noolnool
-
Posts
16 -
Joined
-
Last visited
Posts posted by noolnool
-
-
Re: Power Rating Soccer Predictions
I use polynomial 3rd order functions' date=' because the H and A curves are "sinusoidal", and that's exactly what I would expect. When one team is twice much better then the opposition I'd expect it to win, let's say, 50% of the time, that doesn't mean that if another is 4 times better it would win 100% but, 90 or 80% of the time. The same goes for the opposite side of the curve...and that can't be achieved with 2nd order polynomials. So it's not just for getting higher R-squared, because if I used a 2nd order function, R^2 would be around 0.96 instead of 0.99.[/quote'] I see. That makes a whole lot of sense. Thanks so much for your insightful posts! -
Re: Power Rating Soccer Predictions
No, noolnool. For example the total ratings range is from -30 to +30. Deciles (eg. percentiles) are used to divide the whole date into equal amounts-deciles into 10 (percentiles into 100).The ranges are divided this way (example!) :
(1)-30 to -20; data in this range 625
(2)-20 to -12; data in this range 625
(3)-12 to -5; data in this range 625
...
(10)+15 to +30; data in this range 625
total data =625*10=6250
It means that in the 1st range fall exactly 10% of data. Then I calculate that in this range there are for ex. 15%H; 25%D; 60%A. And I do this for all 10 ranges. I then plot the 3 curves (X-axis=rating difference ... Y-axis= %H;%d;%A) and then i apply a trendline to the data (polinomial fitting of 3rd order) and get an equation, let's say p=-0.00052R^3+0.0564R^2-0.125R+.04780
This way I get different 'p' depending on the exact rating difference 'R'. There are infinite possible probabilities for each outcome, not just 10!
I hope I was clear in my explanations. Thanks again for your interest.
Thanks! Yes, of course. I see now. What a stupid question from me..... It's interesting that you use a polinomial fitting of the 3rd order. I'm really a novice when it comes to all this stuff (but very interested) and after reading negative articles about people boasting their R2 value through using higher polinomial orders, and the possible negative side effects of that, I never go higher than the 2nd order. Maybe I have to rethink my assumptions.
-
Re: Power Rating Soccer Predictions Thanks again, robkor! I'm not that familiar with the term decile but If I understand correctly you can only have ten different Odds outcomes for each match?
-
Re: Power Rating Soccer Predictions Ok thanks, that is even better of course. Which program do you use for that purpose? Sometimes though, it's good to review the regression analysis "manually" so you can remove any outliers. These especially occur at those extreme ratings that do not have a lot of data.
-
Re: Power Rating Soccer Predictions robkor, I have another question for you. Lets say you would have started betting with your system(s) at the start of this season and the results for this season have been good and steady. Then for the next season, would you do a new regression including the new data or would you keep it as it is, since it performed well....?
-
Re: Power Rating Soccer Predictions Ok, thank you! That's a very smart way of doing things. I recently tried to do a multiple regression analysis with a few systems also based on shots, corners etc. But there really wasn't enough data available. I think it's better to do normal regression and then apply your selection methods.
-
Re: Power Rating Soccer Predictions Thanks for your reply! I fully understand you keep the finer details to yourself. Basically, I wanted to know how you implement the 3 systems as I probably don't understand it completely. Forgive me if I come accross a little bit stupid here, but if system 1 produces a match rating then you can calculate the probabilities for that match. But where then, come system 2 and 3 into play?
-
Re: Power Rating Soccer Predictions Great stuff! Very interesting. I have a question for you robkor, You say you are using 3 systems, did you use multiple regression analysis??
-
Re: Converting Elo ratings into odds
Thanks. Must take a while to learn all this right?I have an MS Access database with the match results of all major european leagues since 1993/1994. I wrote code in VBA (Visual basic for applications) to calculate the elo ratings. -
Re: Converting Elo ratings into odds How do you calculate these ratings? I mean, is there some kind of software to do it for you? It seems kind of a big task to calculate the ELO ratings for 1000's of matches manually. At least 2 weeks of work.
Power Rating Soccer Predictions
in Betting Systems & Strategy
Posted
Re: Power Rating Soccer Predictions Great to see that you're still updating this thread, Robkor! ...and with good results too. Keep it up!