GT versus MY versus DF versus QL

When all four algorithms play each other and with wA=0.25, wB=0.75, 6 prices, and 28 pricebots, the MY do the best of the algorithms. The QL follow the MY for awhile, but eventually drop off, see Figure 15. Although not shown in Figure 15, after 250 rounds, there are 6 GT pricebots and 22 MY pricebots; the DF and QL have disappeared from the population. Once there are only GT and MY pricebotsleft, given wA=0.25, wB=0.75 and 6 prices, like in this situation, the GT and the MY appear to stabilize at whatever proportions they were at when both the DF and QL disappeared. We were using 100,000 rounds per generation and this may not have been sufficient time for the QL to learn, and therefore may be responsible for the eventual decline in QL pricebots.

Figure 15. a. Average prices, b. Average profits and c. Number for gt, my, df, and QL pricebots starting with 7 pricebots using each algorithm originally.