Rolling Premier League table

[stu0x]

But unlike the actual league table it projects what things might be like at the finish, using previous data of similar events last year to fill in the blanks of the remaining 29 games in combination with what we know from the opening 9 games.

But that's not how a projection works. In fact, it's the complete opposite. Your statistical calculation is weighted in favour of older, less relevant data, instead of newer, more relevant data. That's why you think it becomes 'more accurate as time goes on' - because the inaccuracy that you've built into the model becomes further marginalised as time goes on. It doesn't become more accurate at predicting at all - it becomes less inaccurate at recording.

 

What you should be doing is pro-rating the deviance of this year's fixtures against last year's and then applying that modifier against this year's remaining fixtures. You could 'tune' the deviation in a million different ways - modify it for form, time of year, etc - but it would then actually be a projection.

 

What you've done is essentially say 'lets work out the difference between last year and this year - and then completely ignore it'. So it's basically worthless.

 

You've just been MLG'd

[Mystic Force]

Oh you people depress me some times.

 

Is this a definitative tool for identifying a final position? No, it is a model, and with all models there have to be some assumptions, the assumption here is that each team will perform similiarly to how they did last year and then give a comparative position. This obviously breaks down when teams do dramatically different than last year, and they would then be given a projected finish way off if THIS YEARS form is dramatically different, however to rule out this approach makes the assumption that thing will carry on as they are this year and FORM TO DATE will continue for the entire year, this simplifies by suggesting all future matches will be like the previous ones, except they are against different teams and different venues which do effect results. All models use an assumption that the past can help indicate the future, but on that basis we would end up with last years table, obviously not helpful.

 

The most interesting observation I find from it is that Arsenal look to be doing well in the league table but comparitively badly when judged against last year, so which Arsenal will we have come the end of the year? The two anamolous results in the final table right now would appear to be Southampton and Man Utd, who are having dramatically different seasons than last year, so far.

 

The way I look at this model is similair to the concept of Par in golf where par is how we did last year. Different models tell us different things, no one way of looking at the data will tell us conclusively the shape of the future but a snapshot of what it might look like given this set of assumptions. A different set of assumptions tells us a different snapshot. With all these analysis methods the accuracy improves with more data. What it does say is we are doing much better than last year.

 

I am not sure if you are all on a wind up becuase it is MLG or if you were not understanding its usefulness and limitations. I am a scientist and like fitting data to models so I find this fun. I can understand why others do not trust this approach.

 

Lies, damn lies and statistics. But is there a better way of way of measuring stuff than statistics? Most bad statistics are used by liers wanting to prove a point which is not valid by cheery picking the statistics or using misleading data sets.

 

Ah you people depress me, back to fitting real data to models...

 

- - - Updated - - -

 

Oh you people depress me some times.

 

Is this a definitative tool for identifying a final position? No, it is a model, and with all models there have to be some assumptions, the assumption here is that each team will perform similiarly to how they did last year and then give a comparative position. This obviously breaks down when teams do dramatically different than last year, and they would then be given a projected finish way off if THIS YEARS form is dramatically different, however to rule out this approach makes the assumption that thing will carry on as they are this year and FORM TO DATE will continue for the entire year, this simplifies by suggesting all future matches will be like the previous ones, except they are against different teams and different venues which do effect results. All models use an assumption that the past can help indicate the future, but on that basis we would end up with last years table, obviously not helpful.

 

The most interesting observation I find from it is that Arsenal look to be doing well in the league table but comparitively badly when judged against last year, so which Arsenal will we have come the end of the year? The two anamolous results in the final table right now would appear to be Southampton and Man Utd, who are having dramatically different seasons than last year, so far.

 

The way I look at this model is similair to the concept of Par in golf where par is how we did last year. Different models tell us different things, no one way of looking at the data will tell us conclusively the shape of the future but a snapshot of what it might look like given this set of assumptions. A different set of assumptions tells us a different snapshot. With all these analysis methods the accuracy improves with more data. What it does say is we are doing much better than last year.

 

I am not sure if you are all on a wind up becuase it is MLG or if you were not understanding its usefulness and limitations. I am a scientist and like fitting data to models so I find this fun. I can understand why others do not trust this approach.

 

Lies, damn lies and statistics. But is there a better way of way of measuring stuff than statistics? Most bad statistics are used by liers wanting to prove a point which is not valid by cheery picking the statistics or using misleading data sets.

 

Ah you people depress me, back to fitting real data to models...

[hypochondriac]

And another example of this rubbish multiquote system.

[Daren W]

Interesting that Arsenal got 27 points from their corresponding fixtures last season!

 

And even more interesting considering that they got slated for not buying big and then surprise everyone by their start to this season... which is not as good as last year...

Odd...

[Window Cleaner]

And even more interesting considering that they got slated for not buying big and then surprise everyone by their start to this season... which is not as good as last year...

Odd...

 

their start is in fact better but against the teams they've already played so far this season they've picked up less points. Last season they had 15 points from their first 9 games,4 wins,3 draws, got beaten by Norwich and Chelsea.

[Matthew Le God]

Updated table after 20 Premier League games...

 

[trousers]

Looking good

[syd_barrett_saint]

Looking good

 

It'll look even better if we beat West Brom.

 

Sent from my GT-I9190 using Tapatalk

[Hatch]

I think Arsenal will get more than 25 points from their last 18 games tbf

[Saint86]

Updated table after 20 Premier League games...

 

 

Newcastle and crystal palace should do far far better than predicted here, but an interesting share so thank you.

[nta786]

Interesting how Everton doing only one point better than last season.

[Dalek2003]

Think we will need a bit more than 37 points for safety.

[SaintBobby]

Think we will need a bit more than 37 points for safety.

 

yeah, but you kind of got that spectacularly wrong last season too....

[nta786]

MLG, Rolling Table Update?

[Coxford_lou]

MLG, Rolling Table Update?

 

You can follow this guy on Twitter:

 

@SimonGleave: Biggest declines compared to equivalent fixtures last season: Everton -20, Liverpool -16, Palace -13, West Brom -8: http://t.co/2T3hxs93Ws