I downloaded the new release this morning. It fixed a number of bugs that I had discovered over the last few weeks. I like the new NGO 2.6 release. It is a little quicker, and the double validation feature between NGO and Profit will really come in useful.
I agree on the copy protection. I rebuild my systems almost monthly (I am an MCSE, so I am constantly 'playing' with new software to learn it!) so I can see myself phoning them up quite a bit to get activation/deactivation codes.
Now to your questions:
1)I have had luck paper trading the models. I haven't placed any *real* trades with a Profit/NGO model. Having only had the package two weeks, I am not comfortable enough to trade with it yet. It has given some really good signals on the SP500 though, and I only spent a day or so working on the models. I hope to trade these in August, once I have refined them a bit, and paper traded them enough that I am comfortable putting real money on the line.
2) That is a tricky question. Some of these models will be dead on in their predicitions. But I find these are the models that tend to stop working suddenly. For example, if you keep your training set short (>18 months) and use short term oscillators as inputs, you will usually get really high returns. But these models have a short life, often becoming useless as market dynamics change 4 to 8 weeks out. I prefer a more fundamental, longer term approach, although it makes a little less money. I don't want to retrain every few weeks, as it invalidates the previous results, as you are then trading new models.
As far as curve fitting, as long as optimization does not occur on the validation set, you should be okay. I prefer not to optimize my models, choosing instead to build them up to better results.
3) I use Reuters Trend Datalink (http://www.equis.com/sales/rtd/index.html) as my data provider. I have seen a number of people using data from CSI (http://www.csidata.com) as they have a good selection of Futures data. You can try Pinnacle Data Corp (http://www.pinnacledata.com/) for some really hard to find stuff, and some great data management tools (Datamaker).
4) There is a list of books I thought were good a while back in the thread. One book that is worth its weight in gold in Cybernetic Trading Strategies, from Murray Ruggiero. He works with Futures magazine I believe. Inside you will find a lot of timely information on all the high tech trading methodologies including genetic algorithms, neural nets, cycles (MESA), etc. What I liked about this versus all the other books is that he doesn't bore you with all the math and theory, but gives enough of explanation to get you going. I go back every once in a while, and it gives me some new ideas to try out. There is a wealth of information here.
Looking back this sounds kind of like an infomercial. Act now and I'll though in a free set of Ginsu knives... :)
Optim |
