Expected Position, Goals & Goals Against

[HolmesfieldBlade]

We’ve conceded less than
City, Leeds, Liverpool, Leicester, Wolves, Man Utd & more.

Equal to Chelsea and Southampton.

We’ve conceded 6
2 more than expected. (4.2).

We’ve scored 1
3 less than expected. (3.9).

We sit 19th
Expected we should be 13th.

Using the expected goals model it does not look great for Palace, Fulham and West Brom. Burnley have had a similar start to The Blades, being unlucky in their games.

And how good does that 0-1 with 10-men at Villa park look right now?

These things always even themselves out over the course of the season.

As Didzy said, we’re not pushing any panic buttons, yet. There’s still 3 worse teams in this league than Sheffield United.

Get behind your club, Chris and the squad are trying their very best in ‘unprecedented times’ as Matt Hancock would say.

 

[BladesOnToast]

We have conceded the seventh fewest goals in the Premier League, the fewest in the bottom half of the divison.

Can't quite remember off the top of my head where we are in the Goals Scored rankings...

 

[tincan]

"And how good does that 0-1 with 10-men at Villa park look right now?"
How many would the scousers have conceded if they had VVD, sent off? Always a blades lining...

 

[Latters]

The “Expected Goals” stat is a bag of bollocks though isn’t it. You either have the players to convert chances to goals or you don’t. Hopefully now we have Brewster, we’ll have somebody who takes the majority of chances we create for him, as well as bagging some “unexpected goals”, if there’s such a thing.

 

[Revolution]

These things always even themselves out over the course of the season.

No they don't. You can score less than your expected goals, or concede more. Look at Didzy last season.

A team that is conceding more than expected, or scoring less than expected, might just be a bad team.

 

[PokerBlade]

The “Expected Goals” stat is a bag of bollocks though isn’t it. You either have the players to convert chances to goals or you don’t. Hopefully now we have Brewster, we’ll have somebody who takes the majority of chances we create for him, as well as bagging some “unexpected goals”, if there’s such a thing.

It's far from bollocks. It's a useful statistic. If the stat were bollocks you'd expect to see lots of players outperforming it by having superior finishing, and lots of player underperforming it by a distance by being weak finishers. The reality is that players like McGoldrick coming massively under expectation are the anomalies and I'm not sure there's anyone who's consistently outperformed their xG. What xG seems to show is that scoring goals is much more about players creating chances and getting shots away (i.e. the ability to beat a man, make a run, find a pocket of space and so on) than it is about being a good finisher.

It's far from the be all and end all of football but it is a useful indication of how well you're likely to do in future games.

 

[FallowfieldBlade]

Think last year the XG we over performed if I remember right this year were underperforming so far


Bad start but nowt to worry about just yet if Brewster hits the ground running we'll be fine

 

[raul]

It's only relevant when the league is decided by expected points.

 

[Kev Gee]

The only stats that matter are goals scored and points.
We currently have 1 on stat 1 and 0 on stat 2.
Hopefully that will change soon!

 

[BlairBlade]

No they don't. You can score less than your expected goals, or concede more. Look at Didzy last season.

A team that is conceding more than expected, or scoring less than expected, might just be a bad team.

I don't believe that things even themselves out over a season.

No idea where this principle came from and I've never seen it happen.

If it did, those cheating bastards at Villa would have been relegated last season.

 

[MalvernBlade]

A

The “Expected Goals” stat is a bag of bollocks though isn’t it. You either have the players to convert chances to goals or you don’t. Hopefully now we have Brewster, we’ll have somebody who takes the majority of chances we create for him, as well as bagging some “unexpected goals”, if there’s such a thing.

Any McGoldrick goal is classified as unexpected isn't it??

 

[PokerBlade]

It's only relevant when the league is decided by expected points.

xG and stats like it are used to analyse the standard of the performances. Same way you can look at tackles made, pass completion, distances covered. Unless you think there's no point in analysing games then it's daft to dismiss it.

 

[PokerBlade]

I don't believe that things even themselves out over a season.

No idea where this principle came from and I've never seen it happen.

If it did, those cheating bastards at Villa would have been relegated last season.

I might butcher this explanation but the idea that past events will be evened out by future events is the gambler's fallacy e.g. I flipped four heads in a row so in the future I'll flip some extra tails to even things out. But what stats do tell you is what you should expect in the future e.g. if I've had an expected goal stat of one per game then I should expect to score one per game in the future and the fact I scored 0 or 2 goals in the last game is misleading when predicting future performances. So if we played last season's Villa games several times under the same conditions then we could be confident of being on the right end of the results.

For anyone who might be interested in playing with graphs try this poker ev simulator:

Cash Game Variance Calculator

Poker Variance Calculator for cash games. Analyze Risk of Ruin, Confidence Intervals and other data to protect your poker bankroll.

www.primedope.com

Hit calculate and then increase the line for hand number and see what happens. The lines between expected and actual get closer as the sample size increases (this is a function of the scale of the graph as a single loss starts to represent less and less). What you'll also see is that over a small sample a positive expectation has a reasonable chance of losing overall, but the higher the sample size the more and more likely it becomes that an actual profit is shown. But if you're £20 under expectation right now you should expect to be there forever, you'll just not mind because you've made money overall.

The goal of xG and other football stats is to paint a similar picture to the poker simulator - will this level of football give me a high degree of confidence in positive future results - because what you don't want to do is get unlucky a few times, panic, and then make changes to your game that aren't warranted. And equally important is not to get deluded because you have a few good results.

In the long term results are a function of how well you play, so you need to look at how well you're playing and not always what the result was in order to improve. Where it gets really messy is that unlike a coin toss, football matches aren't truly independent probabilities. Egan's red card impacts the game after it, JOC's injury affects the entire season, and the buying of Brewster means our true xG might now be different.

 

[HolmesfieldBlade]

I might butcher this explanation but the idea that past events will be evened out by future events is the gambler's fallacy e.g. I flipped four heads in a row so in the future I'll flip some extra tails to even things out. But what stats do tell you is what you should expect in the future e.g. if I've had an expected goal stat of one per game then I should expect to score one per game in the future and the fact I scored 0 or 2 goals in the last game is misleading when predicting future performances. So if we played last season's Villa games several times under the same conditions then we could be confident of being on the right end of the results.

For anyone who might be interested in playing with graphs try this poker ev simulator:

Cash Game Variance Calculator

Poker Variance Calculator for cash games. Analyze Risk of Ruin, Confidence Intervals and other data to protect your poker bankroll.

www.primedope.com

Hit calculate and then increase the line for hand number and see what happens. The lines between expected and actual get closer as the sample size increases (this is a function of the scale of the graph as a single loss starts to represent less and less). What you'll also see is that over a small sample a positive expectation has a reasonable chance of losing overall, but the higher the sample size the more and more likely it becomes that an actual profit is shown. But if you're £20 under expectation right now you should expect to be there forever, you'll just not mind because you've made money overall.

The goal of xG and other football stats is to paint a similar picture to the poker simulator - will this level of football give me a high degree of confidence in positive future results - because what you don't want to do is get unlucky a few times, panic, and then make changes to your game that aren't warranted. And equally important is not to get deluded because you have a few good results.

In the long term results are a function of how well you play, so you need to look at how well you're playing and not always what the result was in order to improve. Where it gets really messy is that unlike a coin toss, football matches aren't truly independent probabilities. Egan's red card impacts the game after it, JOC's injury affects the entire season, and the buying of Brewster means our true xG might now be different.

That's a really good explanation of xG. Ultimately, it's an indicator that you can use to assess a team's performance.

And the takeaway point, we haven't performed that badly in the games we've played. Even if we play to the standard we have over the first 4 fixtures (removing finishing), we should comfortably stay in the league.

I keep thinking about that Meslier save vs Lundstram, that was fucking ridiculous.

 

[beatific blade]

It's far from bollocks. It's a useful statistic. If the stat were bollocks you'd expect to see lots of players outperforming it by having superior finishing, and lots of player underperforming it by a distance by being weak finishers. The reality is that players like McGoldrick coming massively under expectation are the anomalies and I'm not sure there's anyone who's consistently outperformed their xG. What xG seems to show is that scoring goals is much more about players creating chances and getting shots away (i.e. the ability to beat a man, make a run, find a pocket of space and so on) than it is about being a good finisher.

It's far from the be all and end all of football but it is a useful indication of how well you're likely to do in future games.

In James Tippett's book, The Expected Goals Philosophy, he says there is only one player in Europe who has overachieved their xG consistently over 5 seasons:

Messi - which says it all really.

xG is incredibly powerful. I think it's better for analysing attacking/defensive effectiveness of a team rather than the input of an individual.

 

[PokerBlade]

That's a really good explanation of xG. Ultimately, it's an indicator that you can use to assess a team's performance.

And the takeaway point, we haven't performed that badly in the games we've played. Even if we play to the standard we have over the first 4 fixtures (removing finishing), we should comfortably stay in the league.

I keep thinking about that Meslier save vs Lundstram, that was fucking ridiculous.

Yeah, it's a lot easier to see in a game like poker where the element of chance is well defined. The cards are random and there's nothing you can do to change them. Big swings happen to even the best players just by dumb luck. I remember following a player who ran a huge bet on turning $100 into $10k in 15 days. He did it, which is insane, but even that run involved something like a few hundred games break even stretch. Point being that during that break even stretch he had to keep the confidence that he was doing the right things. If he suddenly changes a winning strategy then everything goes to pot.

Football's less chaotic than poker but the same principles apply. You need to know whether you've been good or lucky or bad or unlucky or you don't know what to do about it.

In James Tippett's book, The Expected Goals Philosophy, he says there is only one player in Europe who has overachieved their xG consistently over 5 seasons:

Messi - which says it all really.

xG is incredibly powerful. I think it's better for analysing attacking/defensive effectiveness of a team rather than the input of an individual.

I nearly mentioned that but I wasn't sure I was remembering properly. Someone's mentioned that, and that book, before (probably you?).

 

[robbiez666]

I'm confused how anyone would think Palace have had a bad start?

Beat Man U and Southampton. Lost to the league's form team (Everton) and Chelsea. They have more points than expected

 

[Bladesman]

 

[Bladesthorpe]

All these stats and predictions do my head in, it's simple look at the players in your team and what they create or score, how good they are defensively and the quality of the other teams in your division. On that basis I predict we will be relegated by February with the least number of points in the Prems history.

 

[jono_t2000]

If you flip a coin and get heads 100 times in a row, ur next flip still only has 50/50 chance of being heads or tails regardless of past performance. Dem da facts.

 

[Latters]

It’s all well and good applying chance and statistics to flipping coins and drawing cards because outside factors don’t influence these things. As jono_t2000 said above, regardless of what you’ve just flipped, there’s an equal chance of heads or tails. There’s a one in 52 chance of any given card being drawn, diminishing by one each time. What the stats can’t do, as far as I can tell, is account for an individual’s ability. If a striker is off form, or just bad at finishing, like at least one, maybe a couple of ours are, that won’t balance out over a season. As I said above, hopefully Brewster can tip the scales back in our favour.

What would the expected goals be here? You’d at least expect one to hit the target, or would you...?

 

[PokerBlade]

It’s all well and good applying chance and statistics to flipping coins and drawing cards because outside factors don’t influence these things. As jono_t2000 said above, regardless of what you’ve just flipped, there’s an equal chance of heads or tails. There’s a one in 52 chance of any given card being drawn, diminishing by one each time. What the stats can’t do, as far as I can tell, is account for an individual’s ability. If a striker is off form, or just bad at finishing, like at least one, maybe a couple of ours are, that won’t balance out over a season. As I said above, hopefully Brewster can tip the scales back in our favour.

What would the expected goals be here? You’d at least expect one to hit the target, or would you...?

We use statistical modelling for all sorts of complicated interactions like this. Economics for example is entirely made up of individual choices (how businesses are run, how consumers buy all ultimately come down to free choices at the individual level) and yet we can predict trends and make models. Our models are always imperfect but at the same time there's not much a player does on a pitch that can't be turned into a statistic.

If finishing were the big factor then you'd expect to see the top finishers exceeding their xG. As BeatificBlade pointed out, only Messi has managed that. What seems to make the difference between the top strikers is the ability to get into good positions and get shots away. And it's exactly that type of insight into the game that we get from looking into the stats.

 

[Latters]

We use statistical modelling for all sorts of complicated interactions like this. Economics for example is entirely made up of individual choices (how businesses are run, how consumers buy all ultimately come down to free choices at the individual level) and yet we can predict trends and make models. Our models are always imperfect but at the same time there's not much a player does on a pitch that can't be turned into a statistic.

If finishing were the big factor then you'd expect to see the top finishers exceeding their xG. As BeatificBlade pointed out, only Messi has managed that. What seems to make the difference between the top strikers is the ability to get into good positions and get shots away. And it's exactly that type of insight into the game that we get from looking into the stats.

Economics are controlled though. Models affect consumer choices because the predictions have an impact on demand and prices. The predictions become a self-fulfilling prophecy, in a way. A silly, basic example would be the nonsense with toilet rolls. People expect the demand to go up, so everyone rushes out and buys them, thus increasing demand. I know that’s a fundamental example, but it can be scaled up and down to all sorts of trends and transactions.

With regard to chances, the stats can’t account for somebody who’s got into the right position being slightly off balance, distracted by something or just not very good. The ball in could be a mile an hour too fast or too slow, or an inch too far to one side. I just can’t see how anybody can take comfort from stats. As Kev Gee said, the only stats that matter and that can be relied on are goals and points.

 

[PokerBlade]

Economics are controlled though. Models affect consumer choices because the predictions have an impact on demand and prices. The predictions become a self-fulfilling prophecy, in a way. A silly, basic example would be the nonsense with toilet rolls. People expect the demand to go up, so everyone rushes out and buys them, thus increasing demand. I know that’s a fundamental example, but it can be scaled up and down to all sorts of trends and transactions.

With regard to chances, the stats can’t account for somebody who’s got into the right position being slightly off balance, distracted by something or just not very good. The ball in could be a mile an hour too fast or too slow, or an inch too far to one side. I just can’t see how anybody can take comfort from stats. As Kev Gee said, the only stats that matter and that can be relied on are goals and points.

The stats DO account for all those things, they're included when looking at the % of shots scored in that position. For xG you take a sample of a large number of shots and divvy them up by distance from goal, angle, defenders in the way, whatever criteria you choose to roughly determine what a "similar" chance is. You then look at the % of shots scored in each similar chance. This is the xG. You then take a bunch of players and work out their individual xG from this.

If it were the case that the ability to finish a chance were greatly different between strikers (i.e. some are really good finishers, some are really bad) then what you'd see is that xG wouldn't correlate well to the number of goals a player actually scores. You'd see really good finishers would consistently be well over their xG, and you'd see bad players would be well below it. But that's not what we see. We see that only one player in Europe has consistently outperformed their xG, and very few consistently underperform. That tells us that there's little difference in finishing at the elite level.

You're right that xG is far from perfect. The criteria used is somewhat arbitrary and imprecise, but if it were useless then we'd be able to see empirically that it didn't correlate to actual results. That's not what the data shows.

 

[BlairBlade]

Glad to see it all goes to show there is no evidence things even themselves out over a season.

I have always assumed that was a cop-out to stall too many recriminations, frustrations, debate, etc.

I do not want to get "too" involved in this, but tossing/flipping or whatever you do with a coin is a simple matter of probability. It will be either heads or tails.

Things just might be a little more complex when it comes to a game of premiership football or, for that matter, a game of football on Hackney Marshes on Sunday morning in the rain.

However, the elements of team investment, preparation, training, quality of players on the bench, fitness, attitude, etc. may well be a little different when comparing and contrasting the two. Those variables will exist in every game over a whole season and will, no doubt, change regularly and in totally unpredictable ways.

Still looking for that evening up over the Tevez saga.

 

[PokerBlade]

Glad to see it all goes to show there is no evidence things even themselves out over a season.

I have always assumed that was a cop-out to stall too many recriminations, frustrations, debate, etc.

I do not want to get "too" involved in this, but tossing/flipping or whatever you do with a coin is a simple matter of probability. It will be either heads or tails.

Things just might be a little more complex when it comes to a game of premiership football or, for that matter, a game of football on Hackney Marshes on Sunday morning in the rain.

However, the elements of team investment, preparation, training, quality of players on the bench, fitness, attitude, etc. may well be a little different when comparing and contrasting the two. Those variables will exist in every game over a whole season and will, no doubt, change regularly and in totally unpredictable ways.

Still looking for that evening up over the Tevez saga.

Football is more complicated than coin tosses, yes, but we can still use statistics to analyse performance. We build a model, we test it empirically, then we see if it makes useful predictions. We all look for things like how many chances we create, how well we keep possession, the quality of chances in a game, how hard the players are working. Stats allow us to make a more precise measurement of those things.

 

[Big Hitter]

If you flip a coin and get heads 100 times in a row, ur next flip still only has 50/50 chance of being heads or tails regardless of past performance. Dem da facts.

Then you get the law of diminishing returns coming into play Bro.