|
Post by datguy on Dec 12, 2019 15:29:07 GMT
Gregory is shite at shooting.
If your stat doesn’t show that then sorry, it’s a shit stat.
|
|
|
Post by harlequin on Dec 12, 2019 15:33:39 GMT
Because a person gives a chance a rating based on a range of things. Different people have different ways of measuring xg. It's not this bee all and end all stat, it's subjective and people don't treat it as such. No-one has ever claimed it is the be all and end all. But it is demonstrably a better indicator of future perfromance than actual goals and isn't prone to the numerous congitive biases that go into purely opinion based assessment (now that is subjective). It is the mainstay of every analytically run football (just for Bojansstalker :-)) from Liverpool downwards and appeared to project three and a half months ago that table topping Swansea were just a bog standard mid table team & Stoke would be on the verge of climbing out of the bottom three by this coming weekend. I prefer it to "Gregory is terrible at finishing", but each to his own :-) I don't think that's always the case, at least using the same players. Expected goals uses an average benchmark for finishing, if the strikers are significantly below that benchmark and missing the big chances like Gregory, we'll always be below expected goals.
|
|
|
Post by bayernoatcake on Dec 12, 2019 15:36:41 GMT
Because a person gives a chance a rating based on a range of things. Different people have different ways of measuring xg. It's not this bee all and end all stat, it's subjective and people don't treat it as such. No-one has ever claimed it is the be all and end all. But it is demonstrably a better indicator of future perfromance than actual goals and isn't prone to the numerous congitive biases that go into purely opinion based assessment (now that is subjective). It is the mainstay of every analytically run football (just for Bojansstalker :-)) from Liverpool downwards and appeared to project three and a half months ago that table topping Swansea were just a bog standard mid table team & Stoke would be on the verge of climbing out of the bottom three by this coming weekend. I prefer it to "Gregory is terrible at finishing", but each to his own :-) People refer to is as fact. It isn't. We haven't been unlucky this season, anyone who has watched can see that. The figures don't tally with what has actually gone on.
|
|
|
Post by tachyon on Dec 12, 2019 15:44:24 GMT
No-one has ever claimed it is the be all and end all. But it is demonstrably a better indicator of future perfromance than actual goals and isn't prone to the numerous congitive biases that go into purely opinion based assessment (now that is subjective). It is the mainstay of every analytically run football (just for Bojansstalker :-)) from Liverpool downwards and appeared to project three and a half months ago that table topping Swansea were just a bog standard mid table team & Stoke would be on the verge of climbing out of the bottom three by this coming weekend. I prefer it to "Gregory is terrible at finishing", but each to his own :-) I don't think that's always the case, at least using the same players. Expected goals uses an average benchmark for finishing, if the strikers are significantly below that benchmark and missing the big chances like Gregory we'll always be below expected goals. The true talent range for "finishing ability" is very narrow and it is very difficult to rule out random variation as a major cause for under/over performance in the sample sizes that are available, particularly with confounding factors such as peak age improvement/decline and changing player environment. Over-fitting recency bias of "Gregory is a rubbish finisher" & "(the young) Rashford is clinical" to a model is a much bigger danger to faulty player evaluation.
|
|
|
Post by tachyon on Dec 12, 2019 15:55:41 GMT
No-one has ever claimed it is the be all and end all. But it is demonstrably a better indicator of future perfromance than actual goals and isn't prone to the numerous congitive biases that go into purely opinion based assessment (now that is subjective). It is the mainstay of every analytically run football (just for Bojansstalker :-)) from Liverpool downwards and appeared to project three and a half months ago that table topping Swansea were just a bog standard mid table team & Stoke would be on the verge of climbing out of the bottom three by this coming weekend. I prefer it to "Gregory is terrible at finishing", but each to his own :-) People refer to is as fact. It isn't. We haven't been unlucky this season, anyone who has watched can see that. The figures don't tally with what has actually gone on. I never refer to it as fact, you'll have to take that up with those who do. Totally agree, "The figures don't tally with what has actually gone on" we've created 28 xG and allowed 26 xG. That's a mundane mid-table process, not a bottom three one. Maybe not what a side with our resources should aspire to but sometimes your probabilistic process flips 2 heads out of ten purely by random variance.
|
|
|
Post by StoKeith on Dec 12, 2019 16:07:00 GMT
|
|
|
Post by sheikhmomo on Dec 12, 2019 16:18:21 GMT
Because a person gives a chance a rating based on a range of things. Different people have different ways of measuring xg. It's not this bee all and end all stat, it's subjective and people don't treat it as such. I prefer it to "Gregory is terrible at finishing", but each to his own :-) Isn't that its weakness though? It basically assumes that Gregory should finish a goal like Mitrovic should because they're in the same section!
|
|
|
Post by tachyon on Dec 12, 2019 16:31:55 GMT
I sit somewhere between Bayern and tachyon wrt xG. I think it could be improved. Firstly, I don’t get some of the percentages it come up with, but I’m looking at it subjectively and the calculation is purely objective as far as I understand it. A bigger problem is the tendency to simply add the chances together. Currently (as I understand it), if a player misses a penalty (80%) and the rebound is spooned over by another player (say 50%), then the team gets 1.3 xG for that phase of play. Clearly makes no sense. Perhaps xG could provide individual chance data, but “sum” them more cleverly to acknowledge that an individual attack could never generate more than one goal. I wrote a little script on Python last year (just for coding practice) that generates a few graphs of the season. You can choose how many matches you want to average over (I’ve done 5 in the ones below). I included xG, just because it gave me more data to play with. This is our season so far (for what it’s worth). The “goals minus xG” essentially shows you how well or badly we’re doing vs what’s “expected”. At the beginning of the season, we were conceding a goal more than was expected every game as Butland was having his nightmare period. We also went through a phase of scoring fewer goals than would be expected. Both of these have settled down a bit now. Nice graphs, Related chances have been addressed. The under the hood simulations don't allow you to score more than once in simulations from related chances. Aggregated totals trade simplicity for granular insight. We think it's the better of two compromised choices. There's lots of issues that have been addressed, but they usually stay in house. xG timelines (when you create chances) and current score-line and time remaining impact on xG, also was it better to create lots of small chances or fewer better ones. Most issues just tweak, rather than fundamentally alter the conclusions. We tend to run 10 game rolling averages spread over multiple seasons to look at longer term trends. Single game xG is more a descriptive than predictive metric. For overall team ratings we take the last 40 games, weighted towards the most recent xG outcomes.
|
|
|
Post by StoKeith on Dec 12, 2019 17:03:16 GMT
Nice graphs, Related chances have been addressed. The under the hood simulations don't allow you to score more than once in simulations from related chances. Aggregated totals trade simplicity for granular insight. We think it's the better of two compromised choices. There's lots of issues that have been addressed, but they usually stay in house. xG timelines (when you create chances) and current score-line and time remaining impact on xG, also was it better to create lots of small chances or fewer better ones. Most issues just tweak, rather than fundamentally alter the conclusions. We tend to run 10 game rolling averages spread over multiple seasons to look at longer term trends. Single game xG is more a descriptive than predictive metric. For overall team ratings we take the last 40 games, weighted towards the most recent xG outcomes. That’s good to know. I guess if you can look back at your predictions and see they are generally correct, then it’s an excellent system!
|
|
|
Post by tachyon on Dec 12, 2019 17:04:49 GMT
I prefer it to "Gregory is terrible at finishing", but each to his own :-) Isn't that its weakness though? It basically assumes that Gregory should finish a goal like Mitrovic should because they're in the same section! Conversion rates are both volatile and the skill gap between players is narrow. Lewandowski is over performing his xG by 11% this season. Last season he under performed it by around 20%. Ronaldo under performed his xG last season and Mitrovic also did last season. If you take the biggest over performers, "clinical finishers" if you like, as a group they always regress to being neither under or over performers in future trials. Unless your name is Messi and it is certainly counter intuitive, but finishing IS a skill, of course, but the talent differential between players in the same leagues is invariably too small to measure. The observed difference is almost always random variation that eventually regresses to the expected xG rate.
|
|
|
Post by GoBoks on Dec 12, 2019 18:29:41 GMT
The biggest value expected goals or xG brings to the table is in analysing longer term trends. If you're continually creating more chances than the opposition, even if you're not always converting those chances, it's very likely that over time your outcomes (results) will adhere more closely to your process (xG differential). After six games Stoke had one whole point from a possible 18 (results dire), but a fairly equal share of the xG (process not great, but not terrible) in their matches to go with a similar trend from the previous season. At that season low in September, Infogol simulated the remainder of the season using the xG ratings of all 24 teams to see when it would be more likely than not that Stoke would claw their way out of the bottom three and posted it on Twitter. View AttachmentWe estimated Stoke would start to be slightly more likely than not to escape from the drop zone by game week 22. That's next week and with favourable results we could hit the scheduled recovery at the earliest predicted time.....or we could take until week 30 or we might not manage it at all. That's what projections are, they come with an uncertainty, grouped around the most likely outcome. XG from the start of September has done a pretty good job of projecting where Stoke would be half way through December and currently it thinks our median final points total will be 54, our final median finishing position will be 18th and we currently have a 21% chance of relegation. Obviously you are a supporter of the process, so why title the thread "nonsense"? Wouldn't "Nostradamus step aside, Expected Goals is here" be more catchy?
|
|
|
Post by tachyon on Dec 12, 2019 18:44:45 GMT
The biggest value expected goals or xG brings to the table is in analysing longer term trends. If you're continually creating more chances than the opposition, even if you're not always converting those chances, it's very likely that over time your outcomes (results) will adhere more closely to your process (xG differential). After six games Stoke had one whole point from a possible 18 (results dire), but a fairly equal share of the xG (process not great, but not terrible) in their matches to go with a similar trend from the previous season. At that season low in September, Infogol simulated the remainder of the season using the xG ratings of all 24 teams to see when it would be more likely than not that Stoke would claw their way out of the bottom three and posted it on Twitter. View AttachmentWe estimated Stoke would start to be slightly more likely than not to escape from the drop zone by game week 22. That's next week and with favourable results we could hit the scheduled recovery at the earliest predicted time.....or we could take until week 30 or we might not manage it at all. That's what projections are, they come with an uncertainty, grouped around the most likely outcome. XG from the start of September has done a pretty good job of projecting where Stoke would be half way through December and currently it thinks our median final points total will be 54, our final median finishing position will be 18th and we currently have a 21% chance of relegation. Obviously you are a supporter of the process, so why title the thread "nonsense"? Wouldn't "Nostradamus step aside, Expected Goals is here" be more catchy? It's been portrayed as nonsense by the majority on this forum, with some admirable exceptions. Hence the added "" ""
|
|
|
Post by rawli on Dec 13, 2019 11:21:39 GMT
It's not though, is it? Example - Under Jones, the ball comes in to a single striker who is not reknowned for his goalscoring - misses - Under MON, the ball comes in to a Premier League player and he scores, or it is picked up by the other 3 players in the box who weren't there previously Yes I get all that. But it's not without substance. Sure I saw somewhere it is on average a better indicator of how you will do for example in the second half of a season than your points tally in the first half. Doesn't changing manager and system effectively negate it for us ?
|
|
|
Post by rawli on Dec 13, 2019 11:23:40 GMT
I know why EFL Stats present xG as a table, it’s to breed familiarity & make it accessible. But it simply replaces one unique set of outcomes (the actual results) with another unique set based on xG. xDraws are assigned if the XG differential lies between a narrow xG band & xWins or xLosses are decided when the match expected goals differential lies outside these bounds. Average points and average final positions also trade brevity for loss of insight. The aim of xG is to illustrate the uncertainty & randomness that exists is a low scoring sport, such as soccer. You can do that by simulation every game played so far using each individual xG chance that was created, repeating 10,000 times and plotting the range of current league positions played out in those simualtions. For Stoke it is the most wide ranging in the division. We’re as likely to be 2nd as we are to be 19th, but from a performance viewpoint, there are 16 teams who are more likely to be in the bottom 3 than us, given the chance creation performance of all 24 teams to week 21. Overall, we’re currently most likely to be in the “not going up, not going down” group of 15 What the fucking hell is soccer? Football.
|
|
|
Post by Gods on Dec 13, 2019 15:05:12 GMT
Yes I get all that. But it's not without substance. Sure I saw somewhere it is on average a better indicator of how you will do for example in the second half of a season than your points tally in the first half. Doesn't changing manager and system effectively negate it for us ? Perhaps, it certainly messes with it I'm sure. I still think it's something, not everything, but worth keeping an eye on, like you might have the sat nav on but still keep an eye on the road signs!
|
|
|
Post by georgieboy52 on Dec 13, 2019 15:07:17 GMT
I'm an 'expected goals' believer. It's as close as you'll get to a scientific explanation of 'unlucky' and we all know luck changes. How many goals did YOU score ? When you played football like.
|
|
|
Post by Gods on Dec 13, 2019 15:10:54 GMT
I'm an 'expected goals' believer. It's as close as you'll get to a scientific explanation of 'unlucky' and we all know luck changes. How many goals did YOU score ? When you played football like. Not many mate, I seemed to be cursed with bad luck!
|
|
|
Post by georgieboy52 on Dec 13, 2019 15:14:19 GMT
How many goals did YOU score ? When you played football like. Not many mate, I seemed to be cursed with bad luck! I had bloody decent right foot-could do anything with it. I could never deliver a left-footed cross though. No matter how I tried.
|
|
|
Post by CBUFAWKIPWH on Dec 13, 2019 16:56:59 GMT
That's why this stat will eventually be driven out of football and looked back it with hilarity. Even the heaviest dose of weapons grade acid couldn't convince anyone that we should be 6th this season. It's idiotic. What's idiotic is your apparent lack of understanding of statistics. It isn't an exact science - it's an approximation. For the majority of clubs their actual and EG position will be pretty close - but the nature of any statistical measure is the potential for "outliers" - oddball scenarios that seem to fly in the face of the "expected" outcome. "Outliers" do not disprove the value of the statistical measure - they are an expected feature. An outlier is evidence of shit happening - it doesn't mean it's all shit. Looking at that table Stoke (bad shit) and Bristol (good shit) happen to be this years "outliers".
|
|
|
Post by sheikhmomo on Dec 13, 2019 17:27:05 GMT
That's why this stat will eventually be driven out of football and looked back it with hilarity. Even the heaviest dose of weapons grade acid couldn't convince anyone that we should be 6th this season. It's idiotic. What's idiotic is your apparent lack of understanding of statistics. It isn't an exact science - it's an approximation. For the majority of clubs their actual and EG position will be pretty close - but the nature of any statistical measure is the potential for "outliers" - oddball scenarios that seem to fly in the face of the "expected" outcome. "Outliers" do not disprove the value of the statistical measure - they are an expected feature. An outlier is evidence of shit happening - it doesn't mean it's all shit. Looking at that table Stoke (bad shit) and Bristol (good shit) happen to be this years "outliers". So the clubs they have most wrong prove it's a good system?! I work in gambling and firms are becoming more and more dubious about it. The basic premise that all players are equal is just wrong.
|
|
|
Post by thepremierbanksy on Dec 13, 2019 17:39:49 GMT
Isn't there a simple method of factoring in who a chance has fallen to and using it as a multiplier of the xG? Ie if a clean through is 0.5 xG then multiply by maybe 1.1 if it falls to Messi, 0.95 if it falls to Mitrovic and 0.05* if it falls to Kyle Lightbourne? Or is your arguement about the difference in finishing ability saying that actually makes the model less accurate?
*obvious joke
|
|
|
Post by StoKeith on Dec 13, 2019 17:55:41 GMT
Isn't there a simple method of factoring in who a chance has fallen to and using it as a multiplier of the xG? Ie if a clean through is 0.5 xG then multiply by maybe 1.1 if it falls to Messi, 0.95 if it falls to Mitrovic and 0.05* if it falls to Kyle Lightbourne? Or is your arguement about the difference in finishing ability saying that actually makes the model less accurate? *obvious joke I reckon they might already do that to a certain extent (perhaps averaging within one league), but they need thousands of shots in different locations to be able to build up statistical significance. I don’t think you would get enough data from one player to be able to generate the required statistical significance.
|
|
|
Post by rawli on Dec 13, 2019 17:58:48 GMT
Not many mate, I seemed to be cursed with bad luck! I had bloody decent right foot-could do anything with it. I could never deliver a left-footed cross though. No matter how I tried. Between you and Incey we'd have a decent winger.
|
|
|
Post by Little Gary Patel on Dec 13, 2019 18:01:39 GMT
Bayern showing himself up as usual
|
|
|
Post by Deleted on Dec 13, 2019 19:21:48 GMT
Can I see any stats on how many Danny Batth headers go to a team-mate, rather than spazzed into touch, or straight up in the air?
|
|
|
Post by tachyon on Dec 13, 2019 19:25:12 GMT
What's idiotic is your apparent lack of understanding of statistics. It isn't an exact science - it's an approximation. For the majority of clubs their actual and EG position will be pretty close - but the nature of any statistical measure is the potential for "outliers" - oddball scenarios that seem to fly in the face of the "expected" outcome. "Outliers" do not disprove the value of the statistical measure - they are an expected feature. An outlier is evidence of shit happening - it doesn't mean it's all shit. Looking at that table Stoke (bad shit) and Bristol (good shit) happen to be this years "outliers". So the clubs they have most wrong prove it's a good system?! I work in gambling and firms are becoming more and more dubious about it. The basic premise that all players are equal is just wrong. The basic premise isn't that all players are equal. They differ in their ability to get on to the end of chances and teams differ in their ability to create those chances. It is the conversion rates of those chances that is a narrowly banded skill and under or over performance at the sharp end of a scoring opportunity invariably regresses to the population mean (except Messi). It is used in the betting industry, for instance Matthew Benham's Smartodds uses xG extensively.
|
|
|
Post by sheikhmomo on Dec 13, 2019 19:28:14 GMT
So the clubs they have most wrong prove it's a good system?! I work in gambling and firms are becoming more and more dubious about it. The basic premise that all players are equal is just wrong. The basic premise isn't that all players are equal. They differ in their ability to get on to the end of chances and teams differ in their ability to create those chances. It is the conversion rates of those chances that is a narrowly banded skill and under or over performance at the sharp end of a scoring opportunity invariably regresses to the population mean (except Messi). It is used in the betting industry, for instance Matthew Benham's Smartodds uses xG extensively. I've met Benham many times. He's a genius. I know he relies in his own models much more than any external ones.
|
|
|
Post by tachyon on Dec 13, 2019 19:34:03 GMT
Can I see any stats on how many Danny Batth headers go to a team-mate, rather than spazzed into touch, or straight up in the air? You can. But then we'd have to start talking about non-shot expected goals............
|
|
|
Post by tachyon on Dec 13, 2019 19:38:15 GMT
The basic premise isn't that all players are equal. They differ in their ability to get on to the end of chances and teams differ in their ability to create those chances. It is the conversion rates of those chances that is a narrowly banded skill and under or over performance at the sharp end of a scoring opportunity invariably regresses to the population mean (except Messi). It is used in the betting industry, for instance Matthew Benham's Smartodds uses xG extensively. I've met Benham many times. He's a genius. I know he relies in his own models much more than any external ones. Sure. He collects his own data in house & uses it at Brentford. But the basic premise of xG is the same. IIRC Tippet's recent book explicitly states that there's no skill differential in conversion rates.
|
|
|
Post by sheikhmomo on Dec 13, 2019 19:43:51 GMT
I've met Benham many times. He's a genius. I know he relies in his own models much more than any external ones. Sure. He collects his own data in house & uses it at Brentford. But the basic premise of xG is the same. IIRC Tippet's recent book explicitly states that there's no skill differential in conversion rates. I do see that as the major flaw though mate. I think the differentials are poorly calibrated. I genuinely know that traders at other firms are rowing away a little bit but that's not saying that the system itself can't evolve. It comes up with too many absurdities in my eyes.
|
|