Two firms compete to provide a service using predictions from their machine learning systems. Demand for the product is confined to a fixed number of users, all of whom buy as long as they receive an offer, V-P, of at least zero. Total surplus created per transaction is equal to $500 minus (the firm's prediction error-rate) * (cost of an error), or S = 500 - p*y where p is the error rate (distinguishing from P which is price) and y is the cost of an error. Error rates are a function of the experience a firm has had; the more experience, x, the lower the error. In particular, suppose p = 200/x^2 and y = 10,000. If Firm A has xA = 100 and Firm B has xB = 70.72, then

A has a competitive advantage of about 100

A has a competitive advantage of about 200

B has a competitive advantage of about 50

B has a competitive advantage of about 100

Neither firm has a competitive advantage