The following tip is mostly applicable to the NeuroShell Trader and NeuroShell Trader Professional (although the ETF info at the bottom should appeal to DayTrader users too). When using the NeuroShell DayTrader Professional, there is no problem getting plenty of data for most contracts, because you can get bars smaller than daily bars. The ability to get lots of bars to build models is often helpful even if you aren’t “daytrading”.

There are special considerations for predicting futures contracts with daily data. In a single contract, there is not enough (daily) data to train a network much less do walk forward testing. Futures contracts only last a few months. Even if you aren’t using neural nets, you’ll still need enough data for decent backtesting, both in-sample and out-of-sample.

Therefore, futures data is often “concatenated” to provide “continuous futures” contracts. This is data that combines futures contracts to provide a long, continuous stream of data. Now you have enough data for training and walk forward testing. Some data vendors also provide “adjusted” continous data, in which the older contracts have their prices adjusted to match the next contract. This avoids “jumps” in the data where the contracts are connected.

When you go to predict the future, the most recent part of the continuous data is the same as one of the most recent contracts (check with your data vendor to when they switch contracts).

Here’s another technique we have found useful in our own trading in making models of emini futures contracts. We don’t use the emini data at all. For the S&P emini we model with SPY data. For the Nasdaq emini we model with QQQQ. For the DJ emini we model with DIA. Those three ETFs are excellent proxies for the emini data itself, and with less noise we believe. Therefore, we think the ETFs are even better than using the eminin data.

If you want to trade using our trading interface in release 5.0, you can even “map” the ETF symbols, SPY to ESxx etc. The only problem with such mapping is that the entry fill prices that get reported back will be at emini price levels but the intratrade fill prices are still SPY and thus until trades close out NeuroShell will be showing invalid in-trade profit.

You can also build your model with an ETF, then execute trades using the current emini contract if that seems more comfortable to you. You start with the single futures contract you are interested in predicting into the future, i.e., make your chart with that contract. But since you can’t train with so little data, you insert the ETF data using “Insert Other Instrument Data”. Then you build your rules or your neural net based upon the Other Instrument data prices. If it is a net, even your output will be based on the Other Instrument data, and when you are done you insert your net into a trading strategy for the futures contract. The buy/sell signals will apply to the single contract, however, instead of the ETF, even though the net and/or rules were built with the ETF data. That is because the chart was built with the single contract, and that is what you want.