<p class="MsoNormal" style="text-align: justify; margin: 0in 0.5in 0pt; mso-pagination: none; mso-layout-grid-align: none;"><em style="mso-bidi-font-style: normal;"><span style="color: black; font-size: 10pt;"><span style="font-family: Times New Roman;">Genetic programming is employed to develop trading rules, which are applied to test the efficient market hypothesis. Most previous tests of the efficient market hypothesis were limited to trading rules that returned simple buy-sell signals. The broader approach taken here, developed under a framework consistent with the standard portfolio model, allows use of trading rules that are defined as the proportion of an investor’s total wealth invested into the risky asset (rather than being a simple buy-sell signal). The methodology <span style="mso-bidi-font-weight: bold;">uses average utility of terminal wealth as the fitness function, as a means of adjusting returns for risk. With data on </span>daily stock prices from 1985 to 2005, the algorithm finds trading rules for 24 individual stocks. These rules then are applied to out-of-sample data to test adaptive efficiency of these markets. Applying more stringent thresholds to choose the trading rules to be applied out-of-sample (an extension of previous research) improves out-of-sample fitness; however, the rules still do not outperform the simple buy-and-hold strategy. <span style="mso-bidi-font-weight: bold;">These findings therefore imply that the 24 stock markets studied were adaptively efficient during the period under study.</span></span></span></em></p>
On AIE-ASM: a software to simulate artificial stock markets with genetic programming