draw inflation tool

[joneyyy1981]

i read about this the other day, now i see it has a special offer with the geek..

anyone tried this, looks very interesting.. seen plenty of green screenshots, but worries me more is the lack of red ones..

anyone had a crack yet?

[WhyAlwaysMe]

Think I might join this. You can do a trial for £14 according to the website.

Best to wait a few weeks though! I'm not joining with the expectation of making money but to hopefully learn something which could lead somewhere.

[doubleback]

It sounds to me like they are using a diagonally inflated bivariate model ...

"The majority of football models are based on the poisson distribution and so is the PMT tool - a heavily modified poisson model which allows for teams correlation and leagues specific corrections.
Historically poisson models have been predicting draw outcomes to occur less often and the error has been around 10%, therefrom the draw inflation - inflating/increasing the probability of draws and of course draw scores 0-0/1-1 etc."

Source linky


Funnily enough, the model that I completed in Excel at the end of last season, was using this method ...

The paper that I took it from is:

Analysis of sports data by using bivariate Poisson models, Dimitris Karlis and Ioannis Ntzoufras

I would be interested in what numbers the DI guys churn out, just as a comparison to my own work.

[WhyAlwaysMe]

Sounds like you have a big tool yourself DB .

[dude67]

It sounds to me like they are using a diagonally inflated bivariate model ...

It could also be a univariate model.

I would be interested in what numbers the DI guys churn out, just as a comparison to my own work.

That's probably going to cost you a subscription.

Doubleback, what did you end up using as the inflation function? Seems to be somewhat empiric whether people are using geometric distribution, Bernoulli or whatever.

[xtrt]

but worries me more is the lack of red ones..

Of course we have red ones, nobody wins them all.
The idea is that the further the DI from the average the higher the probability of a winning trade there.

[xtrt]

It sounds to me like they are using a diagonally inflated bivariate model ...

I would be interested in what numbers the DI guys churn out, just as a comparison to my own work.

Spot on.
Diagonally inflated bivariate poisson + additional corrections which the model cannot capture (price of 2-1 to faves being one).
Please use the contact form at PMT site and we'll activate your access, PMT offers the same DI tool plus CS, o/u, half time prices etc.

[xtrt]

Anyone into football modeling is welcome, be happy to exchange ideas and improve.

[custard]

I love this "science stuff" . Please will somebody give the correct scores for tonight's games. ( No graphs or matrices, please, as they take me back to my student days and make me feel old.)

[xtrt]

I love this "science stuff" . Please will somebody give the correct scores for tonight's games. ( No graphs or matrices, please, as they take me back to my student days and make me feel old.)

There was a big move in denmark v germany 2.5 market a couple of hours ago, AUQ was available @9 for a few minutes while "science" said new price should be around 7, look at it now

[custard]

I think the Germany v Denmark game will end 1 - 0, 2 - 0 or a 1 -1 draw. If I should be correct, it will be interesting to see how the science accounts for these results.

[WhyAlwaysMe]

I think the Germany v Denmark game will end 1 - 0, 2 - 0 or a 1 -1 draw. If I should be correct, it will be interesting to see how the science accounts for these results.

That's not very sporting - you're on about 30% of the market!

[dude67]

I love this "science stuff" . Please will somebody give the correct scores for tonight's games. ( No graphs or matrices, please, as they take me back to my student days and make me feel old.)

Strictly speaking it's not science, it's mathematics. Or statistics, to be more exact.

One of the strengths of a statistical model like this is that it can explain the correlation between markets which are basically projections of the same one market: Correct Score, Match Odds, Over/Unders, Draw No Bet, Double Chance, Both Teams to Score, Teams to Score, Odd or Even, Winning Margin, Total Goals, Home Goals, Away Goals, Handicaps and Asian Handicaps.

This can for instance be used to predict market corrections as xtrt describes, where the prices in liquid markets like O/U 2.5 will drive the prices in the less liquid markets like CS Any Unquoted.

The results are only as good as the parameters you put into the model, but using it to identify mispriced markets doesn't require any opinion on the actual odds.

[custard]

I studied statistics as part of my first year mathematics course at university. I was young and silly in those days and had much to learn about life. My initial thoughts were basically - statistics is a load of bollocks.

Now, I am much older and very much wiser. I am able to say, after all these years, that "statistics is a total load of bollocks".

[Emkayracing]

total load of bollocks

It's not just statistics.

Websites too can he summarised by the same phrase