Improved Pricing on the Stock Market with Trading Agents

More documents

Recommendations

Info

e able to compare agents with different parameter settings, and results pertaining todifferent samples.I must admit I do not know why the trading agent did better than its constituentsubagents did in a falling trend. All it does is to look back at previous performance ina rather undisciplined manner, incorporating a virtual random game. Also, it did notperform well in a rising trend, just like the paradoxical game of Parrondo et al.[Parrondo et al, 2000]. Whether this could reduce volatility on the stock market as awhole or not, I cannot tell. At least it should give an incentive to people to stay on thestock market during a slump, investing in shares. This could subdue a fall of theexchange. One has to keep in mind though that the trading agent would only ownshares part of the time it is running. Furthermore, a lot of money leaves the marketdue to the cost associated with the numerous transactions the agents initiate. This isanother good reason for keeping the transaction fees of the agents at a lower levelthan ordinary transactions do.As it turned out, the cost of transaction was of paramount importance. The agentI have implemented engage in transactions frequently, typically around a hundred ayear. It would have been interesting to see how a comparatively large change of thecost of transaction would have affected the performance of the agents. Having lookedat just a few samples, my estimate is that the total cost of transaction cannot exceed afigure of somewhere in between 0.15 and 0.5 per cent for the agent to remainprofitable.Another option for people who want to stay on the market, in spite of fallingrates, or who want to beat the index in a rise, would be to just combine freelysubagents with differently set parameters. In this way people could decide forthemselves how much risk they would like to be exposed to. I would not be surprisedif a set of subagents put simply together could achieve results similar to those of thetrading agent I put together by design.What I have done is a continuation of what Boman et al. did in their study of theParrondo strategy for artificial traders [Boman et al, 2001a]. One could say that theiragents depended on how rates would change (which could be looked upon as a roughform of the first derivative) and was inspired by a capital dependent version of theparadoxical games of Parrondo [Harmer and Abbott, 1999]. I have gone one stepfurther. I have based my agents on a function conceptually similar to the secondderivative, and it was inspired by a temporal version of the paradoxical games ofParrondo et al. [Parrondo et al, 2000]. Furthermore, I have incorporated the cost oftransactions in my tests. I have also increased the number of samples. The results Ihave obtained are valid within the framework of the set-up of my experiments, as Ihave described them. As I have also shown, they have some academic significance.The question whether the work I have done has any practical significance or not, Ihave to leave to future research.5.2 Commercial ApplicabilityIn auctions, time is very important, especially facing a reality where split secondsmight make the difference between loss and profit. If a customer is confined tosending his orders via the Internet he is at a great disadvantage compared tocompetitors that have direct links to the servers of a stock market. There is researchthat shows the importance of monitoring network resources on a competitive market[Boman, 2001]. Artificial agents could remedy injustices that unequal networkresources create.19
As I have already indicated in a previous section, it would be desirable withtrading agents that could subdue overreactions. Investors would like stock prices to beas close to the equilibrium price as possible. A volatile market carries with it a greaterrisk for not trading at this price, which is important to investors. The subagent I haveconstructed may reduce the volatility of stocks as it probably buys and sells stocks alittle earlier than the ‘herd’, thereby facilitating a turn of a trend earlier than it wouldotherwise.Looking at the whole idea of having trading agents implemented on the serversof a stock market, I see a danger in having agent strategies open to public scrutiny,which they would have to be to some extent. The reason for this is that if a strategy isknown it is always easy to find a counter strategy. I do not know whether this couldbecome a real threat in our case or not. Maybe the complexity of the system, whichthe agents would form together with a stock market as a whole, is too great for anyoneto hit upon a working counter-strategy exclusively benefiting from the strategy of ouragents.For future research, other issues that need to be addressed are how taxes wouldaffect profits, and the judicial positions of OM and clients using these agents.5.3 ConclusionsThe test results of the trading agent turned out to be quite interesting. Very much likethe theoretical game of Parrondo et al. [Parrondo et al, 2000], it did better than itsconstituent subagents on a market of falling rates and worse than the constituentsubagents on a market of rising rates. It was very hard for me to draw any conclusionsregarding the core intelligence of the trading agent. All I could do was to observe.Although uncertain, the final run of tests showed that the trading agent achieved aresult 44 per cent greater than that of the ‘Buy and Hold’ strategy, translated to ayearly basis.Keeping the assignment and the hypothesis of the first chapter in mind, theresults I have obtained show that trading agents could very well be a part andfunctionality of the servers of a stock market and at the same time augment thefunctionality of a deterministic matching service.The research I have been doing is still very much at its beginning stages. Theresults I have obtained have academic value, but for practical purposes a lot remainsto be done.5.4 Personal ReflectionsFinally, I would like to add a few words on efficiency of a marketplace. As I havealready mentioned earlier, it is probably important to use one’s intuitive faculties intrying to understand a field as complex as finance. Intuitively, I feel that the onlypossible way to really create a market that does not overreact to events, where pricesare based on the real worth of companies, and where companies do not have to worryabout the going rate of their stock, is to focus on ethics. For this we have to go back toour schools and universities. This could only be a long-term solution though. For ourown generation it is probably too late. In the mean time I think we should face realityand take a look at the character traits of people buying and selling stocks. My intuitiveinsight is that these people (basically all of us including myself) need to experiencethe stock market as boring, at all times. They should feel that there are no shortcuts towealth. In this way the volatility of stock prices would decrease. People would act in away that would be more beneficial for our society as a whole. At the same time I dorealise that it would not be too realistic to assume that people would adopt such a20
Page 1 and 2: Improved P
Page 3 and 4: mathematical sign that all is not a
Page 5 and 6: If the model corresponds well with
Page 7 and 8: this purpose I designed a subagent
Page 9 and 10: exceeded a threshold value, set bef
Page 11 and 12: Chart 2, 26 Feb 99 - 25 Feb 00, sam
Page 13 and 14: Chart 4, Average of 28 samples, 26
Page 15 and 16: Table 4. The choice of subagent.The
Page 17 and 18: separately (compare with results of
Page 19: interesting to see how much time th
Page 23 and 24: [Blume and Easley, 1990] L. Blume,
Page 25 and 26: B, SHB-ATEL2-B,VOLV-B,WM-B, STE-R,A

Improved Pricing on the Stock Market with Trading Agents

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?