Welcome aboard! This is an exciting addition and I'm really looking forward to seeing your contributions.
I have a question, that is admittedly a pet peeve of mine: why do people - not saying you, or anyone here, do this, just the general soccer community - place so much value on single-game xG? There's so much noise in that data in a 90 minute sample that it's essentially useless, and yet broadcasts and analysts report xG at half time and full time as if those numbers are significant. Can you talk a little about why that is, and how long it takes for xG data to actually have meaning?
It feels like it takes about a half season or more for xG to stabilize, but I can't quite explain why with any sort of eloquence.
I share the same sentiment. One of the things that bugs me the most is when the Apple TV guys share single game statistics. While they're helpful, statistics over a single game don't convey the full story of a players season. Just because Evander has a game where he doesn't score, doesn't mean his xG is permanently tanked and will destroy his season.
I'm not a data scientist or data analyst by any means, I've got more of a technical background and have been on a journey of educating myself over the last year or so, and still have a TON to learn, so I won't know the answer for everything.
I would say though that it takes a multi game span for any sort of data to be at a point where it can actually be interpreted, so like half a season is a good metric imo
This is a GREAT addition to the Stumptown Footy offerings. But please, do this for the Thorns as well as the Timbers. As an aside, you are a good writer.
Thanks Lee! I'd love to potentially write about Thorns stuff in the future. I know some of the Thorns writers touch into video and data, and they do a top job. If I can fire out a Thorns piece from time to time, I will!
Thanks, and I totally get your sentiment! I wasn't a super big data guy up until not too long ago. This stuff gets easier to understand as time goes along, and is one of the reasons why I want to write this series in the first place. Hopefully things get easier to understand, and if you ever want clarification on anything, please don't hesitate to reach out either here or on my twitter @LuukasOjala
I don't tweet but appreciate your invitation to connect here. Maybe your metrics can help me understand why I think Pantemis is a better fit for Portland than Maxi is 8>)).
Any strong opinions on the predictive value differential between xG and “post-shot xG”? Intuitively it seems like there should be, but I’ve never seen anything in print…
xG is considered to be a pre-shot model(applicable to attacking players), while post shot xG is considered to be more of an applicable metric for goalkeepers and defensive players. xG is considered up until the moment the shot is taken, while PSxG is after the shot is taken, and is only measured with shots on target.
I know the definitions. My question is more about your personal take on the relative value re: how the difference is useful in assessing attacking pressure.
I tend to see analyses that directly equate xG with “did well applying pressure on goal” but wonder how much the PS-xG might be a better tool. It’s nice to get shots from dangerous areas…but seems more important to get GOOD shots.
Ahh gotcha! sorry for misinterpeting :) I think that the two metrics go sort of hand in hand. I think personally the xG plus PSxG minus goals conceded per 90 is the best way in measuring attacking pressure. Plus I think if you take a look into how ineffective the defending teams PPDA (meaning the attacking team is pressing hard) I think those would all combine into a good interpretation of effectiveness of attacking play.
Also, I am in no way a data scientist or analyst. I come from a more technical, video analysis based background, and I am trying to learn more regarding the data side of the game. The club enviroment I work in currently has no access to any data tools so I'm totally self taught, and continuously learning
Sorry to hear about the resource-poor environment. Soccer is, generally, slow in accepting the idea that you can learn from stats and analysis rather than some sort of mystic soccer nous.
Yeah I agree! I'm working on an article for my personal substack where I talk about the divide between clubs who are data forward, and who arent. The divide is only getting larger, and soon itll be so distinguishable that clubs will have to invest in data. (cough cough portland cough cough )
Having lived through the Baseball Nerd Wars of the early 2000's, it's infuriating to see soccer going through the exact same thing. The same arguments, the same battle lines being drawn - it's so dumb!
The way it was sometimes explained to resistant baseball fans makes a lot of sense, and it has stuck with me ever since. You can plug in your own favorites, but essentially it's "You like pizza, right? You like beer, right? You get that you don't have to choose? You can have them both at the same meal?"
There will always be a need to watch soccer games closely! That doesn't go away, even if you're a hard-core believer in the value of statistical analysis. It's just that, after the game is over, it's also useful to turn to the stats to inform a deeper understanding of what it is you just saw, should you choose to.
The soccer analysis community often talks about how you need to combine various types of data in order to truly understand how a game went, and unfortunately most people just view statistics, and interpret the game purely through the numbers. Unfortunately though these other data types like tracking and event data are often overlooked for surface level statistics
And in terms of defenses, yes; the best analytical tool I’ve seen to date for assessing pure goalkeeper shot-stopping/positioning is PS-xG minus goals allowed per 90 minutes. Comes as close as possible to filtering out the “conceded tons because the backline is shit” elements…c
This was such a good essay, and I never new soccer could go into such detail. I don’t think I can ever look at the sport the same way again!
Welcome aboard! This is an exciting addition and I'm really looking forward to seeing your contributions.
I have a question, that is admittedly a pet peeve of mine: why do people - not saying you, or anyone here, do this, just the general soccer community - place so much value on single-game xG? There's so much noise in that data in a 90 minute sample that it's essentially useless, and yet broadcasts and analysts report xG at half time and full time as if those numbers are significant. Can you talk a little about why that is, and how long it takes for xG data to actually have meaning?
It feels like it takes about a half season or more for xG to stabilize, but I can't quite explain why with any sort of eloquence.
Thanks! Happy to be here.
I share the same sentiment. One of the things that bugs me the most is when the Apple TV guys share single game statistics. While they're helpful, statistics over a single game don't convey the full story of a players season. Just because Evander has a game where he doesn't score, doesn't mean his xG is permanently tanked and will destroy his season.
I'm not a data scientist or data analyst by any means, I've got more of a technical background and have been on a journey of educating myself over the last year or so, and still have a TON to learn, so I won't know the answer for everything.
I would say though that it takes a multi game span for any sort of data to be at a point where it can actually be interpreted, so like half a season is a good metric imo
Nice to see you pursuing your dreams, we all miss you at the Park and hope to see you when you are in town at future Timbers or Thorns matches.
MITCH!!!!! Hi! I miss you all, and I hope to see you soon!
This is a GREAT addition to the Stumptown Footy offerings. But please, do this for the Thorns as well as the Timbers. As an aside, you are a good writer.
Thanks Lee! I'd love to potentially write about Thorns stuff in the future. I know some of the Thorns writers touch into video and data, and they do a top job. If I can fire out a Thorns piece from time to time, I will!
Welcome Luukas, and thank you for bringing this kind of content to STF! Looking forward to reading more of your pieces.
Thanks, Stefan! I look forward to writing more!
Welcome aboard and thanks for sharing this info which, BTW, if way over my head. Looking forward to your future contributions!
Thanks, and I totally get your sentiment! I wasn't a super big data guy up until not too long ago. This stuff gets easier to understand as time goes along, and is one of the reasons why I want to write this series in the first place. Hopefully things get easier to understand, and if you ever want clarification on anything, please don't hesitate to reach out either here or on my twitter @LuukasOjala
I don't tweet but appreciate your invitation to connect here. Maybe your metrics can help me understand why I think Pantemis is a better fit for Portland than Maxi is 8>)).
Any strong opinions on the predictive value differential between xG and “post-shot xG”? Intuitively it seems like there should be, but I’ve never seen anything in print…
xG is considered to be a pre-shot model(applicable to attacking players), while post shot xG is considered to be more of an applicable metric for goalkeepers and defensive players. xG is considered up until the moment the shot is taken, while PSxG is after the shot is taken, and is only measured with shots on target.
I hope my explanation helps!
I know the definitions. My question is more about your personal take on the relative value re: how the difference is useful in assessing attacking pressure.
I tend to see analyses that directly equate xG with “did well applying pressure on goal” but wonder how much the PS-xG might be a better tool. It’s nice to get shots from dangerous areas…but seems more important to get GOOD shots.
Thoughts?
Ahh gotcha! sorry for misinterpeting :) I think that the two metrics go sort of hand in hand. I think personally the xG plus PSxG minus goals conceded per 90 is the best way in measuring attacking pressure. Plus I think if you take a look into how ineffective the defending teams PPDA (meaning the attacking team is pressing hard) I think those would all combine into a good interpretation of effectiveness of attacking play.
Also, I am in no way a data scientist or analyst. I come from a more technical, video analysis based background, and I am trying to learn more regarding the data side of the game. The club enviroment I work in currently has no access to any data tools so I'm totally self taught, and continuously learning
Sorry to hear about the resource-poor environment. Soccer is, generally, slow in accepting the idea that you can learn from stats and analysis rather than some sort of mystic soccer nous.
Yeah I agree! I'm working on an article for my personal substack where I talk about the divide between clubs who are data forward, and who arent. The divide is only getting larger, and soon itll be so distinguishable that clubs will have to invest in data. (cough cough portland cough cough )
Having lived through the Baseball Nerd Wars of the early 2000's, it's infuriating to see soccer going through the exact same thing. The same arguments, the same battle lines being drawn - it's so dumb!
The way it was sometimes explained to resistant baseball fans makes a lot of sense, and it has stuck with me ever since. You can plug in your own favorites, but essentially it's "You like pizza, right? You like beer, right? You get that you don't have to choose? You can have them both at the same meal?"
There will always be a need to watch soccer games closely! That doesn't go away, even if you're a hard-core believer in the value of statistical analysis. It's just that, after the game is over, it's also useful to turn to the stats to inform a deeper understanding of what it is you just saw, should you choose to.
The soccer analysis community often talks about how you need to combine various types of data in order to truly understand how a game went, and unfortunately most people just view statistics, and interpret the game purely through the numbers. Unfortunately though these other data types like tracking and event data are often overlooked for surface level statistics
And in terms of defenses, yes; the best analytical tool I’ve seen to date for assessing pure goalkeeper shot-stopping/positioning is PS-xG minus goals allowed per 90 minutes. Comes as close as possible to filtering out the “conceded tons because the backline is shit” elements…c
I'm excited to see what's next! By the end of this series I fully expect Luukas to work with Tony Bloom as a data analyst for BHAFC.
Thats very kind of you to say! I'm not Tony Bloom smart though haha 😂
Looking forward to your next piece!
Thank you! I hope you enjoy it, currently writing it right now, and its a ton of fun