An analytical assessment of league one this season


Status
Not open for further replies.
Tim hat on but we really should be doing something about how many good chances we allow for the opposition and the fact that we aren’t creating all that many good ones for ourselves - regardless of how good our keeper and finishers are. It’s great to see the lad save two one on ones in a minute like at Donny but the more we allow that sort of thing, the more chance we give to bad luck to occur. Plus I’m not sure my ticker can put up with it in the long term.
 
Is it simply laying off against the money already bet or is there also a strategic element from their own predictions built in ?
If, in-house, they believe in a certain outcome it would be sensible to discourage more money going on it.
Don't know, just guessing.

Just about balancing the book, so they make a profit whatever the outcome. They will initially set the odds based on knowledge/statistics, but once its up and running its just using maths and the amount of money placed.
 
just looking at the team comments at the end they clearly don't like us, yeah right we're just lucky.

If they had made a point on the first graph that the 3 positive discrepancies were the top 3 so their analysis was aimed at the wrong data then that would have helped, but it seems to me the conclusions had been written before the analysis was done.
 
I think it's worth comparing this season and its stats to our first season under O'Neill. In that season it seemed that slot of 'worldies' were flying in and we were getting the rub of the green in many matches. I remember George Caulkin talking with Roker Report and saying pretty much the same thing. It couldn't last and it didn't.

I don't think anybody would argue that we haven't had games where we've been very fortunate to win based on chances. Wimbledon away springs to mind immediately. I think the only game where the opposite is true is probably Peterborough where, based on chances, we deserved to win the game comfortably.

Anyway, I think it makes for an interesting talking point and I, for one, will be interested to see what these stats look like later in the season.
 
Not sure why everyone is being so defensive. Obviously our quality is the difference, the keeper has being key in the recent winning run and we seem to score a goal of the week contender every game.

We have yet to really hit top form and destroy teams, when we do the stats will change and we will see a lot more scorelines with 3, 4 and 5 in the for column.

I always thought if we could be in the mix at Christmas the team will have gelled to take control in the new year, nothing so far seems to contradict this.
 
Tim hat on but we really should be doing something about how many good chances we allow for the opposition and the fact that we aren’t creating all that many good ones for ourselves - regardless of how good our keeper and finishers are. It’s great to see the lad save two one on ones in a minute like at Donny but the more we allow that sort of thing, the more chance we give to bad luck to occur. Plus I’m not sure my ticker can put up with it in the long term.
The vast majority of people on here have been saying the same thing all season, it’s just when an outsider suggests it people seem to get their backs up.

I thought the article, though in my opinion flawed in its conclusion, contained some really interesting information. The lad’s first post on here was decent too but the fact he’s subsequently gone on like a bit of a condescending prick won’t help.
 
I’ve see the expected goals stat used on tipping sites and in the racing post. It’s not something I personally rate. At least the lad has come on and tried to justify it but then says he’s a Luton fan and has watched Barnsley so bases it partly on that.

Has he watched safc, fully more than the Luton 1-1?
He has watched us three times. I enjoyed his post, he makes some interesting observations.
 
As I said I'm happy to accept that we have been running a little 'warm' to date, but any model that suggests we should have lost that game yesterday because they had twice as many shots as us is overly simplistic and deeply flawed. We were at no point ever in danger of falling behind yesterday.

And if the best his model can say is that we are in part getting the rub of the green and in part just have better players, but it can't even start quantifying what difference each makes then that is just stating what we can all see anyway and the ''model" doesn't add anything.

He is of course correct that Maja won't keep scoring one in two chances etc. But maybe other players' clinicality will improve (Sinclair would be the obvious one) or maybe we'll start creating more chances/shots taking more shots if we have to (ie move up the gears as needed).

Another thing I'm not sure of is whether our particular style of play or game management might somehow be confusing the model. I don't know enough about the model to comment on that.
 
Last time I'll be on here. I won't waste too much of my breath (or your time) cos it's clear some of you are adamant you don't want to accept it which, as I said in my earlier post, is fine. Some questions have been asked so I'll answer them.



I said this in my earlier post: "For what it's worth, I've seen Sunderland 3 times this season (vs Luton, Shrewsbury, and Southend). In none of those games would I say Sunderland were obviously the better side in terms of chances created. I'd say that the games were all pretty balanced overall but then you've just scored your chances whilst other teams have missed theirs. Call it clinical finishing, call it killer instinct, call it whatever you want. The point is to imagine what if you had missed those chances, or what if your opponents had converted theirs. You may disagree with me on that if you want."

The data is based on myself and colleagues having watched *every single minute of League One football this season*. The stats aren't just plucked out of thin air, nor are our opinions. I'm sorry you don't like or agree with them.

Kipper said something about teams setting up defensively against Sunderland and therefore expected goals is bollocks:

Kipper, I'm sorry to say that just because a team sets out to play defensively against you, doesn't make expected goals bollocks. Teams have done it against Barnsley, Luton, Portsmouth too. Think of the Premier League using your example. Man City are currently the best team in the league because they're better than anyone else at creating chances, especially against those teams that sit deep which is the majority of the league other than the big 6. All the big 6 face that issue. Man City do it better.





The last sentence of your paragraph is literally what is being shown up in the data so I'm glad you agree with me and it's nice to know someone on here is at least a little open to the idea of searching for an explanation for why Sunderland show up badly in the analysis rather than just rubbishing it. Chickenbaltipie has my respect too because at least the two of you have tried to come up and think of reasons and arguments, which is really good.

The reason why I mentioned "luck" as a reason is because there are very, and I mean very, few occasions where teams outperform (or underperform for that matter) expected goals over a longer period of time which is why it's regarded as an important tool. 16 games is a small sample and for the most part you wouldn't expect teams to get the results they deserve through that sample of games, i.e teams will get lucky or unlucky in that time. The player skill argument is fair but also it just isn't enough of a reason to explain it. Sunderland's players and goalkeeper are not that much better than the rest of the league to explain that large a discrepancy. It could be one factor yes, but not the entirety of it. In the original article I said either Sunderland's "performances" would improve, or results would worsen. Remember results worsening doesn't have to mean a 5 game losing streak. It might just be picking up 18 points from a possible 30, rather than 24 for example.

The Maja-Messi example - that Maja is converting chances at a higher rate than Messi - was to get you to consider that maybe that just isn't going to continue longer term. A good striker at any level will finish his chances at about a 1-in-4, 1-in-5 rate. Maja is 1-in-2. I will bet my net worth with any and all of you that Maja does not score 1 goal for every 2 shots he takes for the rest of the season and his career. If you don't want that bet, I'm afraid that means you agree with me and to some extent my analysis, whether you like it or not.

The 3% conversion rate which Chickenbaltipie raised is based on literally everybody, yes which I'm happy to admit is a flaw within the model (again I've mentioned that it isn't perfect). However, an interesting exercise for you all to consider: Maguire shoots from 30 yards. Just how many times do you believe it goes in out of 100 attempts? He probably doesn't hit the target 60% of the time (pure estimation, it'd probably be even more than that). So. hypothetically, 40% of his efforts from 30 yards hit the target. Then the shot has to be good enough to beat the keeper who will have a long time to see it coming. So it has to be right in the corner and with considerable power. How many times will he hit the ball well enough to be on target, then with enough pace and power, then with enough accuracy, then hoping the keeper reacts late, for it result in a goal? Probably not more than 5%. If you want to argue with me over a shot being worth 3 goals out of 100 attempts or 5 goals out of 100 attempts just because you believe your favourite player *might* be significantly better than an average player, be my guest. As I said, no Sunderland player is that much better at finishing than other players in League One for it to explain the entire discrepancy.

One thing you all might find interesting is that by this type of analysis, David De Gea had Man United overperforming in the league last season - it was estimated he saved them around 7-9 goals fewer than expected. Look how well that's going for them now his form's dropped off slightly.

Anyway, well done on your win today. Despite the animosity by some of you, I've enjoyed the conversation and reading your thoughts. I don't blame people having no interest or time for it whatsoever but, if that's the case, there's no need to wet your pants over it, is there? I do wonder if it would've had the same reaction if it had said Sunderland were good and Portsmouth weren't so good, too.

Cheers all.
If he plays against the shitty 'keepers and defenders in league 1, he might well do. :) If you look at where he gets his chances, they're, I think, what would be classed as relatively high value chances... unmarked, with a clear shot at goal from relatively close in. What is interesting is how well he places shots - a large proportion of his goals are not hit with power, but are very well placed. His current stats are due to high value opportunities, plus being a better finisher. It's looking certain that Josh Maja will play at a higher level than League 1, so your 'bet' is a bollocks analogy. By your own methodology, if he continues to play as he does, now, getting lower numbers of chances, but, high value chances, his conversion rate shouldn't change, significantly, otherwise, the system is flawed, which you're advocating it isn't. Look at his goals, they're good chances, finished well.
 
Last edited:
Afternoon all,

I noticed my article had picked up a fair amount of clicks through your site so I thought I'd check it out. It seems I've struck a nerve. Thanks to the few of you who found it interesting, apologies to those of you who felt their time had been wasted and thought it was "a load of bollocks", and other adjectives.

If you would give me just a couple more minutes I'd like to try and clarify some things. Believe me I'm not going to try and convince you to come around to my way of thinking and I also, for the record, do not believe that I am 100% right but there's been some interesting points raised in this thread that weren't fully explained in the article so I thought maybe if I offered more explanation then it might make more sense. I'll try and be brief but address some of the points raised.

By the way, I completely appreciate this way of looking at football isn't for everyone so please feel free to ignore this and carry on throwing pelters my way.

1) What is Expected Goals?
So, everytime a player takes a shot, it has a probability of going in. That's not me being mathsy or nerdy, it's just truth. Everytime a player picks up the ball, cuts inside and shoots from the edge of the area, there's a probability of that shot becoming a goal. Everytime a player is played in behind the defence and is through one on one, that also has a probability of becoming a goal. More likely than the shot from the edge of the area, right? Every shot is unique, but shots that have similar factors about them (i.e the location the shot was taken from, the body part used to strike the ball, positioning of defenders between ball and goal, etc) can be grouped together and you can use the average conversion rates to predict the likelihood of future shots going in. Over large samples, you can then say "well this is how many goals we would expected team X to score based on the shots they've taken" and vice versa for conceded. It's not perfect and it's not definite. When I say that Chris Maguire's goal against Southend had about a 3% chance of going in, that's because in simple terms shots from that distance have gone in 3 times out of 100 historically. That's not to say it was *exactly* 3%, it's just the most likely estimation we have based on the information we have.

Expected Goals is mostly used in gambling by bettors and bookmakers to benchmark teams ability but is also used within club boardrooms and in recruitment and analysis departments to analyse other teams and players.

2) Why is it important?
If you're still with me reading this then I'm glad you're keeping an open mind to this. I'm sure most of you would agree that creating more chances and better chances than your opponents on a regular basis is an important factor in generating good runs of results. Studies show that it is (currently) the best method of predicting future performance in smaller samples - better than number of shots taken, possession, goals scored, everything. It correlates well with points and goals scored and conceded though not every time - as is the case with Sunderland. That's also the point of using analysis like this though - if it lined up perfectly with the goals tallies of each team, what would be the point of it? The point of it is to raise questions about whether a team is getting "lucky or unlucky" in simplistic terms and do more research. Sometimes, there'll be good reasons as to why a team looks like an outlier. Other times, it really is the case that they may have just been getting lucky or unlucky over a run of fixtures.

3) Does Sunderland being 11th for Expected Goal Difference mean Sunderland aren't good?

No it doesn't, but is does suggest they aren't as good as other teams in the league at creating good chances regularly, and also aren't as good as other teams in the league at preventing chances. However, there are many reasons why it doesn't have to mean that Sunderland aren't good. One is that every model is flawed and a lot of you are completely right to point out that it matters which players the chances fall to and that Sunderland as a team might have better finishers than the rest of the league. Two is that there is other context sometimes missing that is only clear once digging even deeper. Someone mentioned that you've had played with 10 men quite often this season. Valid point. Another is that you've also not been playing with your strongest team for a lot of the season because of injuries. There are things like that that I haven't included because that's information I'd use after looking at the model to contextualise it. Hopefully you can all see that whilst Sunderland stick out, most of the other teams in the league are benchmarked fairly well.

For what it's worth, I've seen Sunderland 3 times this season (vs Luton, Shrewsbury, and Southend). In none of those games would I say Sunderland were obviously the better side in terms of chances created. I'd say that the games were all pretty balanced overall but then you've just scored your chances whilst other teams have missed theirs. Call it clinical finishing, call it killer instinct, call it whatever you want. The point is to imagine what if you had missed those chances, or what if your opponents had converted theirs. You may disagree with me on that if you want.

4) Should a model like this feature player ability?
Yeah, ideally it would. It's very hard to incorporate this for a number of reasons though. I can't just say Josh Maja is a good finisher because he's only taken 20 shots in League One this season. It would take a helluva lot more shots to statistically prove that he was an above average finisher. For the record, he's scored 9 goals from 20 shots this season which is really good, but also a better conversion rate than Messi over Messi's entire career. If you want to argue with me than Maja is a better finisher than Messi then fair play but I'd respectfully disagree. To be clear, I'm absolutely not saying that finishing skill does not exist - obviously some players are better finishers than others. It's just than in Expected Goals, that's something you'd take into account when forming your analysis. I don't think Sunderland are *that* much better finishers than the rest of the league though, because there are plenty of other good players and good goalkeepers in the league that don't play for Sunderland.

By the way, if you want an idea of the work that has to go into quantifying finishing skill, then please google search "Statsbomb quantifying finishing skill" and read the first article. Though if you hated my piece, you'll probably spontaneously combust when reading that.


5) Why does everyone love Barnsley?
I noticed a few people suggest that I said Barnsley were the best team in the league in the article, which I didn't at any point - I literally said they're on a downward trend. But, for what it's worth, from the eye test Barnsley *are* the best side I've seen play against Luton (I'm a Luton fan) this season. I've also watched them in several other games (as I have a job that requires me to watch a lot of League One) and they've also absolutely dominated teams in those games to a degree that I haven't seen from another League One side this season. It's just my opinion, you're allowed to feel differently. From the numbers used in my piece, they profiled as the best side in the league up until a couple of weeks ago. Using other measurements, as someone posted just now you might notice that Barnsley are now virtually level with Sunderland in the bookmakers eyes, despite Sunderland being heavily favoured over them prior to the season. That's despite Sunderland being above Barnsley in the table. Either Barnsley are performing better than the bookmakers thought they would do prior to the season kick off, or Sunderland are performing worse. In my opinion, it's both, again you're entitled to your own thoughts on that.

Anyway, as I said I know not everyone enjoys deeper analysis of football and do not like what I had to say in the article. Some of you probably don't like what I've had to say in this post either. That's fair enough, everyone's entitled to watch and think about football in whatever way they please. For me personally, my profession requires me to look at the things mentioned in the article and think about fooball in a similar way to what was written. One last point I'd like to reiterate is that I do not think that was was in the article is 100% fact and nothing else matters. It's missing some context, but I wasn't about to give 1000 words on every team and why they may look good or not so good by these measures and hopefully you'll forgive me for that and appreciate that you're the only club who I've gone to greater lengths for to offer an explanation! Fans of other clubs, such as Peterborough who I cast in a similar light, have actually told me that what I've said matches with what they've seen with their own eyes. It's interesting that the vast majority of you don't feel that way about Sunderland.

Good luck for the rest of the season. Everyone knows you shouldn't be in this division and aren't likely to be in it for long so please don't let an article written by a "nerd" ruin any of the enjoyment you've had so far.
All models are false but some are useful. I suspect you are invested in the expected goals model and it is obviously very powerful. Although most of the questions on here are about players finishing i suspect the goalkeeper has the largest effect. Did Shrewsbury 'beat' there expected goals last season. I see Man utd did with De Gea and i bet if you went back to the Schmeichel days they outperformed expected goals.
 
As I said I'm happy to accept that we have been running a little 'warm' to date, but any model that suggests we should have lost that game yesterday because they had twice as many shots as us is overly simplistic and deeply flawed. We were at no point ever in danger of falling behind yesterday.

Experimental 361 does a timeline of xG for each game. Yesterday's are here: League 1 timelines, 3 Nov 2018

According to this, Sunderland were ahead in xG up until they scored, then it swing back to Plymouth. Overall xG for the game was Plymouth 1.8 Sunderland 1.3, which suggests it was Plymouth not taking advantage of chances more than Sunderland being clinical.

Did Shrewsbury 'beat' there expected goals last season.
There's a decent analysis of last year's League One stats, including xG, here: Inside League One: Teams, Trends And Metrics | StatsBomb

Shrewsbury were 8th in the xG table last year, with about +5 xG over the course of the season, and we're the second highest overachievers (after Gillingham) with 0.4 xG over predicted per game.
 
Last edited:
Experimental 361 does a timeline of xG for each game. Yesterday's are here: League 1 timelines, 3 Nov 2018

According to this, Sunderland were ahead in xG up until they scored, then it swing back to Plymouth. Overall xG for the game was Plymouth 1.8 Sunderland 1.3, which suggests it was Plymouth not taking advantage of chances more than Sunderland being clinical.


There's a decent analysis of last year's League One stats, including xG, here: Inside League One: Teams, Trends And Metrics | StatsBomb

Shrewsbury were 8th in the xG table last year, with about +5 xG over the course of the season, and we're the second highest overachievers (after Gillingham) with 0.4 xG over predicted per game.
Which had abosolutley nothing to do with Shrewsbury having Dean Henderson in goal a keeper who will be at worst understudy to man u keeper in a few years
 
So is that a 'yes' then? The expected goals model suggests Plymouth ''should" have won 2-1?

What sort of gap is needed to suggest a draw is more likely?

Is it a better or worse predictor than just running the games through Champ Manager by the way?
 
Which had abosolutley nothing to do with Shrewsbury having Dean Henderson in goal a keeper who will be at worst understudy to man u keeper in a few years
They scored way less than the top two, and conceded way less than the teams below them, so you're probably right that the main reason is their keeper/defence, although their disparity between expected and actual goal difference was about 0.4 per game, a third of what Sunderland currently have, which suggests more is at play for Sunderland than just good goalkeeping.

So is that a 'yes' then? The expected goals model suggests Plymouth ''should" have won 2-1?

What sort of gap is needed to suggest a draw is more likely?
I'm no statistician, but I'm guessing a gap of 0.5 or less means a draw is most likely.
 
Last edited:
I appreciate you taking the time to come on here and write this. I can sense an understandable amount of defensiveness in your post. I'm sure you see it as daft Sunderland fans just having their nose put out because a report doesn't blow smoke up their arse. But you seem to be writing off the outlier that is Sunderland's real-world position as a 'luck' rather than acknowledging that the model may be flawed. It doesn't tally up for us, because we are both where we expected to be, and (for the most part) where we deserve to be.

People have mentioned things like the sending offs as being factors in while the model might be skewed, but there's other factors. A key one for me is that Sunderland this season have suffered from the "cup final" effect, where opposition teams raise their game and completely alter their normal tactics against us. This is not Sunderland arrogance... check social media or listen the the podcasts of most of our opponents before they faced us this season. Or listen to the oppo manager's post-match interviews. It's clear they have approached it differently. We got caught out early on, but we seemed to have sussed it now.

But the main thing, by a mile, is player ability.


I think this is at the absolute heart of why it's bollocks. You dismiss the player ability factor as a nice-to-have but basically say it's too hard to do. I'd say it's a fairly massive oversight that virtually renders the model meaningless.

Like them or not, you look at games like FIFA and (to a much better extent) Football Manager and the absolute core part of the game engine is individual player ability. FM is basically just a big number crunching machine. These attributes include things like Finishing and Long Shots. Quite obviously, like the real-world, some players are better at some things than others. All teams have 11 men in them, but Barcelona has 11 men who are better at footy than the 11 men who play for Basingstoke. It's not controversial.

So a 3% goal chance from a shot from 30 yards. But 3% based on who? Just an average of literally everybody? Which presumably includes players as diverse as Cristiano Ronaldo and Jozy Altidore? Would I have a 3% chance of scoring?? (I'm shit by the way). Not taking the player's "long shot" attributes into the equation when calculating the likelihood of a goal is just plain daft. Maguire has done it twice this season, and it's not like he spends the entire game attempting failed worldies. He's done it enough in his career to indicate that his aptitude for long shots is high. So 3% for him is harsh, meanwhile, Maja's "Finishing" attribute is way better than a lot of the other strikers I have seen so far this season. Conversely... the percentage chance of scoring against us, fails to take into account McLaughlin's "Shot Stopping" attribute, which is (trust me) WAY higher than Lee Camp's.

When you boil the piss off it, we've scored the second highest number of goals in the league, because we have the players with the individual ability to take their chances. We've conceded the least number of goals in the league because we have a stout defence and a class goalie. The reality (in terms of the league table) matches what we see; a team of players with the individual attributes to comfortably see us out of this league. To write it off as luck because "Computer says no" feels wrong.
Well said. There's absolutely no doubt that the opposition are raising their game against us. Putting more into it than they would any other team. I've seen Fleetwood twice this season, once against us and against peterborough, the effort and intensity they put into our game compared to there's was like night and day.
We're the main "scalp" in this league, no doubt about it.
 
So once again on Saturday we had the better players to score goals and the better keeper. Obviously this level of quality can not be maintained and both our keeper and strikers will realise this and reverse to the league 1 norm of being crap.
 
Experimental 361 does a timeline of xG for each game. Yesterday's are here: League 1 timelines, 3 Nov 2018

According to this, Sunderland were ahead in xG up until they scored, then it swing back to Plymouth. Overall xG for the game was Plymouth 1.8 Sunderland 1.3, which suggests it was Plymouth not taking advantage of chances more than Sunderland being clinical.


There's a decent analysis of last year's League One stats, including xG, here: Inside League One: Teams, Trends And Metrics | StatsBomb

Shrewsbury were 8th in the xG table last year, with about +5 xG over the course of the season, and we're the second highest overachievers (after Gillingham) with 0.4 xG over predicted per game.
:lol:
 
Status
Not open for further replies.

Back
Top