Beat the Mathematical Model

[AlienWoodward]

Are you accusing Darren of a Militant Third World Lesbian past Alien?


More interestingly......RADIO LUXEMBOURG!!


Grumpy, Broomhill, Trigger - your Hour is nigh.

Nope not me. You try arguing with him

 

[Highbury_Blade]

I'm happy to do it til the end of the season too. At the very least I could end up with some sound dietary advice.

 

[mattbianco1]

The chances are that it would be more accurate than Darren's, Matt B's and HH predictions, but there is no guarantee.

Hang about!! I've not said anything yet.

Here goes...

Blackpool v Ipswich D
Bristol C v Doncaster HW
Cardiff v Middlesbrough HW
Palace v SUFC D
Derby v Watford HW
Newcastle v Barnsley HW
Forest v Swansea HW
Peterborough v Coventry D
Plymouth v Preston D
QPR v WBA AW
Sheff Wed v Leicester AW.

I'm going to put 3 acc's on this week. Mine, darrens and ollys. If Olly wins me a grand i'll take everything back and give you a big sloppy kiss!

Put 3 accumilators on:

Mine - £15,300 return
Darren - £12,900 return
Olly - £28,100 return.

All off quids.

You may wanna pay attention to the next bit.

IF Darren's or Olly's comes in, i will give the relevant person £1000 as a token of gratitude. 100% i will keep my word.

I am that confident that none of us will win! In which case you both owe me a quid!:thumbup:

 

[mattbianco1]

I have delved into my sock draw 11 times, and pulled out a sock for each result. A white sock is a home win, a black sock is an away win and anything else (red sock, packet of crisps, condom) is a draw.

Why have you got white socks? :thumbdown:

do they have two rings around them?

[YOUTUBE]b9lp7q4e9AU[/YOUTUBE]

 

[Ollessendro]

I am quite happy to keep doing this until the end of the season to get a broadly based result but I suspect everyone else might get bored.

I suspect that we have bored people to tears with our debate already!?

No, several people have said that they found this debate interesting. We could try to keep it fun for everbody by making a bit of a league table, or keeping a points system of how people score (bit of irony in there somewhere). this might be a bit time consuming but it will help to analyse predictive power of people vs the model. Additionally, anyone that is bored simply doesn't have to read the posts/look at the thread. Simple!

 

[Pinchy]

Are you accusing Darren of a Militant Third World Lesbian past Alien?


More interestingly......RADIO LUXEMBOURG!!


Grumpy, Broomhill, Trigger - your Hour is nigh.

Fabulous 208. Those were the days. My introduction to 'progressive' music came at the age of 12 at the hands of Kid Jensen, late at night, in bed, with the light off and the radio turned down so my parents would not hear him.

 

[Selly]

Well if its going to be a competition...

Blackpool v Ipswich HW
Bristol C v Doncaster D
Cardiff v Middlesbrough D
Palace v SUFC AW
Derby v Watford HW
Newcastle v Barnsley HW
Forest v Swansea D
Peterborough v Coventry AW
Plymouth v Preston AW
QPR v WBA D
Sheff Wed v Leicester D.

 

[mattbianco1]

We could try to keep it fun for everbody by making a bit of a league table, or keeping a points system of how people score (bit of irony in there somewhere).

I thought this as well. Maybe something Linz could set-up?? I know theirs a prediction table on the main site.

3pts for a "Nailed on score"
1pt for correct result i.e. HW/AW/D (but incorrect scoreline)
Nil-pwa if your wrong totally.

 

[Selly]

We could try to keep it fun for everbody by making a bit of a league table, or keeping a points system of how people score (bit of irony in there somewhere). this might be a bit time consuming but it will help to analyse predictive power of people vs the model. Additionally, anyone that is bored simply doesn't have to read the posts/look at the thread. Simple!

Bear in mind that if Darren and Walthy get going it will end up no fun at all!!!

 

[Linz]

I thought this as well. Maybe something Linz could set-up?? I know theirs a prediction table on the main site.

Foxy's the man to talk to for that one but he's got a lot on with work this week.

I'll point him in this direction though.

 

[Darren]

Hang about!! I've not said anything yet.

Here goes...

Blackpool v Ipswich D
Bristol C v Doncaster HW
Cardiff v Middlesbrough HW
Palace v SUFC D
Derby v Watford HW
Newcastle v Barnsley HW
Forest v Swansea HW
Peterborough v Coventry D
Plymouth v Preston D
QPR v WBA AW
Sheff Wed v Leicester AW.

I'm going to put 3 acc's on this week. Mine, darrens and ollys. If Olly wins me a grand i'll take everything back and give you a big sloppy kiss!

Put 3 accumilators on:

Mine - £15,300 return
Darren - £12,900 return
Olly - £28,100 return.

All off quids.

You may wanna pay attention to the next bit.

IF Darren's or Olly's comes in, i will give the relevant person £1000 as a token of gratitude. 100% i will keep my word.

I am that confident that none of us will win! In which case you both owe me a quid!:thumbup:

I will look forward to my grand

Interesting that my prediction gives you the lowest win which presumably means that my prediction correlates most closely with the bookies' odds.

 

[mattbianco1]

I will look forward to my grand

Interesting that my prediction gives you the lowest win which presumably means that my prediction correlates most closely with the bookies' odds.

I hope i owe you £1000 by saturday evening. (never thought i'd say that!)

 

[RichJay]

My basic scepticism about all alleged mathematical models is that, as a matter of logic, they cannot possibly work in the long run. Firstly because, if they did work all astute gamblers would used them and the bookmakers would all be bankrupted (or bookies would stop taking bets on football). Secondly, I rather suspect that William Hills, Ladbrokes etc have enough dosh to employ people at least as clever as Olly's friend to work out their odds and ensure that they are not out-foxed by these alleged models.

These are statistical models that are being updated constantly so there is no reason to believe that the can't work in the long run. Actually one of the most interesting problems in sports betting models is being able to update models quickly enough so that odds can be calculated for things that happen in play. As I mentioned in the other thread, the paper that Olly provided is by Mark Dixon and Stuart Coles. Mark Dixon now owns ATASS sports (http://www.atassltd.co.uk/about.htm) which is a sports betting company. Stuart Coles works for a similar company in London (sorry can't remember the name) which I suspect is the same one that Olly's friend used to work for.

 

[Highbury_Blade]

QPR v WBA My movements say qpr wanting to play up for the new boss. (3) HW

 

[Ollessendro]

Interesting that my prediction gives you the lowest win which presumably means that my prediction correlates most closely with the bookies' odds.

but we all know that the bookmakers don't always get it right!?

 

[Darren]

These are statistical models that are being updated constantly so there is no reason to believe that the can't work in the long run. Actually one of the most interesting problems in sports betting models is being able to update models quickly enough so that odds can be calculated for things that happen in play. As I mentioned in the other thread, the paper that Olly provided is by Mark Dixon and Stuart Coles. Mark Dixon now owns ATASS sports (http://www.atassltd.co.uk/about.htm) which is a sports betting company. Stuart Coles works for a similar company in London (sorry can't remember the name) which I suspect is the same one that Olly's friend used to work for.

But as soon as a statistical model starts beating the bookies' odds, the bookies will become aware of this and alter their odds accordingly. If they didn't they would either go out of business or stop taking bets on the sport in question. Obviously, bookies will not continue sustaining losses in a particular area indefinitely.

That's why these models can never work in the long run.

 

[RichJay]

But as soon as a statistical model starts beating the bookies' odds, the bookies will become aware of this and alter their odds accordingly. If they didn't they would either go out of business or stop taking bets on the sport in question. Obviously, bookies will not continue sustaining losses in a particular area indefinitely.

That's why these models can never work in the long run.

But how many people are capable of using these models or have the time to do the calculations. The bookies will already be using similar types of models to calculate their odds, in fact I think that ATASS sell odds to some bookies.

 

[Darren]

But how many people are capable of using these models or have the time to do the calculations. The bookies will already be using similar types of models to calculate their odds, in fact I think that ATASS sell odds to some bookies.

Thats my point. The bookies will almost certainly have more resources than the vast majority of punters so will be generally be one step ahead of any system applied by the said punters.

 

[Ollessendro]

But as soon as a statistical model starts beating the bookies' odds, the bookies will become aware of this and alter their odds accordingly. If they didn't they would either go out of business or stop taking bets on the sport in question. Obviously, bookies will not continue sustaining losses in a particular area indefinitely.

That's why these models can never work in the long run.

Simply wrong Darren. You are very narrow minded in this issue.

Thats my point. The bookies will almost certainly have more resources than the vast majority of punters so will be generally be one step ahead of any system applied by the said punters.

You have to understand that the people who do this are few and far between. For starters you need an up to date model. You can't just dream one up or find on the net. Secondly you need statistical/probability expertise. Thirdly you need vast sums of money. These people probably make up less than 1% of the total betting market. The bookies know about them, don't like them and try and stop them. But all in all it doesn't matter. Why? because you, I, Matt B, Foxy, Linz, Swiss, Joe Blogs and whomever else will be betting. We don't have the time, knowledge or resources to beat the bookie. Hence they make (vast sums of) money of the general gambler. This is an important point you keep overlooking.

Oh and my friend has sent me his predictions for the weekend. Drum roll please ..........

 

[Darren]

Simply wrong Darren. You are very narrow minded in this issue.



You have to understand that the people who do this are few and far between. For starters you need an up to date model. You can't just dream one up or find on the net. Secondly you need statistical/probability expertise. Thirdly you need vast sums of money. These people probably make up less than 1% of the total betting market. The bookies know about them, don't like them and try and stop them. But all in all it doesn't matter. Why? because you, I, Matt B, Foxy, Linz, Swiss, Joe Blogs and whomever else will be betting. We don't have the time, knowledge or resources to beat the bookie. Hence they make (vast sums of) money of the general gambler. This is an important point you keep overlooking.

Oh and my friend has sent me his predictions for the weekend. Drum roll please ..........

You speak for yourself. I won't be betting. It's a mug's game, as my old mum used to say

As for the general point, I remain sceptical, but think it is a good idea to test how good this model is for every set of fixtures from now until the end of the season. Every Monday I am prepared the post the fixtures for the forthcoming Championship weekend and predict them on the basis of my strictly half arsed non mathematical model outlined above. Your friend can predict them using his model as can anyone else who cares to join in and we can do a prediction league table at the end of the season.

I think its simplest to just predict HW, AW or D without predicting actual scores and give people one point for each one they get right.

If you and your friend are correct in their estimate of the model you would expect him to easily have most points at the end of the season wouldn't you?

 

[Ollessendro]

If you and your friend are correct in their estimate of the model you would expect him to easily have most points at the end of the season wouldn't you?

If you've been following this thread and reading my posts/replies then you'll know the answer to this question. But yes I would expect his predictions to have more points than everyone else. Lucky and randomness may come into play and 13/14 sets of fixtures may not be enough, but still for the sake of argument I'd expect him to be top of the league table.

 

[Darren]

If you've been following this thread and reading my posts/replies then you'll know the answer to this question. But yes I would expect his predictions to have more points than everyone else. Lucky and randomness may come into play and 13/14 sets of fixtures may not be enough, but still for the sake of argument I'd expect him to be top of the league table.

You see that's where I have trouble with this whole argument. If luck and randomness play such a large part in football then any mathematical model must be pointless mustn't it?

Anyway, as you say, we will see come the end of the season. My extremly non-mathematical hunch is that his predictions are likely to be no more likely to be correct than everyone else's.

 

[Ollessendro]

You see that's where I have trouble with this whole argument. If luck and randomness play such a large part in football then any mathematical model must be pointless mustn't it?

Anyway, as you say, we will see come the end of the season. My extremly non-mathematical hunch is that his predictions are likely to be no more likely to be correct than everyone else's.

Nope, the point is that over time his predictions will be better. Some could keep getting lucky and random factor might bugger my mates results up. However, the more games that are played the more his results are likely to stand out from the rest.

For a bit of fun i'm willing to put my money where my mouth is and say that his results will predict better results than other peoples. I'm confident in the model and his judgements.

 

[Foxy]

These are statistical models that are being updated constantly so there is no reason to believe that the can't work in the long run. Actually one of the most interesting problems in sports betting models is being able to update models quickly enough so that odds can be calculated for things that happen in play.

But that's one of the big issues with modelling for prediction. If the statistical model is needed to be up to the second in play, then it moves further from a prediction and closer to a reaction. I have a fantastic system for modelling football results, it may however be a rebadged version of sky sportscentre at 8pm Saturday night

This is a model used to change in-play odds, but isn't really much use to the punter, as if it's being implemented correctly at the bookies - the punter cannot place the bet as the system has already suspended betting or updated the odds

If it was possible to predict accurately with a system, football would be a boring sport. It isn't, there are too many variables. What is interesting however, is the level of "accuracy" you can place on a prediction based upon opinion and the statistical "probability" from previous results - and at what distance from the event finishing these maintain that level of "accuracy".

My "system" allowed me to win some goods on Monday night based purely on chance (the repeated tossing of a coin), this doesn't mean my system can accurately model the results, it means I guessed correctly based upon a stupid superstition. If the coin had landed the other side up half way through, it wouldn't have been a matter of under/over performing on the coins part Whilst football is a game undoubtdly majorly effected by skill and performance, this was a game with less outside influence and less chance of random events.

 

[RichJay]

But that's one of the big issues with modelling for prediction. If the statistical model is needed to be up to the second in play, then it moves further from a prediction and closer to a reaction. I have a fantastic system for modelling football results, it may however be a rebadged version of sky sportscentre at 8pm Saturday night

This is a model used to change in-play odds, but isn't really much use to the punter, as if it's being implemented correctly at the bookies - the punter cannot place the bet as the system has already suspended betting or updated the odds

If it was possible to predict accurately with a system, football would be a boring sport. It isn't, there are too many variables. What is interesting however, is the level of "accuracy" you can place on a prediction based upon opinion and the statistical "probability" from previous results - and at what distance from the event finishing these maintain that level of "accuracy".

My "system" allowed me to win some goods on Monday night based purely on chance (the repeated tossing of a coin), this doesn't mean my system can accurately model the results, it means I guessed correctly based upon a stupid superstition. If the coin had landed the other side up half way through, it wouldn't have been a matter of under/over performing on the coins part Whilst football is a game undoubtdly majorly effected by skill and performance, this was a game with less outside influence and less chance of random events.

I wasn't meaning up to the second with play more updating odds at half time. To be honest I don't know a great deal about what you can bet on during a game. I do however know quite a bit about probability and statistical models and I've been to a couple of seminars given by one of the authors of the paper from the other thread.


This gives a good description of what the model is all about...

http://www.tech.plym.ac.uk/Research/statistics/overview/seminars.htm

Sports events provide a rich source of statistical modelling problems that can be used to trade on newly-developed betting exchanges such as Betfair. These exchanges work in a very similar manner to standard financial markets, and traders require sophisticated models based upon a detailed knowledge of the sport in question. For example, Dixon and Coles (Applied Statistics, 1997) set the standard for modelling football scores using a Poisson-based generalised linear model to determine the outcome of future matches. The model assigns defence and attack scores to each team, as well as accounting for other factors such as home-advantage when predicting match outcomes. This basic model was extended by Karlis and Ntzoufras (The Statistician, 2003) who introduced the bivariate Poisson model to overcome dependencies between team performances in low-scoring matches. The aim of these models is to provide an accurate assessment of the probability of different match outcomes that can be compared with the “market view” to determine whether or not to trade and, if so, at what price.

 

[happyhippy]

Nope, the point is that over time his predictions will be better. Some could keep getting lucky and random factor might bugger my mates results up. However, the more games that are played the more his results are likely to stand out from the rest.

I think most have missed the point of the system, which is to provide an edge when assessing the relative odds. It's not meant to be used as a crystal ball. It doesn't take much working out to see that 'unbeaten, having only conceded 8 at home all season' Newcastle should see off the Dingles at the weekend. I'm sure that the system and everyone else's predictions will say that they should win.

Is it a prediction by the system? In the sense that it suggests the most likely outcome statistically from the data it is given, I suppose it has to be. Does it identify a sure bet? Does it identify a value bet? Depending on your judgement and personal criteria, it might do.

Taking the above match as an example, in my opinion the 4/9 which can be had about Newcastle is way too short for me to bother with, but it looks a rock solid bet for the big hitters who can smash it to pieces.

What it doesn't do is definitively say that "x versus y = result n".

For a bit of fun i'm willing to put my money where my mouth is and say that his results will predict better results than other peoples. I'm confident in the model and his judgements.

So now we have the nub and gist of it. His opinion is a defining feature of how you observe the model. Randomness

Put it this way. Does he bet on every outcome he inputs? No, of course he doesn't. That's why this exercise and the previous one are both pointless and simply 'pedantry in motion'. If a bit of a laugh as well

Most of us could look down the betting coupon and predict the right result in a large number of matches. It doesn't mean that we'd win by gambling on them (remember the model is about gambling!).

 

[Ollessendro]

Here are my friends (we'll call him the Ginger statistical warrior) models predictions from his model. It doesn't give absolutre predictions but goals scored and concede by each time. I pushed him for a home, awy or draw prediction but he just said 'you have my model predictions as to what the scores should be, the rest is down to luck on the day'

Blackpool 1.75 Ipswich 1.40 D
Bristol City 1.44 Doncaster 1.12 D
Cardiff 1.43 Middlesbrough 1.04 H
Crystal Palace 1.21 Sheffield United 1.09 D
Derby 1.71 Watford 1.04 H
Newcastle 1.77 Barnsley 0.86 H
Nott'm Forest 1.02 Swansea 0.73 D
Peterboro 1.68 Coventry 1.13 H
Plymouth 1.38 Preston 1.12 D
QPR 1.32 West Brom 1.78 A
Sheffield Weds 1.27 Leicester 1.25 D

For the sake of argument and to keep Darren happy I have interpeted his results to give a H, A or D scenario. The more I think about it the more futile this excercise becomes. Still with all the hype around it I'll go ahead.

 

[Darren]

Here are my friends (we'll call him the Ginger statistical warrior) models predictions from his model. It doesn't give absolutre predictions but goals scored and concede by each time. I pushed him for a home, awy or draw prediction but he just said 'you have my model predictions as to what the scores should be, the rest is down to luck on the day'

Blackpool 1.75 Ipswich 1.40 D
Bristol City 1.44 Doncaster 1.12 D
Cardiff 1.43 Middlesbrough 1.04 H
Crystal Palace 1.21 Sheffield United 1.09 D
Derby 1.71 Watford 1.04 H
Newcastle 1.77 Barnsley 0.86 H
Nott'm Forest 1.02 Swansea 0.73 D
Peterboro 1.68 Coventry 1.13 H
Plymouth 1.38 Preston 1.12 D
QPR 1.32 West Brom 1.78 A
Sheffield Weds 1.27 Leicester 1.25 D

For the sake of argument and to keep Darren happy I have interpeted his results to give a H, A or D scenario. The more I think about it the more futile this excercise becomes. Still with all the hype around it I'll go ahead.

Perhaps I am being incredibly thick again here, but does his model predict that Blackpool will score 1.75 goals and Ipswich 1.4 goals? If this is right (and apart from the obvious retort that you can't score fractions of goals) how does that translate to a draw when 1.43 for Cardiff and 1.04 for Middlesbrough translates as a home win?

And isn't the statement "you have my predictions what the score should be the rest is down to luck on the day" just confirmation that the model is a self confirming circular affair? I think Sheffield United should be champions of Europe. The fact that they are not is down to luck on the day...

 

[Ollessendro]

Perhaps I am being incredibly thick again here, but does his model predict that Blackpool will score 1.75 goals and Ipswich 1.4 goals? If this is right (and apart from the obvious retort that you can't score fractions of goals) how does that translate to a draw when 1.43 for Cardiff and 1.04 for Middlesbrough translates as a home win?

And isn't the statement "you have my predictions what the score should be the rest is down to luck on the day" just confirmation that the model is a self confirming circular affair? I think Sheffield United should be champions of Europe. The fact that they are not is down to luck on the day...

Ok let's adjust it.

Again you go with your farcical comments. The model is predicts what the scores should be based on past performance, form, players la de da. But (as I've always maintained) there will be some element of randomness. That is the beauty of football Darren. You know that. If the best teams won all their games and no one could beat them it would be boring. Just think how boring the premiership was last year when Man U only dropped 2 points all season against non top 4 teams.

Here are my friends (we'll call him the Ginger statistical warrior) models predictions from his model. It doesn't give absolutre predictions but goals scored and concede by each time. I pushed him for a home, awy or draw prediction but he just said 'you have my model predictions as to what the scores should be, the rest is down to luck on the day'

Blackpool 1.75 Ipswich 1.40 H
Bristol City 1.44 Doncaster 1.12 H
Cardiff 1.43 Middlesbrough 1.04 H
Crystal Palace 1.21 Sheffield United 1.09 D
Derby 1.71 Watford 1.04 H
Newcastle 1.77 Barnsley 0.86 H
Nott'm Forest 1.02 Swansea 0.73 D
Peterboro 1.68 Coventry 1.13 H
Plymouth 1.38 Preston 1.12 D
QPR 1.32 West Brom 1.78 A
Sheffield Weds 1.27 Leicester 1.25 D

For the sake of argument and to keep Darren happy I have interpeted his results to give a H, A or D scenario. The more I think about it the more futile this excercise becomes. Still with all the hype around it I'll go ahead.

 

[Geordie Blade]

My basic scepticism about all alleged mathematical models is that, as a matter of logic, they cannot possibly work in the long run.

Your logic doesn't necessarily make sense. See below.

Firstly because, if they did work all astute gamblers would used them and the bookmakers would all be bankrupted (or bookies would stop taking bets on football).

That is illogical. Many people simply can't be bothered with 'systems', or don't believe in them (like you). You might say the same about stockbroking. There are people making vast sums of money from that - so why doesn't everyone do it?

Fortunately for the bookies, there are enough people that don't follow systems and, hence, the bookies can make profits. It's fairly difficult for the bookies to know who is using systems and, although they have techniques, there are so many systems as to make things difficult to spot. Some punters win, some lose. Most of this is pure luck, so you can't guarantee that a 'winner' must be cheating.

Secondly, I rather suspect that William Hills, Ladbrokes etc have enough dosh to employ people at least as clever as Olly's friend to work out their odds and ensure that they are not out-foxed by these alleged models.

This isn't necessarily true. The bookies don't always try to calculate the 'true' odds. They sets their odds such that they can make a profit. Think of the following example. Suppose that the 'TRUE' odds of tossing a coin are 50% heads and 50% tails. If the bookies took bets on this, and 95% of punters bet on tails (maybe based on spurious 'hunches'), the bookies would be facing a loss if it did come up tails. As such, the bookies would offer longer odds on heads in order to entice some more bets on that. If you were an astute gambler, you would spot that the long odds are offering a decent reurn to you, and you'd stick the money on. You'd be happy (remember that you're not guaranteeing that heads will win - just that you're getting good odds on it). However, the bookies would also be happy, as they'd be hedging their own outputs and will make a profit no matter what, even though the odds are favourable to you.

I think that most 'systems' don't aim to actually predict WHO will win each game. They simply aim to spot where bookies are offering slightly generous odds. In reality, these odds will only be generous to a degree of around 1.5% or 2.0% but if you do this over a long time you'll get a nice return.