Conclusion

It has been shown that the past work by Jansen et al. [17][18] is reproducible and it is possible, by the new technique outlined in this thesis, to infer the search strategy. The compression achieved on the chessmasters, although not as favourable as before, is still good. The conclusion that chess masters follow a strategy closer to brute force technique rather than quiescent, was also an interesting finding. The success of inferring this information from purely the recorded data is very beneficial, as it demonstrates this method can be used on fields of study that have intrinsically uncontrollable data. These fields can be in areas where it is impossible or very difficult to have controlled experiments and use real time learning such as financial markets, psychology, crime studies and other realms. The work done in this paper has duplicated the previous work and elaborated on it. The extension has found new and interesting facts such as that the style of play in the data used for Kasparov is more innately compressible than that of Fischer and that, as such, it dictates that different people can be predicted at different levels and thus it may be possible to find an optimal way to compress each different player. This fact become more important, considering it can be used to differentiate between the various styles displayed by a variety of adversaries. This thesis also shows that programs can be written to learn and adopt some aspects of a person's strategy.