Hey everyone,
Stumbled across this forum and figured I could test out a model I've been working on trying to predict soccer matches. Hope to get some useful insights.
My model uses only statistics and available information as input. A flaw in the model is that it is unable (for now) to take in abscent players into it's calculations. Anyway, this is how it works:
1. Uses a Bivariat Poisson distribution curve to assess correct score and calculates Three-way odds according to it. The two variables used is the home team's scoring capability (Average scoring rate at home together with the away team's conceded goals while away, this and last years seasons) and the away team's scoring capability (Average scoring rate while away etc...)
2. Uses a Glick rating system to determine the rating of each team in the league. Then uses the rating difference between the two teams to determine the probability of a home win/away win. (This is based on statistical data, i.e. when two teams with this difference met what was the outcome.)
3. Trend Analysis. Looking at the last 4 matches played by each team and adjusts the odds depending on which team has the best outcomes (i.e. if they are on a roll or not.)
4. Match ups. Looks at the last 4 matches these teams have played against eachother and adjusts the odds according to which team seem to have the edge.
5. Distance. It's a well-known fact that home team advantage can be quite high. This home team advantage generally goes up depending on how far the team has travelled (due to player fatigue / fewer fans cheering their favorite team on etc...). This is why Derbies have a lot lower home team advantage. Anyway, the model determines how far the away team has travelled, compares it with the average distance travelled in the leage and adjusts the odds accordingly.
6. Top team / New teams. Since all of the above are based on generalizations, one needs to account to the differing in top team performance and "bottom" team performance. If a top team or a new team is present in the match up then the odds are adjusted accordingly.
These odds have been tested on data from English Premier League/Italian Serie A/German Bundesliga/French Ligue 1/Portugese Liga Sagres/Spanish Primera Division (La Liga)/Dutch Eredivisie from the past 7 years and it seems to be performing quite well at setting odds.
So to the true testing of the model.
Each week I will run the model and take out a certain number of matches where the model predicts the home favorite odds much lower than the bookies (i.e. where the model tells me there's value to bet on the favorite). I'll run the model on these 7 top leagues and hopefully pull out around 10-15 bets each week.
This would be a great place to use Kelly staking, but I'll settle with Level staking and use: around 5% of bank.
I'll start with a bank of $100 and hopefully build on that
Start Level: $5
Wish my luck on this endeavor
Stumbled across this forum and figured I could test out a model I've been working on trying to predict soccer matches. Hope to get some useful insights.
My model uses only statistics and available information as input. A flaw in the model is that it is unable (for now) to take in abscent players into it's calculations. Anyway, this is how it works:
1. Uses a Bivariat Poisson distribution curve to assess correct score and calculates Three-way odds according to it. The two variables used is the home team's scoring capability (Average scoring rate at home together with the away team's conceded goals while away, this and last years seasons) and the away team's scoring capability (Average scoring rate while away etc...)
2. Uses a Glick rating system to determine the rating of each team in the league. Then uses the rating difference between the two teams to determine the probability of a home win/away win. (This is based on statistical data, i.e. when two teams with this difference met what was the outcome.)
3. Trend Analysis. Looking at the last 4 matches played by each team and adjusts the odds depending on which team has the best outcomes (i.e. if they are on a roll or not.)
4. Match ups. Looks at the last 4 matches these teams have played against eachother and adjusts the odds according to which team seem to have the edge.
5. Distance. It's a well-known fact that home team advantage can be quite high. This home team advantage generally goes up depending on how far the team has travelled (due to player fatigue / fewer fans cheering their favorite team on etc...). This is why Derbies have a lot lower home team advantage. Anyway, the model determines how far the away team has travelled, compares it with the average distance travelled in the leage and adjusts the odds accordingly.
6. Top team / New teams. Since all of the above are based on generalizations, one needs to account to the differing in top team performance and "bottom" team performance. If a top team or a new team is present in the match up then the odds are adjusted accordingly.
These odds have been tested on data from English Premier League/Italian Serie A/German Bundesliga/French Ligue 1/Portugese Liga Sagres/Spanish Primera Division (La Liga)/Dutch Eredivisie from the past 7 years and it seems to be performing quite well at setting odds.
So to the true testing of the model.
Each week I will run the model and take out a certain number of matches where the model predicts the home favorite odds much lower than the bookies (i.e. where the model tells me there's value to bet on the favorite). I'll run the model on these 7 top leagues and hopefully pull out around 10-15 bets each week.
This would be a great place to use Kelly staking, but I'll settle with Level staking and use: around 5% of bank.
I'll start with a bank of $100 and hopefully build on that
Start Level: $5
Wish my luck on this endeavor
Wish me luck 
