Investigate ideal moving average alpha

[peter-yuma]

• Moving average alpha determines how much the most recent score impacts the weight set.

• Low moving average alpha (which we have currently), means that very little of the weight is driven by the most recent score. This leads to slow changes in evaluation, and smoothing of rewards distribution.

• We need to investigate what the ideal moving average alpha parameter is to increase the steepness of the rewards distribution, while maintaining consistent evaluation, and not completely disregarding historical performance.