Calculating Probability for Rematch in FIFA12 on Xbox Live

[15123]

Hello

I am having a hard time understanding which factors should be taken into account and which ones should not be taken into account when calculating probability.

I have the following situation:

Last night I played FIFA12 on xbox live against an unknown opponent. He challenged me to a rematch after the match was over.

I would like to calculate the chance I would have had of winning if I would have said "yes" to the rematch.
However, upon trying to calculate the chance (using Bayes' theory for example), I quickly gathered countless factors such as:

• previous personal win/loss ratio
• the fact he first scored a goal in the first 10 minutes of the first half
• the fact I scored two goals the last 20 minutes of the second half
• his predictability
• my predictability
• my fatigue at that hour; mentally as well as phyiscally
• the knowledge the opponent has about my strategy after having played a match with me
• how tired my opponent is
• the opponent's fatigue; mentally as well as physically
• if he is a night person or not

I keep thinking of factors such as these and I am not sure which ones would be valid factors for probability calculation. I am getting lost.

Does anyone have an idea on this?

Thank you

 

[MarneMath]

In the real world, we often have an extreme long list of variables that may effect an outcome. There are many methods on how to construct probability models and thus make a prediction on an outcome. The only way you are going to make a good model is if you test it repeatedly against the data you aggregate.

Once you make a model, how you constructed it, will often dictate how you can test if a variable contributed a significant amount of information or not. For example, if you choose to make a multivariate linear regression, you could use partial F test or additional sum of squares.

As a side note, I would definitely not recommend you make a Bayesian model using your current knowledge of Bayesian statistics. Bayesian models need a prior probability and finding such distribution for a parameter is will work for your model is non-trivial.

If you want to try something, you could go for a simple model and see how often you tend to win a rematch regardless of an opponent.