Search results

Absolutely, no argument there. The LSP introduction was somewhat of an eye opener. We did understand some of the problems with correlations and tried to fix it. But we have always known that our model was a bit naive. One issue is that as a result of how our trading system works our market...

Interesting ideas in this thread. Thanks to all contributors. I've just read the introduction paper on LSP. It looks promising and I'm trying to figure out if and how we could use these ideas for our trading systems. For the moment we use resampling simulations that use historical data to find...

Haven't done any real work with CUDA yet but I have read the documentation as well as visited their forum. To gain a large performance boost you do need to parallelize your algorithm, but unfortunately that is not enough. Equally important are the memory access patterns of your algorithm. The...

Did a quick search and found a few articles. Even found a few on island/migration GA. From these it seems that the Markov chain analysis is mostly directed towards population convergence (most stipulate population to be ”large enough” or use the population size as an input to the analysis)...

Thanks, will do if time permits. The Grefenstette study, even if old, is interesting since they use a meta level GA to optimize GA parameters over a number of benchmark problems. The Goldberg study is based on schema theory. /Hugin

Don’t want to turn this thread into a GA specific thread, but here we go. We started with population sizes recommended by Grefenstette (i.e. pretty small ~30) but found that we had to increase it a bit, possibly due to the fact that our problem differs from standard benchmark problems. We...

OK, if you say so. In my naivety I thought that a system with high success probability required many samples to be confirmed but only a few to be invalidated. I think I'll have to take it up with my old statistics professor. All this hard work for nothing... /Hugin

OK, I'll take the bait... My guess is that we all do. People in this thread just don't want put in a lot of work into systems with close to zero probablilty of working. Really? This is quite interesting... Care to tell us how you know that a system works after a few trades? By working I...

Markov Chains as a model for how GA works or used as a trading model? /Hugin

Actually the method we use could be classified as reinforcement learning. Most (?) evolutionary algorithms can use the concept of reinforcement learning to explore the state space of the model (if I remember correctly this was included in Holland’s work on GA). We have looked at SVM but...

If you want a useful answer you need to provide information about the risk of the system. ~25% annually is good if the Sharpe Ratio is >2, not so good if its <0.5. A problem with a relative small edge is that it can easily disappear and it is hard to know whether it is gone forever or if you're...

What I say here comes from my experience working with NN some 15 years ago so it may be outdated. Also, our focus then was on feedforward networks trained by the back propagation training algorithm so my answer is biased by that. The main problem we found with changing the goal function is...

That's correct, the reason for us to do it this way is to get away from the limitations of the standard textbook NN training algorithms. There are probably other ways to do it, e.g. by using other network topologies (e.g. non-feedforward) and other training algorithms than gradient based. We...

We use the GA to tune the weights of the NN (our hybrid models may also include other parameterized stuff). The output from the NN is used in the signal generation. The signals are then used to generate trades. Somewhat simplified the GA goal function looks at these trades and see how their...

I have read and I like this thread and agree with what is said above. Which technology to use in order to optimize your model is only one part, and not the most important one. Since we use both Genetic Algorithms (GA) and Neural Networks (NN) in our trading systems I may have some input on...

From your performance graph I would agree that the system is I/O bound. I don't know how the Blocks system works with respect to updating data but one thing that is always important in a system with a lot of I/O (e.g. high usage databases) is to separate disk reads from disk writes. I have...

I think Monte Carlo resampling is a good method. We use it as the major decision support tool to decide if a system has any probability of working. We use two different variants. One takes the same number of trades for random stocks at random times (within the walk-forward interval). The...

Thank you for your comments. I found some research reports related to the topic on the Internet. It seems like the consensus is that that implied is a better predictor than historical. As nitro indicates, if both are included in the trading system, at least we have the opportunity to not take...

The background to the question comes from findings when developing an automated equity trading system. When we analyzed the trades generated by some of the rules we found the signals they generated coincided with breakouts in price and volatility. Now the time has come to take a closer look into...

Thanks Ron, I’ll have a look at it. We have tested a number of simple exit methods (like various trailing stops schemes in combination with hold time limits, profit targets etc) running MC simulations on the trading system level to analyze the results. As always it is possible to find things...