If you’re not familiar with the term, “trading system” is actually a common name for a number of very different applications.
Some systems use the newest technologies and know-how, and are state-of-the-art applications. Such applications may be very complex and employ principles that may seem counterintuitive to the general public at first glance. Retail traders almost never use them.
Other trading systems implement a handful of specific proprietary trading ideas, which aim to eliminate the trader (and therefore, emotions) from the trading process.
Either way, automated trading systems are no longer unusual. In fact, 70%-80% of the trades on the NYSE and NASDAQ are generated by automated trading systems.
So, whether you have your own great idea for a trading strategy and are thinking of hiring a qualified developer to help you program a trading robot of your own, or you simply want to purchase an existing system from a reliable trading software vendor, here are seven critical truths about these systems that you may not have realized.
1. Previous Success Can Be Real And Yet Still Be Misleading
Have you ever seen ads on the Internet for trading systems that promise five or even ten-fold ROI?
Usually in big flashing letters, right?
Click on the ad and you’ll find a limited-time offer from the owner of an “exclusive” trading system that may include trading reports showing excellent results.
While this may or may not be real performance, you need to know that almost any strategy can show very good results if you carefully select the time period. The appearance of a maximization of returns can be achieved by applying a mathematical technique called “optimization” that only shows the results for the strategy’s best performance in a specific date range.
This range is, of course, always in the past. For example, an over-optimized trading system can show exceptional results for the last 5 years up until now, but from tomorrow the system will probably begin to lose money.
Don’t fall for historical results. Ask for a trial period of at least one month, preferably 3-4 months, and see how the strategy performs in live results.
2. You Need To Review EVERY Parameter
Some trading systems work on a specific symbol only. Some only work at a specific time. Others have specific requirements related to the dynamics of volume.
You MUST know the right settings for your strategy. The correct parameters can turn a losing strategy into a winner. If someone develops a strategy for you, ask them to parametrize everything possible so you can fine-tune your system over time.
3. Stop Loss Order Strategies Are NOT Going To Be Enough
Strategy must provide you a way to control your losses. Most strategies don’t do anything beyond trivial stop loss orders and that clearly isn’t enough. For any trade you must know your maximum possible loss to the dollar.
At any moment your strategy must know “do I have enough money for the next order?” or even “do I have enough money for the next 5 minutes?”
Every time you enter a position, ask yourself: How much could I lose here? If the strategy doesn’t answer this question then you should think twice before using it in a live environment.
4. Backtesting Will Fool You If It Can’t Handle Intra-bar
Every strategy is back tested on historical data before going live. Trading platforms today offer various backtesting tools, however it’s important to understand one critical nuance:
Backtesting can be run on bar’s close or on each tick.
These two approaches are very different. In the former case the strategy just gets four values: Open, High, Low, and Close; in the latter the strategy processes each tick that actually occurred on the exchanges.
If your strategy works Intra-bar (i.e. it needs to see each tick rather than just bars) then you MUST back test it using Intra-bar as well.
This seems rather simple and obvious but, in reality, many people only test their strategy on bar values because tick-based backtesting is sometimes unavailable or has limitations. For example, due to the massive amount of data involved, tick-based backtesting may only be available for the last month. It also takes much longer to test the strategy this way, but then that’s the price you should be willing to pay to obtain precision.
A simple Renko strategy may win 100% of trades on historical data while in reality it will just keep losing. This is a good illustration of the difference between testing on bars and then trading Intra-bar.
5. Demo Trading Can NEVER Perfectly Replicate Live Trading
Even if you do it correctly, don’t expect the live performance to mimic the demo precisely. Generally, if the live strategy performance matches the demo strategy 90% of the time, you should consider this a great result.
The reasons for the occasional discrepancies are factors such as market data differences and order execution logic differences.
You can enhance your performance by, first of all, making sure that your live data feed comes from the same provider you use for live trading.
As for order execution flow, it’s not really possible to precisely re-create the conditions in a demo environment that are identical to a live environment. At any given moment, there are other traders competing for a given price. Order book is also constantly changing and the dynamics of these changes depend on market, time, and other factors.
While some trading platforms provide sophisticated simulation of the order flow, it will never be able to fully replicate the real market activity.
There also may be major differences in terms of costs between live strategy and back test strategy. Since the former has to deal with real brokerage, some orders may not get filled and some may be rejected. There may even be a limit for requests, and the connection to a broker could go down.
To handle all such events properly may require more effort than it takes to build the strategy itself. So, don’t sweat every single detail. Just plan for a certain percentage slippage and accept that some of your orders will never get filled.
6. “Raw” Data Feeds Can Produce Strange Results
Some data providers send data to their clients exactly as it comes from the exchanges, i.e. in a “raw” format. Such data will contain everything, including unusual non-public trades, and even errors.
If you use these “raw” feeds your strategy should be programmed to deal with the occasional strange single trade, which has a price a long way away from the actual action.
Otherwise you could experience, for example, a prematurely triggered stop loss.
In contrast to raw data, other providers filter out all irrelevant ticks from the feed, only leaving in trades that change the Last price (so-called, “qualified trades”).
Make sure that you understand the difference between raw and filtered data and that your trading system is also aware of the difference.
7. Some Vendor Support Can Be Patchy At Best
This one is always neglected.
The strategy is ready, isn’t that enough? Well, in some cases, yes. If a strategy shows no results and is quickly forgotten, then ongoing support is redundant.
But what about on those 5%-10% of occasions when the strategy DOES make money?
When this happens, sooner or later you’re going to want to make some adjustments to make your winning strategy even more successful. It’s these ongoing tweaks that require solid support from your vendor.
In order to make changes quickly and safely there must be a good development process established but, unfortunately, amateur developers don’t bother with such things. Sometimes the first couple of changes can be implemented with brute force, the adjustments work and the client is happy. But then the next change throws up more problems than it solves. New bugs appear, the performance degrades and, finally, the vendor stops returning your calls. The build was likely so unprofessional and poorly documented that the vendor can’t fix the issues. So now you have to pay for the whole thing to be rebuilt from scratch.
If you plan to quickly test your trading ideas, you may not need a process. However, if you plan to go live and especially if you have specific requirements to trading (say, for instance, you watch 200 symbols) then make sure your strategy is always in a maintainable state, and make sure you pick a reliable vendor, right from the start.
I hope that you now have a better understanding of the criteria you should use when buying or ordering a trading system. Here is a quick summary of the things to look for:
- Live results vs historical results
- Trading parameters
- Risk per trade
- Bar-based vs tick-based backtesting
- Demo trading vs live trading
- Raw market data vs filtered market data
- Strategy support and maintainability