Goal Difference!

[Rev Algenon Stickleback H]

GD is a very very strong indicator of the actual quality (relative quality) of sides, and "GD will out".

... but over a season GD is a great predictor.

Again, this is where you are reading far too much into stats.

Goal difference is a good indicator of teams that are probably over/underachieving. It is NOT a good predictor for final position.

Yes, the team with the third best goal difference will probably finish somewhere around third, but having the third best goal difference in December is no indicator that the team will have the third best in May.

[Mr Angry]

Actually, it's easy to say screamingly obvious.

But we could say

"The top teams score the most goals." (Therefore the "FOR" column is a good predictor)

That is in fact NOT true

In the last 10 seasons, in the Division that Reading have been in, only on 2 occasions has the team that finished top NOT scored the most goals; therefore, it IS true that generally the top teams score the most goals.

"The top teams have the best defences" (Therefore the "AGAINST" column is a good predictor)


That is in fact NOT true

In those last 10 seasons, 6 out of the 10 times the team that finished top had conceded the fewest goals. Again, generally that means that the top team concedes fewer goals.

GD is a very very strong indicator of the actual quality (relative quality) of sides, and "GD will out".

Totally agree with this; in the last 10 seasons, the division that Reading has been in has been won by the team with the best GD on 8 of those occasions. Therefore, it IS obvious to anyone that the top teams will have the best GD.

However, what IS daft is to attempt to state is that because a team has the best (or third best) GD after a (randomly chosen) number of games, that team will finish 1st (or 3rd) - which is precisely what you are trying to argue. There are far too many variables to state with any degree of certainty, after a randomly selected number of games, ON THE BASIS of GD, where any single team will finish up in the final table. On that basis, after the first game of the season, you would think "right, Team X will finish top as they have the best GD".

Football is a sport played on turf, not in an excel spreadsheet; might be worth remembering that from time to time......

[Arch]

The fact that we currently have the third best GD could be taken as an indicator that we're more likely to move up than down. Of course, it's just one among many that you could look at. The strongest correlation is between final position and points acquired. The question is will our GD move more into line with our points or will our points move more into line with our GD (or neither). Who knows?

[Ian Royal]

This suggestion is so backwards its untrue.

If we finish in the top 4 - given the way we play of course we're going to have a good goal difference, almost certainly one of the best 5 in the division. But that's because we'll have won more games than we lose, therefore making a good goal difference almost inevitable.

The fact we have a good goal difference now is because halfway through the season we're well placed in the league because we've won more games than we've lost. And the nature of our play means we'll quite often punish beaten teams in the latter part of the game, while generally we only lose by the odd goal or two when it does happen.

If we are going to finish high up that trend will continue. Our current GD doesn't predict what will happen though. We could quite easily fall apart with a bad injury or two and some poor form and slide down the table. In which case our good GD will erode significantly and it will then once again match our league placing.

[Snowball]

However, what IS daft is to attempt to state is that because a team has the best (or third best) GD after a (randomly chosen) number of games, that team will finish 1st (or 3rd) -

which is precisely what you are trying to argue.


NO it isn't


There are far too many variables to state with any degree of certainty, after a randomly selected number of games, ON THE BASIS of GD, where any single team will finish up in the final table. On that basis, after the first game of the season, you would think "right, Team X will finish top as they have the best GD".

Hmmm....

I wonder what I ACTUALLY SAID?

If we finish the season with the 3rd best GD we'll finish 3rd or 4th, probably 3rd

[Harpers So Solid Crew]

thing is SB you say we will, if you say historically we SHOULD finish 3rd or 4th then you would not get pulled up all the time. Same as 92 points should get top two, but it is not guaranteed.

[Wycombe Royal]

Two seasons ago we had the best goal difference and finished fourth. We also scored the second highest number of goals.

The season before that Bristol City finished 4th with the 12th best goal difference (+1).

All stats have to be taken with a pich of salt, and if presented by Snowball they need to be taken with a table spoon of salt.

[Snowball]

thing is SB you say we will, if you say historically we SHOULD finish 3rd or 4th then you would not get pulled up all the time. Same as 92 points should get top two, but it is not guaranteed.

If we finish the season with the 3rd best GD we'll finish 3rd or 4th, probably 3rd

Ah, I see. My bad. My statement is not meant to be "factual" just my 100% bet, my "statement". Me saying IT WILL HAPPEN.



A bit like saying, "If we win at Swansea we will finish the season in the top 4, and if we also win at Cardiff we will finish in second place."





Like anyone else, I can read and see that sides with lower GDs occasionally do well, sides with good GDs miss-out


but the predictability is still 91%... meaning 1 in 11 times it'll be less-so!


BTW it's pointless picking out individual cases. That isn't what correlation is about, unless it's a perfect correlation of 1 (=100%)

The point is OVERALL what is the relationship between GD and final position.

[Rev Algenon Stickleback H]

thing is SB you say we will, if you say historically we SHOULD finish 3rd or 4th then you would not get pulled up all the time. Same as 92 points should get top two, but it is not guaranteed.

If we finish the season with the 3rd best GD we'll finish 3rd or 4th, probably 3rd

Ah, I see. My bad. My statement is not meant to be "factual" just my 100% bet, my "statement". Me saying IT WILL HAPPEN.



A bit like saying, "If we win at Swansea we will finish the season in the top 4, and if we also win at Cardiff we will finish in second place."





Like anyone else, I can read and see that sides with lower GDs occasionally do well, sides with good GDs miss-out


but the predictability is still 91%... meaning 1 in 11 times it'll be less-so!


BTW it's pointless picking out individual cases. That isn't what correlation is about, unless it's a perfect correlation of 1 (=100%)

The point is OVERALL what is the relationship between GD and final position.

where are your stats to back up the statement that 91% of the time, the team with the 3rd best goal difference will finish third?

A very quick check showed that of the last 11 teams in the championship to have the 3rd best goal difference, only 4 finished third (couldn't be arsed to check further back than that, as http://www.rsssf.com didn't show goal difference on the league tables before that).

Four teams finished 3rd. Four also finished second. Two finished 4th and one finished 5th.

Rather more importantly, where do you offer anything at all to back up the claim that having the 3rd best goal difference in December indicates a 3rd place finish?


All you can do with goal difference is highlight teams that are probably overachieving and those that are probably underachieving. The law of averages says that overachievers and underachievers probably will even out over a season, but that's really about as far as it goes. You can't use December goal differences as predictors of final positions with anything more than blunderbuss accuracy.

This is the problem with your stats obsession. You use stats like a man carving a turkey with a cricket bat, because you don't stop to think how illogical the conclusions you draw from them are.

[Snowball]

where are your stats to back up the statement that 91% of the time, the team with the 3rd best goal difference will finish third?

I didn't say that. A correlation of .91 more-or-less means you can say X maps to Y with an accuracy of 91%

IT IS NOT A PREDICTION FOR SINGLE TEAMS or single instances. THAT'S NOT HOW CORRELATION WORKS.

A very quick check showed that of the last 11 teams in the championship to have the 3rd best goal difference, only 4 finished third (couldn't be arsed to check further back than that, as http://www.rsssf.com didn't show goal difference on the league tables before that). Four teams finished 3rd. Four also finished second. Two finished 4th and one finished 5th.

So what you are saying is that the third-best GD predicted the finishing position within a single place for 10/11 (91%) of the teams and was only 2 places out for the last club. Oh, and Looooook. I said the correlation is 91%. Correlation is not an absolute predictor. It doesn't even necessarily mean causality. But 4 x 3 (12), plus 4 x 2 (8) plus 2 x 4 (8) + 1 x 5 (5) total 33. Divide by 11 teams = 3rd.

People picking single instances of teams means NOTHING. Correlation is about trends and predictability

However you have pretty much proven the point I was making. Even someone from HNA should understand that just occasionally teams have an odd set of results and buck a trend, or that one very bad defeat or huge win can distort things... but the generality. over the whole season is highly, highly predictable

Rather more importantly, where do you offer anything at all to back up the claim that having the 3rd best goal difference in December indicates a 3rd place finish?

That is not what I said.

If we finish the season with the 3rd best GD we'll finish 3rd or 4th, probably 3rd

can't use December goal differences as predictors of final positions with anything more than blunderbuss accuracy.

I didn't

This is the problem with your stats obsession. You use stats like a man carving a turkey with a cricket bat, because you don't stop to think how illogical the conclusions you draw from them are.

or you could learn to read?

[Svlad Cjelli]

Does anyone have any stats on how many times Snowball has turned abusive when someone has challenged his use of stats?

[Archie's penalty]

Snowball is a goon. But he's the only person investing any time into posting on the team board anymore. So kudos to him.

[Svlad Cjelli]

He's obviously never heard the phrases "less is more" or "quality over quantity".

Frankly there's so much data being presented that none of it is effective - over-communicate and you're not communicating any more.

Any why do you think hardly anyone else puts any time into posting here any more? Any opinion that Snowball doesn't agree with gets submerged under a sea of numbers.....

[Archie's penalty]

He needs to post a statistic and let everyone argue with or against it, not weigh in.

Perhaps we could create a snowball thread and all of his statistics junk could be thrown in there? Make it a sticky and then we could confine him to his statistics 'ghetto.'

[weybridgewanderer]

problem with stats is that you can, make them tell you anything you want and Snowball frequently does


For example, 20% of road traffic accidents are caused by people drinking and driving

That means 80% are caused by sober people so i say ban the sober people!

The telling point missing form this thread is the point "how often does the team with the 3rd highest goal difference, sitting 6th/7th, in december finish in the top 6"
Snowball provide to us statistically we would get automatic promotion 2 seasons ago...

[Svlad Cjelli]

problem with stats is that you can, make them tell you anything you want and Snowball frequently does


For example, 20% of road traffic accidents are caused by people drinking and driving

That means 80% are caused by sober people so i say ban the sober people!

The telling point missing form this thread is the point "how often does the team with the 3rd highest goal difference in december finish in the top 6"
Even them Snowball provide to us statistically we would get automatic promotion 2 seasons ago...

Even better is the stat saying that when the wearing of tin helmets was introduced in WW1 the number of head injuries increased.

That was, of course, because before that most shots to the head killed people, but if you look at that single statistic in isolation, as Snowball often does, it's easy to argue the case that wearing tin helmets increases the number of head wounds.

[weybridgewanderer]

did you know that, statisically, the number of road accidents around speed cameras has INCREASED!

ban them I say!

[Rev Algenon Stickleback H]

So what you are saying is that the third-best GD predicted the finishing position within a single place for 10/11 (91%) of the teams and was only 2 places out for the last club. Oh, and Looooook. I said the correlation is 91%. Correlation is not an absolute predictor. It doesn't even necessarily mean causality. But 4 x 3 (12), plus 4 x 2 (8) plus 2 x 4 (8) + 1 x 5 (5) total 33. Divide by 11 teams = 3rd.

"within a single place"

A team with the third best goal difference is highly likely to finish somewhere between 2nd and 4th, which includes 2nd and 4th places. The fact that it "averages out to 3rd" means nothing.

I checked again for League 1&2 for the same period.

Collating them all gives

2nd - 11 / 37
3rd - 12 / 37
4th - 8 / 37
5th - 4 / 37
7th - 2 / 37

so yes, the most likely final place of a team with the third best goal difference is third, but it only happens 32% of the time.

In other words, the team with the third best goal difference is more than twice as likely to NOT finish third than it is.

That is why saying having the third best goal difference means we'd probably finish third is utter crap. The team with the 3rd best goal difference is more likely to finish 3rd than any other individual position, but probably will NOT finish third.

You don't understand the fundamental difference between something been the most likely outcome, and something being the probable outcome.


For example, if you had a doctored roulette wheel, where the number 15 was wider than the other numbers, there would be a higher probability of 15 being the number the ball rests in than any other number. It would be the most likely number, but you could not say that the ball will probably end in 15 as that's just wrong.

However you have pretty much proven the point I was making. Even someone from HNA should understand that just occasionally teams have an odd set of results and buck a trend, or that one very bad defeat or huge win can distort things... but the generality. over the whole season is highly, highly predictable

Yes, but you are presenting it as some kind of revelation, while most are expressing a "no shit, sherlock" response.

That is not what I said.

It's not, but it's kind of the crux of the whole thread. We have the third best, and you claim that if we maintain that we'll finish 3rd.

That's a rather big if to gloss over.

And that's also the fundamental point of predictive statistics - looking at the situation now to try and predict future events.

Now either...

a) the fact we have the 3rd best goal difference was pure coincidence to you starting this thread, and you made no connection between our current GD and a hypothetical one in May - which in itself would be pretty stupid as you might as well just say "if we have 75 points in May then we should be in the play-offs". i.e a pointless hypothesis.

or

b) you were taking it as probable that we'd have the third best goal difference, but because you implication wasn't written down, you are now denying it in an attempt to move the goalpost.

or you could learn to read?

see "b"

[Snowball]

Does anyone have any stats on how many times Snowball has turned abusive when someone has challenged his use of stats?

Rarely. I love DEBATE but react to ABUSE

[seahawk10]

Does anyone have any stats on how many times Snowball has turned abusive when someone has challenged his use of stats?

Rarely. I love DEBATE but react to ABUSE

Self abuse > all

why is everyone picking one snowball these days?