Combining Analysis

[AndrewKent]

Hi folks, First off, Merry Christmas! Okay secondly, I have a question regarding combining different types of analysis. From a database of football matches I have set up I have analysed and have % probabilities of a Home Win, a Draw or an Away Win based on some of the following criteria: Based on match outcomes Form of the Home Team in the Last 6 Home Games Form of the Away Team in the Last 6 Away Games Form of the Home Team in the Last 6 Games Form of the Away Team in the Last 6 Games Form of the Home Team in this seasons Home Games Form of the Away Team in this seasons Away Games Form of the Home Team in this seasons Games Form of the Away Team in this seasons Games I have a list of other things I am calculating which include goals scored, odds offered by different book makers ect. Please remember this list is not exhaustive, I have a few more ideas but please feel free to suggest anything new that I have missed Problem is although I want to arrive at the same outcome for each measure (a % chance of a Home win, Draw or Away Win), I can't think of how to use these measures in relation to each other? i.e. I want to have an OVERALL % chance of a Home win, Draw or Away win. I thought about taking an average of all the % outcomes, which would be easy but too general, and then I thought of having some sort of ranking or points system for the outcomes based on importance but I'm not sure of order of precedence. Can anyone point me in the right direction? Either from a betting point of view or from a statistical analysis point of view? Cheers, Andy

[Datapunter]

Re: Combining Analysis I'd recommend starting with a simple rating system. Assign values to different criteria based on common sense and see if you can get close to simply predicting the average odds in the industry. From that examine each criteria in turn to attempt to measure it's validity and importance. Not an easy task but then again no-one has ever said this is easy. Be carefull with what you measure. "Form of the Home Team in the Last 6 Home Games" I take it that means Win-Draw-Lose ? If so it is an extremely dangerous measurement possibly very misleading. Assuming the team was actually the same in those last 6 matches: did they play high quality opponants or lower quality opponants. In the first case the "Form" may appear bad, in the second it may appear good. But the difference is not with the team itself. Give the exact same team the "Form" will appear differently due to the differences in opponants. Not exactly hepfull here but it is an important thing to be aware of. A simple rating system, by that i mean assigning values to criteria and then simply add up all the values to get to a relative measure, allows you to start with a list of criteria that you can examine one by one. Consider that a tool for the purpose of learning about statistical research. Ultimately whatever your prediction method ends up as it will probably need to be a little more complex than a simple rating system. But having gone through the process with the simple rating system should provide you with valueable insights on different individual criteria.

[davidc]

Re: Combining Analysis

I'd recommend starting with a simple rating system. Assign values to different criteria based on common sense and see if you can get close to simply predicting the average odds in the industry.

I don't have the knowledge or skills in Excel or Access to advise you more constructively than to say that I think that you should be trying to predict winners that will outperform the average odds in the industry rather than just predicting the average odds in the industry. Also the fewer elements you need to incorporate to achieve this the better. GL