Here's what you could have won!

[JJ Sefton]

Geez this is getting dull. There was little difference between United and Ipswich in the 2nd half of the season in 2009/2010. My mate rated Ipswich as better than us and the league table did not show that. The initial stats (games played, won drawn and lost) suggest that we were pretty similar (from January onwards - please don't come back to me looking a figures from before that). So this comes back to the age old argument. Randomness. Ipswich did not play to their full potential, had some bad injuries, had some bad luck and had some poor refereeing decisions given against them. These are not taken into consideration in the stats I mentioned andof course not in the league table. So we move back to square one.

No, thats not correct. Your mate made calculations in January 2010 of which team was better. The league table produced different results. Youve tried to redeem his calculations by stretching the period to encompass the entire calendar year of 2010 which, given the difference in the teams (the samples to use the jargon) in the two periods is obviously flawed. Unless, that is, your mate could predict in January that we would sign Jordan, Ertl, Britton, Doyle, Bogdanovic, Calve, Vokes...

So yes, were back to randomness and the fact that your mates tea leaf gazing cant account for, seemingly, any great degree of it. Not much of a system.

 

[Ollessendro]

No, thats not correct. Your mate made calculations in January 2010 of which team was better. The league table produced different results. Youve tried to redeem his calculations by stretching the period to encompass the entire calendar year of 2010 which, given the difference in the teams (the samples to use the jargon) in the two periods is obviously flawed. Unless, that is, your mate could predict in January that we would sign Jordan, Ertl, Britton, Doyle, Bogdanovic, Calve, Vokes...

So yes, were back to randomness and the fact that your mates tea leaf gazing cant account for, seemingly, any great degree of it. Not much of a system.

Et voila. The league table looks different to what my friends model predicted. We look at one isolated example and suddenly his tea leaves lead to nothing. I maintained all along the model cannot predict randomness.

 

[JJ Sefton]

Et voila. The league table looks different to what my friends model predicted. We look at one isolated example and suddenly his tea leaves lead to nothing. I maintained all along the model cannot predict randomness.

Not much point in it then.

 

[judge]

I like people are self depricating. You are one of my favourite posters judgey (even if you are a bit of a potty mouth).



Geez this is getting dull. There was little difference between United and Ipswich in the 2nd half of the season in 2009/2010. My mate rated Ipswich as better than us and the league table did not show that. The initial stats (games played, won drawn and lost) suggest that we were pretty similar (from January onwards - please don't come back to me looking a figures from before that). So this comes back to the age old argument. Randomness. Ipswich did not play to their full potential, had some bad injuries, had some bad luck and had some poor refereeing decisions given against them. These are not taken into consideration in the stats I mentioned andof course not in the league table. So we move back to square one.

Thanks ol lucky for you I like people who can't spell although I wouldn't say you are one of my favourite posters as you are really boring and not even accidentally
funny

 

[moDtheGod]

Not much point in it then.

But statistical models ARE useful for predicting short term results. That's how odds compilers arrive at their prices.

Look at the betting at the start of the season compared to the actual league table after 46 games and notice how the final table often bears little resemblance to the betting. Middlesbrough were pre-season favourites for last year's Championship title. Would you say that the work of odds compilers is useless? Of course you wouldn't, so what it boils down to is that this model is being used incorrectly. It can't PREDICT how the table is going to look months down the line, but it can be helpful in spotting value bets in the immediate future.

If I remember rightly, they tested the accuracy of his model on what it SHOULD be used for and on one particular weekend, it didn't do very well. But that obviously proves nothing, because one losing week is a ridiculously small sample size.

 

[JJ Sefton]

But statistical models ARE useful for predicting short term results. That's how odds compilers arrive at their prices.

Look at the betting at the start of the season compared to the actual league table after 46 games and notice how the final table often bears little resemblance to the betting. Middlesbrough were pre-season favourites for last year's Championship title. Would you say that the work of odds compilers is useless? Of course you wouldn't, so what it boils down to is that this model is being used incorrectly. It can't PREDICT how the table is going to look months down the line, but it can be helpful in spotting value bets in the immediate future.

If I remember rightly, they tested the accuracy of his model on what it SHOULD be used for and on one particular weekend, it didn't do very well. But that obviously proves nothing, because one losing week is a ridiculously small sample size.

It didnt work any better if you expand the sample size from January to May 2010. Theres been an attempt to defend it by extending the sample size further to include the whole calendar year of 2010 which, given the turnover of players halfway through, means that your samples are ruined and the attempt breaks down.

 

[SouthEssexBlade]

But statistical models ARE useful for predicting short term results. That's how odds compilers arrive at their prices.

Look at the betting at the start of the season compared to the actual league table after 46 games and notice how the final table often bears little resemblance to the betting. Middlesbrough were pre-season favourites for last year's Championship title. Would you say that the work of odds compilers is useless? Of course you wouldn't, so what it boils down to is that this model is being used incorrectly. It can't PREDICT how the table is going to look months down the line, but it can be helpful in spotting value bets in the immediate future.

If I remember rightly, they tested the accuracy of his model on what it SHOULD be used for and on one particular weekend, it didn't do very well. But that obviously proves nothing, because one losing week is a ridiculously small sample size.

In other words odds reflect form over the last few games!

 

[moDtheGod]

It didnt work any better if you expand the sample size from January to May 2010. Theres been an attempt to defend it by extending the sample size further to include the whole calendar year of 2010 which, given the turnover of players halfway through, means that your samples are ruined and the attempt breaks down.

No, you misunderstand me. I'm saying the statistical model is only valuable for the short term. Like over the coming week or two. You can't use it to predict what will happen months from now and you need to keep it constantly updated.

Olle's friend will use it to pick out his bets each weekend and then update it accordingly; what he won't be using it for is making long term predictions about who will win the league. You simply can't use it to do that, because there's no way it can predict the variance.

I'm not sure why anyone ever suggested that it COULD be used for that.

 

[JJ Sefton]

I'm not sure why anyone ever suggested that it COULD be used for that.

Youd have to ask him.

 

[Darren]

. Ipswich did not play to their full potential, had some bad injuries, had some bad luck and had some poor refereeing decisions given against them. .

Where's your evidence of this? Or is it just that this must have happened because your mate's model must be accurate?

 

[Ollessendro]

In other words odds reflect form over the last few games!

That is part of the equation. But equally they reflect where the money is being placed. They will also take into consideration factors such as home advantage, distance travelled by away teams to game etc

It didnt work any better if you expand the sample size from January to May 2010. Theres been an attempt to defend it by extending the sample size further to include the whole calendar year of 2010 which, given the turnover of players halfway through, means that your samples are ruined and the attempt breaks down.

On one team. Geez. You are basing your whole argument against this on one team over 6 months. If so, you've completely missed the point.

Where's your evidence of this? Or is it just that this must have happened because your mate's model must be accurate?

Read MoD's posts. He gets it (and is doing a pretty good job of explaining/defending it). I simply don;t have the time to gather evidence on no's injured, the times they hit the post, shots on target in games they'd lost, player values, miles travlled to game etc. So .... let's sya my mate's model was not accurate for Ipswich. There were plenty of other teams in the division.

 

[JJ Sefton]

On one team. Geez. You are basing your whole argument against this on one team over 6 months. If so, you've completely missed the point.

We can easily compare to every other team in the division.

 

[bricktop]

From a bloke who frequently posts about his wish to see people blown up by terrorists I find that a little rich.

I regard people and members of the Conservative party as two completely different entities.

 

[JJ Sefton]

I regard people and members of the Conservative party as two completely different entities.

No, you dont.

 

[bricktop]

2 days to think that up.

 

[JJ Sefton]

2 days to think that up.

About two seconds actually. Why on earth would you assume that I read your daft post the second you posted it?