CroJack's Corner

CroJack

Key Player
I have finally finished my expected goals model based on exact data (goal opening, distance and angles). My model is based on the real life action, and it's similar to goal-line technology and VAR. I use geometry. The expected goals models based on statistics are fraud. The people who make expected goals models based on statistics don't watch real games and their models give a totally wrong picture of what is going on.

Today I am going to make charts with Swans' expected points, expected goals, and expected goals against for this season. You will be surprised as much as I was when I finished calculating Swansea data. Stay tuned.
 

CroJack

Key Player
This is a table with all Swansea games, expected goals, results, expected results, points and expected points.
We have been lucky in 10, and unlucky in 9 games. Who says that luck doesn't even out over the course of a season? :)
You will be surprised when you see that we should have beaten both Norwich and Brentford in the first half of the season.

Expected Points.jpg

2. All Expected Goals-1800.jpg

This is interesting!

1.jpg
I got shocked when I finished calculating Swansea's expected goals for and against.

First, we have created enough chances to score 68 goals, but we have scored only 45. That's a deficit of 23 goals! Some credit have to go to the opposition goalkeepers, but most of these unconverted chances reflect our own wastefulness up front. Had we won 5 out of those 9 games where we were 'unlucky', then we would have been in a much better position.

Second, we have conceded 29, but we should have conceded 51 goals. Again a huge difference of 22 goals. We should give some credit to Woodman, he has after all made some excellent saves, but again, the vast majority of these unconverted chances can be attributed to the opposition's wastefulness in front of goal.

Conclusion

We create plenty of chances, but our wingers who play as strikers are not clinical enough.
We have some good defenders, but we don't defend as a team. 1,39 expected goals per game is simply too much.


This is a table with all data for the Swansea games prior to Hourihane's arrival

2.jpg
The same wastefulness in front of goal. Not defending as a team.
During this period we conceded 30% fewer goals than our season average. We didn't covert 35% of the chances we created. Also, during this period we should have earned 5 more points. We were unlucky.

This is a table with all data for the Swansea games since Hourihane's arrival

3.jpg

Slightly more clinical in front of goal. 27% chances not converted. We have earned 5 more points than we should. We were lucky.

Have a look at Goals Against Per Game. Before Hourihane arrived, we conceded 0,54 goals per game, and after, almost double as many, 1,23. From 1,1 we went up to 1,88 expected goals against per game.

Also, the number of non-penalty goals per game went down from 1,16 to 0,92, and expected non-penalty goals per game went down from 1,86 to 1,51.

It's worth mentioning that before Hourihane's arrival the vast majority of the goals we scored were goals from open play, and after Hourihane's arrival we scored more goals from set-pieces than from open play.


Conclusion
1. We are not so poor at creating chances. We have created many more chances than expected goals statistical models have (wrongly) calculated. Both Infogol and Experimental361 give us 42 expected goals for. This is not a surprise, because both of them probably use Opta's statistical data and Opta's expected goals model. Trust me, they are wrong. They don't watch live football.

2. We are worse defensively than many of us thought we were. First of all, we don't press and defend as a team. That's why we allow relatively many shots against. Second, our defensive record has lately been spoiled by Cooper's change in formation and personel. Third, there have been too many personal defensive errors, especially Woodman's. Though, it's not all Cooper's fault - we had injuries to Bennet and Guehi, but most of it is. Infogol have calculated 38, and Experimental361 40 expected goals against us. Both are wrong. We should have conceded 51 goal this season.

3. With Hourihane we have got an excellent set pieces specialist who can also score goals out of box. 5 goals and 1 assist in 10 x 90 minutes of play is a fantastic record, but, but...in my opinion since he arrived we haven't had a real balance in the starting XI. In the most (not all) games our back line have been too far away from our midfield line, so the back line haven't had any natural passing options. We have simply been too disjointed and therefore forced to hoof instead of pass the ball. This has resulted in struggles both up front and at the back. While we managed to compensate our lack of build-up flow with some set pieces goals, our defensive line has been shaky and conceded goals at a double rate than before Hourihane's arrival. In my humble opinion, we have played most balanced football this season when we played Grimes - Fulton - Smith midfield combo, and Guehi-Bennett-Cabango back three combo in 3-5-2 formation.

4. Since Hourihane arrived Lowe has scored 0 and Ayew 1 goal (from open play). I don't care about the goals from the penalty spot. Who says Lowe wouldn't have scored them? Only 1 goal between our two forwards is not a coincidence. They haven't suddenly forgotten how to score or become worse players. No, it's something else. We don't create chances by running and passing through the middle anymore. We hoof it.

5. The number of points we have earned this season (69), and the number of expected points calculated by using my expected goals method (68), are proof of the model's validity. The difference is only 1 point, which is insignificant.
 
Last edited:

Yankee_Jack

Key Player
Excellent analysis. As you say the model matches reality so it is validated.

After all that it comes down to a simple answer: drop Hourihane, play Cabango, Fulton, Bidwell. We have few games left. Play our core strength.

Or put another way Andy Pandy can't manage his new toys and when he tries to think we're lost.
 

CroJack

Key Player
As you say the model matches reality so it is validated.
I should have added that the model is accurate because in our case there is almost an identical ratio between 'lucky' and 'unlucky' games. Of course, some teams can be more lucky than unlycky, and vice versa. Then the number of the real points would differ more from the number of the expected points, but never as much as in the statistical expected goals models. Experimental361 want to make us believe that Swansea have been soooo lucky this season that the difference between the real points and the expected points is +17 which places us in the 11th place in the league table with a total of 42 expected points. That's simply not true.

I don't blame Infogol, Experimental361 and other statistical expected goals models too much. Their models are probably valid, but at the same time they are fundamentaly wrong, because their data is false. It is possible to build a valid theory on false premises. So I blame Opta for introducing an idea that the quality of the real life football chances can be calculated by using a database with all goals scored from a certain position and angle. Imagine if I said that an offside was offside just because most offsides have been recorded when players were placed in certain positions? How ridiculous that would be?

When we calculate the value of chances in football we should follow common sense. For god sake, every football player and every football fan knows what a good chance is. We don't need to complicate things. How many times have we screamed "He missed an open goal!"? We use terms "closing down opposition strikers" or "the goalkeeper should narrow the angle" etc. And that's exactly what Opta don't understand. The goal opening should be the most important data when you calculate the value of chances. Then distance, and then angle. The more the goal is open the better the chance is. It's so simple. And that's exactly what I am using in my model. I use camera angles and geometry to calculate the value of chances. Just like the goal-line technology use high-speed cameras to see whether the ball has crossed the goal-line or not. Fortunately, in modern football we have replays of chances from different angles, so it has been possible for me to calculate expected goals with a pretty good accuracy.
 
Last edited:

Behindthegoal

Key Player
Yes.
To translate that to another endeavour. If I compile an eclectic golf card of my best ever on each hole, I would record a round of 50, when truth be told I would be delighted to post a 90. You can’t be good and lucky all the time.
 

CroJack

Key Player
@CroJack ... so what’s the answer to the age old question .... is it better to be lucky or good?
Both :giggle: .

Jokes aside, in long term when it comes to football it's better to be good. A good football team has a perfect balance between attack and defence, is creative and effective up front and resilient at the back, does a lot of runnung and pressing, and has players who are good at reading the game. A good team has to have an excellent goalkeeper as well.

In my opinion, and Swansea expected points confirm that, we are exactly there where we should be: in the play-off places. We are better than Reading, Barnsley, Bournemouth, Middlesbrough and Cardiff, we are neither better nor worse than Watford, and we are worse than Norwich and Brentford.

Saying all that, Swansea are in the play-off places not because we are a good team, it's the rest of the Championship teams that are worse than we are.
 

CroJack

Key Player
Ok, it took me three days to watch the replays of all Brentford chances and chances of their opposing teams, and to compile expected goals and points data. I thought the Brentford expected goals and expected points numbers would be higher.

I was surprised to see that they have created only 3,1 expected goals more than Swansea. But, as you can see, they have coverted 93 % of all expected goals to actual goals, which is extraordinary. The difference between Toney, Canos, Forss, Dasilva and Mbuemo on one side, and Ayew, Lowe, Whittaker, Dhanda and Hourihane, on the other other, is obvious. Brentford have much better players up front than Swansea.

Defensively, Swansea are much better. Swansea have allowed 51,35 expected goals against, but conceded only 29. Brentford have allowed 6 fewer expected goals than Swansea, but conceded 10 more goals. Brentford defend better as a team, but Swansea play a more defensive football and are more difficult to score against.

Swansea have 1 point more then they have deserved, and Brentford have 2 points fewer than they have deserved. Not a big difference.

What I have noticed is that Brentford create a high number of high quality chances, but they also allow a relatively high number of high quality chances. Both are the result of their attacking style of play.

Brentford have been 4 times unlucky and lost 8 deserved points, but they have also been 5 times lucky and earned 9 undeserved points. Brentford results don't depend on luck, when they create more chances than opposition they score and win their games. Games like Swansea v. Bristol, where Swansea created 4,85 expected goals and lost, don't happen to Brentford.
 
Last edited:

CroJack

Key Player
More expected goals stupidities from the totally unreliable statistical models.

This is just one out of thousands chances I have examined. Have a good laugh.

On Feb 6, 2021, Brentford played an away game at Middlesbrough. In 38' Ivan Toney had probably one of the easiest tap-ins in his career. Open goal from 10 yards, and without any defensive presure a chance almost impossible to miss.

ToneyChance.jpg

Here is how Infogol (and others) have rated this chance.

Screenshot 2021-04-11 at 14.11.08.png

Why is this calculation stupid? Well, all statistical models agree that a penalty kick is a chance with 76% probability to score. And this is the only time I agree with them. But how on earth can a chance from 11 yards distance, with a goalkeeper between the sticks, have 76%, and a chance from 10 yards distance with open goal (no goalkeeper), 45% scoring probability?

In my book this is a 96% chance. Actually, for a professional football player this should be a 100% chance, but in a real life football incredible misses happens.
 
Last edited:

Behindthegoal

Key Player
Well, did he score? The suspense is killing me!
Acrually this expected goals malarkey has always flummoxed me, so it’s good to know it’s not set in stone
 

CroJack

Key Player
Key passes per X minutes (players who have played at least 180 minutes)

Screenshot 2021-09-24 at 12.56.14.png

Conversion rate (Piroe 22 shots 4 goals, Paterson 7 shots 2 goals, Ntcham 7 shots 1 goal)

Screenshot 2021-09-24 at 13.13.41.png
 
Last edited:
Top Bottom