Game trading engine

Has become Algo-Trading Game

Client: Radmilo Racic, IMC (Netherlands) (Radmilo.racic@imc.nl)

Data mining and statistical modeling is an essential part of generating successful models. There are various approaches to generate a statistical model, but verification of ideas are usually standardized to a couple of criteria. In this project, the aim is to create a back-testing engine that is capable of evaluating different models according to well-known success metrics such as P&L, and risk metrics like returns, alpha, beta, max drawdown and sharp ratio. The project involves efficient parallelization of experiments with parameter sweeps and visualization of the success of the models. Succinctly put, the input to this engine is a strategy and the output is a vector of success metrics mentioned above.

feedback

I've been trying to think of a way to incorporate this into a product design problem of a kind that would allow a team to work together on different aspects of the problem, but I'm afraid I've been struggling with it. It seems like quite an interesting research problem for a single student to work on, or perhaps a pair working together, but we try to find ideas where students with a range of interests and abilities can make different kinds of contribution to a team project. We also need to keep in mind that these are only second year undergraduates, so asking them to create a novel machine learning engine that has a chance of outperforming existing algorithms may be a little ambitious.

At present, I'm wondering whether we might be able to take elements of several of the proposals that have been coming from your colleagues at IMC, and combine these to make a single project that would be better suited to having a design team work on it.