Hello all guys, just for the record I will say that I am completely new to automated systems. I wanted to know this: If I have a strategy that has been backtested already and I want to use it 24/5, what should I do? In which programming language should I write it? Where should I upload it (what options do I have?) ? etc. I understand that this is a long topic, so if instead of writing an answer you could pass me some links, books or even words to search on google I would be glad anyway. Thanks a lot.
Running a system 24/5 requires a reliable infrastructure (power & internet), so typically a server or VPS (virtual private server) hosted in a tier-1 datacenter. Even hosted in a tier-1 datacenter, you will experience occasional losses of connectivity. Your system will need to handle these, in particular download the missing market data once the connection is restored, and update your positions & open-orders when appropriate. You will also have to deal with positions & open-orders rollover in futures markets, and the transition from one contract-month to the next. You might have a broker connection that does not accept orders during session-breaks, and have to deal with it. Of course, you need to be able to stop & restart your trading strategies without impacting them, whether it is for planned maintenance of your server, change of your strategy parameters, upgrade of your strategy software, or reasons completely out of your control (forced update or crash). I have done all of the above for my strategies running 24/7 on NinjaTrader (over 1-year live for CLAlwaysIn, which as its name suggests, is always in the market, long or short). My infrastructure software is very stable and I offer it for lease, PM me if you are interested.
Nice dom, thanks for your answer. Since I published this thread I have been investigating a little while and I decide to use Ninjatrader also. But, isn't there any free platform that allows automated trading without paying a fee? Just curious. Ninjatrader looks for a nice solution, I'm actually waiting for my demo-account data to test it. Thanks again.
The cost of a single trading error might far exceed the price of a NinjaTrader or similar license. If you are serious about automated trading, be willing to pay for the required infrastructure aspects, server or VPS, reliable datafeed, trading platform license, and don't forget quality historical data. Which markets and type of trading are you considering getting into?
Futures and trend trading. Yea you are right about the infrastructure, I will have to look for some of that. I have already backtested the strategy in TOS (just for the last 10 days though) and looks good, now I'm checking it at real time. My biggest problem is to get my strategy reports into excel to develop my own statistics... Mmmmm maybe you could help me with that? I asked it here: http://www.elitetrader.com/vb/showthread.php?t=283117 but seems anybody has an idea
10 days ??? Just to avoid you any serious "market injuries", I strongly recommend that you backtest on at least 5 years AND minimum 1000 trades (this is a very low minimum, but a good start nonetheless)
Yea, I know it's few, but is all I could got with TOS feature. I'm now looking for a better backtesting platform, thanks a lot for your advise.
I just wanted to follow up on what dom993 said. 10 days may or may not make much sense for backtesting, and a certain number of simulated trades may or may not mean much. Things depends heavily on the timeframe for which you strategy applies, how frequently it trades, and the variability in its performance. The important thing is that you collect enough data to accomplish two things: (1) accurately represent the market conditions for which you plan to apply your strategy (you will nonetheless have to periodically adapt to changing conditions), and (2) perform a statistical analysis on it. What this means is that you must particular attention to the error bars associated with your calculations. Neglecting error bars in a statistical calculation literally makes it meaningless (yet this is frequently done). As for backtesting software, many of the available programs do similar things. However, in my opinion, all seem to have their drawbacks. Robert Pardo mentioned something along these lines not too long ago as well, if I remember correctly. In any case, because of this, we have written our own from scratch in C++. If you are serious about automated trading and are a proficient coder, you might consider writing your own as well.
If you back test on anything less then 3 years of intraday data you are setting yourself up for a big fall. There is 10 years of Intraday Futures data available for system testing on the Tradestation platform and many others. There are decades of stock and Index data available. Remember EVERY system will work at some point (RSI's,Chart patterns,Mean Regression, Coin flips, Trading moon cycles,Dart throws etc). Often it may be profitable for months or longer, BUT in the longer term (several years plus) may have a negative expectancy. Conversely a system tested on several months worth of data may be shown to be have negative expectancy, but over many years is consistently profitable. If you dont do the required work to develop a robust system there are many THOUSANDS of other traders around the world who will, and do. These traders will take your money gladly.