STFC stat attack 2022/23

[Benzel]

Apologies had blanked that god awful season from my mind!!

20th after 4 games and finished bottom - so started shit and got even worse

We hung on to Wilson for too long and then had the whole Paul Hart debacle, if it comes to needing to move SL on it needs doing early.

[RobertT]

The only stat that should matter is points.  We currently have 3 after 4 games, which is the same pace as the worst season in our history (we have had shitter seasons for points but at higher divisions).  That year we were also on 3 points after 4 games and finished in 17th on 58 points.

I still remain firm on my belief that anything outside the top 7 is abject failure.  Any decisions should be taken with the prospect of achieving that in mind.

xG irritates me, it seems to be based purely on the location of the ball when it is struck to give a % occurrence that people score from that spot.  It doesn't take into account the context (how many defenders, how did the ball arrive at the strikers foot etc.).  An xG over 2 means we got pummeled I believe.

[reeves4england]

xG irritates me, it seems to be based purely on the location of the ball when it is struck to give a % occurrence that people score from that spot.  It doesn't take into account the context (how many defenders, how did the ball arrive at the strikers foot etc.).  An xG over 2 means we got pummeled I believe.

Whilst you're right about points being the only stat that really matters, the above simply isn't true. Different stats providers have different models but variables can include:

Location of shooter: How far was it from the goal and at what angle on the pitch?
Body part: Was it a header or off the shooter's foot?
Type of pass: Was it from a through ball, cross, set piece, etc?
Type of attack: Was it from an established possession? Was it off a rebound? Did the defense have time to get in position? Did it follow a dribble?

[Posh Red]

The only stat that should matter is points.  We currently have 3 after 4 games, which is the same pace as the worst season in our history (we have had shitter seasons for points but at higher divisions).  That year we were also on 3 points after 4 games and finished in 17th on 58 points.

I still remain firm on my belief that anything outside the top 7 is abject failure.  Any decisions should be taken with the prospect of achieving that in mind.

You can use stats to prove any point, for instance after 4 games under PDC we had 3 points, in fact we only had 3 points after 5 games & one of those was a home defeat to the Piss Stains.

The difference (as pointed out before) was that we let Paolo buy his way out of trouble and once he got the right formula HMS Piss the league was on track.

Not sure that Lindsey is going to be able to get us playing that well, he certainly won't have the budget to buy another team if they have got recruitment wrong.

[RobertT]

On the xG, something like a penalty shows how far it can be trusted, just in my opinion.  A penalty has a default value of about 0.76/0.77.  If you get Matt LeTissier to step up, that should actually be something like 0.96.  Peter Shilton in goal - 1.00,  Davison taking it - 0.25 (sorry, just being churlish).  It just seems like a daft stat to use at the team level.

Anyway, if we do use it, I think >2 in a game for the other team means they had a lot of chances - that's basically 3 penalties.

[Nijholts Nuts]

Are any TEF'ers actual professional Data Analysts?

[Batch]

Are any TEF'ers actual professional Data Analysts?

I'm going to guess a professional days analyst will tell us it's really good if that helps

[RobertT]

I'm going to guess a professional days analyst will tell us it's really good if that helps

What, xG?  Nope - I may not exactly be a Professional Data Analyst, but have worked in Data Analysis for over 20 years through various roles.

I'm not saying data does not have a role to play, it does.  The reliance on gut feel was certainly the old way, but Data can be used in ways that seem good on paper without producing "good football".  I refer you to Howard Wilkinson's influence on the National teams for a period.

xG is an oddity one for me - it's like a stat was generated just to generate a stat, and I think that is likely partly true.  I believe it began life as a data point within gambling.

Trying to describe this in football talk - you don't have very many games where a team has an xG of 4.7 for example.  Yet we all have witnessed games where we would have said "they were lucky to only concede 4", or, "we could have had another three or four".  The argument would be that the team has outperformed their xG and that is how you use the stat, but it still doesn't describe the game very well.  Data is only useful if you can do something with it - in football, how can I use that stat to improve my team.  I'm just not sure it does that, based on limited research to date.  It seems to have been adopted because it sounds fun.

[@MacPhlea]

What, xG?  Nope - I may not exactly be a Professional Data Analyst, but have worked in Data Analysis for over 20 years through various roles.

I'm not saying data does not have a role to play, it does.  The reliance on gut feel was certainly the old way, but Data can be used in ways that seem good on paper without producing "good football".  I refer you to Howard Wilkinson's influence on the National teams for a period.

xG is an oddity one for me - it's like a stat was generated just to generate a stat, and I think that is likely partly true.  I believe it began life as a data point within gambling.

Trying to describe this in football talk - you don't have very many games where a team has an xG of 4.7 for example.  Yet we all have witnessed games where we would have said "they were lucky to only concede 4", or, "we could have had another three or four".  The argument would be that the team has outperformed their xG and that is how you use the stat, but it still doesn't describe the game very well.  Data is only useful if you can do something with it - in football, how can I use that stat to improve my team.  I'm just not sure it does that, based on limited research to date.  It seems to have been adopted because it sounds fun.

Data patterns… you don’t see xG of 4.7 because the pattern frequency doesn’t exist - it’s an outlier, an anomaly - data analysis in football is all about data replicating itself to create a patter or view of how a player, plays… it form a persons ‘modus operandi’… we look at a player and call him a fox in the box or a poacher because of this pattern of play… the problem with humans is that we are swayed about one bad habit that distract everything else that makes a player good.. so someone who has natural ability in 9out of 10 things can be side lined because the one bad habit they have doesn’t fit with the ‘norm’ and because of that we may perceive that actually they think he’s only good with 4 things out of 10.  Using stats dismisses that bias and solidly confirms both positive and negative patterns making it easier to spot potential.

Do you remember when cars used to be unreliable and you only had oil, temperature and battery monitors - well data changed all that because now it’s possible to detect a car has a problem before you even notice it..

[RobertT]

I'm not arguing against the use of data, just that one particular stat :-), for now anyway.

If a player takes 10 penalties in a season, and then didn't take any other shots, they'd have an xG for the season of about 7.6 to 7.9.  Clearly, a good penalty taker would actually get more like 9 goals, but comparing actual goals to the xG and saying the player has over performed is disingenuous at best.

It would also not be fair to compare that players xG to someone else who took 40 shots in the season from a variety of situations creating the same xG of 7.6.  Especially if that player did not take any penalties. 

Yet, somehow, that single stat has been latched onto as some sort of catch-all to show how well or badly a team is performing, not just a player.

It may well serve a useful purpose when combined with a myriad of other data points (like the anecdotal ones listed above), but I worry when people, especially those running the football clubs, start to rely on it (it seems).

[Pericardinho]

XG, just like any other stat, needs to be used with context and some common sense.

I seem to remember Rochdale last season were in the top 6/7 for XG for quite a while. I thought to myself they'll become clinical eventually and climb up the table. They never did.

[Jimmy HaveHave]

The best stat of all is the points total at the end of the season it never lies!

[jayohaitchenn]

The best stat of all is the points total at the end of the season it never lies!

What about the season when we won the league on PPG?

[Jimmy HaveHave]

What about the season when we won the league on PPG?

Exceptional circumstances that season and hopefully a one off.

[@MacPhlea]

Are any TEF'ers actual professional Data Analysts?

I'm a data scientist...