I have some code that generates a list of paired of values for various buy/sell strategies. The expected (predicted) value is generated at the beginning of the trading period for each period in the give time period being tested. The actual value is captured at the end of each trading period. The values are paired for each time period and are represented as percentage changes-- the predicted percentage change and the actual percentage change. I have generated these paired list for several different time frames for several different strategies. I am wondering if there would be value in exploring the relationship between the predicted values and the actual values using these lists. Would the presence of a relationship (or absence of one) indicate anything meaningful about the strategy being tested for the given time period. If so, which statistical tests would be the most informative to use? (cointegration, one of the correlation tests, t test, wilcoxon test, etc). What kind of confidence levels should I be looking for? How many data points would be needed to make a meaningful test? Any and all feedback would be greatly appreciated.