) football for years and your gut tells your that a player performed well, but the stats tell you the opposite, you should usually question the stats more than your gut. Quite often they're wrongly interpreted or a lot of key data ist missing.

 

[Chipper]

No great to place to post this so this thread will do. I was curious about the xG of Vardy's goal against Liverpool with an empty net from about 7 yards, Understat gave it a 0.59.

That instintictevly sounds wrong to me. A penalty is about 0.76 from memory, are they really saying he had less chance of scoring there than a player taking a pen and that your average player fails to score 4 out of 10 times from there? Can't see how that's possibly true.

Would like to see what other xG models made of that chance.Would assume their's can't be taking account no goalie being there.

 

[giorno]

Infogol has it .68

Fbref uses statsbomb which have the most advanced/sophisticated model(taking defenders position into account, height of the ball, etc) and i would guess for them it's something like .9 something...

 

[bosnian_red]

https://www.skysports.com/football/...hy-this-stats-revolution-will-change-the-game

Good read, hope United gets on board with this asap too as I think it's going to be the next big revolutionary step in scouting and analyzing both players and coaches. So much of what was previously hard to capture or track will be available to see.

 

[RooneyLegend]

Won't help a jot.

 

[Smores]

https://www.skysports.com/football/...hy-this-stats-revolution-will-change-the-game

Good read, hope United gets on board with this asap too as I think it's going to be the next big revolutionary step in scouting and analyzing both players and coaches. So much of what was previously hard to capture or track will be available to see.

Yeah an interesting piece albeit a bit of a sales pitch.

You wonder how those guys would come in and review some of our players and movement. There's bits of our play as a team and as individuals that repeatedly show up without correction, makes me wonder about our analysts.

Mind you LvG couldn't even get them to watch their own footage without them sabotaging it.

 

[bosnian_red]

Won't help a jot.

It already made a big difference to the few teams that took it on when they first started out and only a few teams were using it. And then the others had to catch up. Will be the same right now, and you'll see a massive use of it for sure when recruiting new players.

 

[bosnian_red]

Yeah an interesting piece albeit a bit of a sales pitch.

You wonder how those guys would come in and review some of our players and movement. There's bits of our play as a team and as individuals that repeatedly show up without correction, makes me wonder about our analysts.

Mind you LvG couldn't even get them to watch their own footage without them sabotaging it.

I think certain aspects are more what does a player inherently do, with little influence from coaching. Space awareness and decision making at all times is just natural movements and more down to the individual imo, with only so much you can do from a coaching perspective. Hard to coach how someone like Scholes was literally always available in space to pick up the ball, and always knew how to play the line breaking passes or make the right decisions to get the ball in the best areas - even if sometimes there are a few decent options. So this new tech will theoretically be able to quantify/rank players who always seem to be open and the value of passes they play.

 

[TheMagicFoolBus]

https://www.skysports.com/football/...hy-this-stats-revolution-will-change-the-game

Good read, hope United gets on board with this asap too as I think it's going to be the next big revolutionary step in scouting and analyzing both players and coaches. So much of what was previously hard to capture or track will be available to see.

Can confirm that Knutson is as insufferable in person as he is in interviews, but StatsBomb is definitely legit. I suspect that it might be hard to use this sort of thing for training as the level of repetition needed might prove burdensome, but these tools should be very valuable at measuring instinctive reading of the game for recruitment purposes. And actually maybe for training, integrating this kind of thing with VR might be interesting - maybe a way to improve spatial awareness without necessarily raising injury risk?

 

[Champ]

Analysis of football has been taken way too far, for starters football is absolutely 50% luck and 50% skill, the managers job is inherently to try to reduce the luck factor down as much as possible, and yes those figures are correct (football analysis is to thank for that).

This is where things like xT (expected threat) as mentioned in a few posts on this thread is taking things way to far - this trys to prove, by way of a mathmatical sum, how credit should go to the players who create the pass for the assist, or who play the pass before the assist pass etc, or how dangerous a player is depending on their whereabouts on the pitch.
This is all well and good, and certainly has it's uses, however the example the creater of xT used on his website was Kolasinac providing an assist for an Arsenal goal V Burnley, and using the sum he created he proved that Ozil, who played the pass to Kolasinac, should get around 85/86% of the credit for that goal.
This to me is absurd. Many factors need to be taken into consideration, especially in the goal in question (Arsenals third v Burnley in 2018 I think), such as the dummy run from Guendouzi, which created space for Ozil, the run from Kolasinac without which Ozil wouldn't have a pass, the static defending by Burnley to allow Abameyang to nip in a tuck the ball away. the original movement from Aubameyang to take the defender out of position.

The above scenario shows that Ozil is to be credited for a great pass, but using the baseline of 85% credit that xT gives Ozil, thats taking away from some other fine play from his teammates.

Analysis undoutedly has it's benefits and will provide teams with superior intelligence, however it is in danger of being relied upon too much. The human element of football cannot be undone by stats and mathmatical equations.

 

[Pagh Wraith]

Can confirm that Knutson is as insufferable in person as he is in interviews, but StatsBomb is definitely legit. I suspect that it might be hard to use this sort of thing for training as the level of repetition needed might prove burdensome, but these tools should be very valuable at measuring instinctive reading of the game for recruitment purposes. And actually maybe for training, integrating this kind of thing with VR might be interesting - maybe a way to improve spatial awareness without necessarily raising injury risk?

I like Knutson The StatsBomb podcasts have always been pretty enjoyable. I'd rather hear those guys talk about football than the pundits on TV who can't come up with anything insightful beyond meaningless platitudes.

Not surprised Liverpool are one of the first to use 360.

 

[RooneyLegend]

It already made a big difference to the few teams that took it on when they first started out and only a few teams were using it. And then the others had to catch up. Will be the same right now, and you'll see a massive use of it for sure when recruiting new players.

Who has it helped in recruitment.

 

[TheMagicFoolBus]

I like Knutson The StatsBomb podcasts have always been pretty enjoyable. I'd rather hear those guys talk about football than the pundits on TV who can't come up with anything insightful beyond meaningless platitudes.

Not surprised Liverpool are one of the first to use 360.

Agreed that that sort of content is far better than TV punditry! "Insufferable" is a bit harsh, he's just quite full of himself and has a tendency to talk down to others, I've found. Nowhere near as aggravating as the likes of Michael Caley though.

 

[TheMagicFoolBus]

Who has it helped in recruitment.

RB Leipzig and Salzburg have been some of the early adopters and had massive success in identifying undervalued players with the potential to go on to bigger and better things.

 

[bosnian_red]

Who has it helped in recruitment.

Well Liverpool very publicly and recently put a lot into advanced analytics to fuel a lot of their decision making, led to some pretty decent results over the past few years for them... Knutson often references his time with Midtylland and they all of a sudden turned into a top danish side. It's pretty much everywhere now, but the early adopters definitely benefitted hugely and led to tangible success. It's nothing surprising though, you would expect advanced stats to always help out with making better recruitment decisions than if you didn't have this data. Happened a lot in every other sport and it's happened with football too. Getting the early edge is where the biggest benefits lie though, because eventually everyone else will catch on (and if not, they'll really fall behind as we've seen often).

Anyway, there's loads of data out there that I can't be arsed to pull up. Also not like I have the specific information. But using advanced data can only help with recruitment.

 

[RooneyLegend]

Well Liverpool very publicly and recently put a lot into advanced analytics to fuel a lot of their decision making, led to some pretty decent results over the past few years for them... Knutson often references his time with Midtylland and they all of a sudden turned into a top danish side. It's pretty much everywhere now, but the early adopters definitely benefitted hugely and led to tangible success. It's nothing surprising though, you would expect advanced stats to always help out with making better recruitment decisions than if you didn't have this data. Happened a lot in every other sport and it's happened with football too. Getting the early edge is where the biggest benefits lie though, because eventually everyone else will catch on (and if not, they'll really fall behind as we've seen often).

Anyway, there's loads of data out there that I can't be arsed to pull up. Also not like I have the specific information. But using advanced data can only help with recruitment.

There's nothing all that impressive about liverpool's recruitment, their coaching is where it's at. Football is a very complex sport with alot of variations. Alot of different types of tactics and types of players that suit this tactics. Pardon me if I don't think a bunch of geeks are going to unlock the secret to recruitment. They lost their defenders and their models helped them get a bunch of defenders that won't cut it.

 

[Cheimoon]

Who has it helped in recruitment.

Brentford would be the main example here. From your comments, you don't seem particularly open to be convinced otherwise, so I can't really be bothered to spend much time on this; but see the following article and thread for more info:

https://www.redcafe.net/threads/ivan-toney-brentford.461156/
https://talksport.com/football/fa-c...-smartest-club-championship-leicester-fa-cup/

How are these stats measured? Are people literally watching and doing data entry in some system? Oh look this player had a completed pass, add 1 to his passing stats.

Always wondered this myself. Definitely used to be like this at first with the simpler stats like passes completed/missed where it's easy to tick, but when you're now talking about every little action/type of shot/positioning of defenders/shot placement and power for a variety of stats... It'd be impossible to do without programming.

Statsbomb recently bought an Egyptian company called Arqam that had been doing the raw data for them. People watch each match and count a billion different events. Where was the shot taken from ? How many defenders between the shot taker and the goal ? Where is the keeper stood ? What happened just before the shot ? What height was the ball off the ground when the shot or header happened ? Each year Statsbomb add more detail. Pressures is a big thing for them. It must take hours to do each match.

This is what their website says:

StatsBomb Data
Our unique event data collection spec has over 3,400 events per match of on and off the ball data including pressures, ball carries, possession chains and more. Data generated from a blend of Computer Vision and human driven collection with automated validation checks and a highly experienced quality assurance team, makes it the most accurate event data in the industry

Who records this sort of data?

Can't be people watching TV coverage as replays are often shown when the game is ongoing so passes, tackles, etc would be missed. Do they have a different broadcast feed or are the numbers just wrong?

I know these are old posts, but I only just noticed this thread. Just to add that, some months ago (or more), my Dutch newspaper ran a long piece on how these stats are compiled. They went to one of the UK companies, and their office really was full of people that re-watch games and code everything that's happening into their database, following a clear set of rules on how to assess events. So the recording of events really is very manual; no programming involved - or at least not in that story.

 

[Liver_bird]

Well Liverpool very publicly and recently put a lot into advanced analytics to fuel a lot of their decision making, led to some pretty decent results over the past few years for them... Knutson often references his time with Midtylland and they all of a sudden turned into a top danish side. It's pretty much everywhere now, but the early adopters definitely benefitted hugely and led to tangible success. It's nothing surprising though, you would expect advanced stats to always help out with making better recruitment decisions than if you didn't have this data. Happened a lot in every other sport and it's happened with football too. Getting the early edge is where the biggest benefits lie though, because eventually everyone else will catch on (and if not, they'll really fall behind as we've seen often).

Anyway, there's loads of data out there that I can't be arsed to pull up. Also not like I have the specific information. But using advanced data can only help with recruitment.

It’s a really good point, because analytics across the board weren’t widely embraced even 5-6 years ago, so we were able to take advantage of that to gain an edge on the market. When you’re competing with clubs with better finances than you it’s essential to have.

However, these days such a gap and advantage doesn’t exist because these stats are widely available and used by all clubs. Players that you could get for a relative steal analytics wise you can’t now because everyone is aware of them. Of course one thing worth mentioning is these stats at least the public ones are far more advanced for attackers than they are for defenders. Defenders are largely dependant on the system and it’s harder to analyse how one will transfer over another. That’s why for instance we didn’t move to the next level until we invested in VVD and Alisson for huge sums.

One of my main concerns going forward is where the next edge so to speak can be gleaned from, any club that isn’t city or PSG needs to get their recruitment spot on most of the time to be able to build sides that can compete with their resources. It’s not to say we and other clubs don’t have money there’s just a disparity that has to be combated in other ways.

 

[Liver_bird]

There's nothing all that impressive about liverpool's recruitment, their coaching is where it's at. Football is a very complex sport with alot of variations. Alot of different types of tactics and types of players that suit this tactics. Pardon me if I don't think a bunch of geeks are going to unlock the secret to recruitment. They lost their defenders and their models helped them get a bunch of defenders that won't cut it.

The January window was a farce anyways and the defenders we picked up are the ones that the owners were willing to buy or loan. So I don’t think that’s a completely fair assertion. I do agree that you can’t use analytics alone, but in combination with things it’s a useful tool. Having a manager with a clear style and a recruitment department who understands how to recruit for that style can pay dividends. They’re not really unlocking secrets but these are highly paid people at the top of their profession who’ve been employed by the clubs to find an edge. It’s just obviously not as cut amd dry as that, all the moving parts have to be working in conjunction with one another. It’s also just far easier for owners to recruit the best in the world in other departments rather than paying and losing exorbitant sums on transfer duds.
But you are right, it won’t solve anything alone, the committee and Rodgers is a prime example of how it shouldn’t work.

I do believe having said all that however the most important factor is usually money, the more you can pay your players generally the better players you have. It’s why the top four barring the odd exception is always the clubs with the highest wage bills.