Data Mining Approach suggestions needed

I have written scripts that perform analysis on stock market equities and then compute equity derivative (stock options) strategies that potentially are properly suited for profit. I keep my own historical data on strategy parameters and update the data over time. Some win and some lose (btw, if the strategies are held to expiration the outcome will be binary in nature). I'm trying to mine my data to better understand what initial parameters may be adequate predictors of win/lose. The strategies are all time constrained and all have discrete outcomes. Outcomes are updated throughout the duration of the strategy.

I'm at the point that I don't really know where to begin mining my present data. One thought is to implement a machine learning technique and look at every individual time step for every strategy as inputs with the final win/lose as the output. I've used TreeBagger methodology in the past, but I think I may be hindered in this approach as I presently do not have lots of target/output data. Because of the binary nature of the outcomes, it seems like a good classification problem.

Any thoughts or suggested directions?