Trading Blog: TnT Group

Since my last couple of posts dealt with simple backtesting concepts, I wanted to write today about some potential pitfalls of using backtesting.

When conducting a backtest, always remember that what you’re viewing is “simulated” data. There’s no guarantee that a program executing live will behave exactly the same as it did in a backtest. Unless you’re looking at individual ticks, most backtesting systems will build OHLC bars of the timeframe you give it. The reason why this is important to keep in mind is that this means that you can not determine what the price action looked like mid-bar. This is especially evident on days when major economic numbers come out. Looking at historical data, bars with violent moves simply look like any other bar except with a larger range. Anyone who has traded the US jobs report knows that there’s nothing smooth about it. It only takes one trade to trigger a stop and if it’s a stop market order, there’s no telling where you’ll be filled. If it’s a stop limit order, there’s no telling IF you’ll be filled.

The bottom line is that there is no substitute for testing with live data and if your results look too good to be true, they probably are. Always be mindful of outlying days and try to figure out a reason for the abnormal results.

Above all, remember that computers make great tools and as much as it may seem otherwise, they are ultimately stupid. This means that for any system, it is only capable of performing as instructed.

Until the next time, Happy Trading and good luck!

It’s been awhile since I posted the first part of this series so I hope everyone remembers what I mentioned.

The importance of backtesting should be obvious. However, it can be difficult to determine the best way to do this. It is impossible to say that any method is better than any other or that a single method will provide the best results for all strategies, but there are certain things to keep in mind when backtesting any strategy.

When I am testing a strategy, I like to look back at a couple month’s worth of data. I know a lot of people run backtests on years of historical data. Personally, I think this is a waste of time since markets are constantly changing and the conditions that existed a year ago will not exist today. I can only assume that people do this because they are trying to create a track record for how their strategy would have performed. The problem with this is that a track record created in a simulated environment does not hold much value. Everyone makes money paper trading but performance live is a very different story.

Ultimately, the most important thing I’ve learned about ATS development is DO NOT get too caught up with backtesting. Any ATS is attempting to take advantage of a scenario in a certain market. These scenarios do not last forever so if you spend too much time backtesting, it is entirely possible that you will miss your window of opportunity to run the strategy. Trading is inherently risky and sometimes you just have to take that plunge.

Well, that’s about all for now. I’ll probably have more on this subject within the next couple days. Until then, good luck with the rest of the week and happy trading!

Hi there all!

From the title it should be fairly clear that today I will be talking about backtesting.

Backtesting is a critical phase in the life of any automated system. When I say backtesting, I’m actually including backtesting, optimization, and forward testing in the same group. Backtesting is running your strategy on historical data and seeing how it would have executed. Generally, this is followed by optimization where you attempt to tweak various parameters of the strategy to increase your profits. Sometimes it may take many iterations of backtesting and optimization to find the best parameters.

Once the best parameters have been determined, forward testing begins. Forward testing is running your strategy on live data while executing in a simulated market.

Only after completing these three steps should you feel comfortable running your strategy using real money. The whole process can take some time but it’s definitely worth it in the money you can save yourself in poor execution.

I realize that this post may be somewhat short and light on material but I wanted to make sure everyone recognizes the basic vocabulary. Next time I’ll get more into the methodology of testing your strategy.

Until next time, I hope all of you have a great weekend!

Everyone has their own set of indicators that they use to trade the markets. There are hundreds if not thousands of indicators out there. The question is, how do you know which ones are better than others? Or, which combination will give you insight into the perfect trade? I wish I had the answer for you but I too am searching for the holy grail.

Normally, traders will have maybe 2 or 3 indicators that they use to trade off of. Much more than that may be too much to keep track of. But hey! What if you had a computer that could watch hundreds of indicators? Sounds like you’d have a guaranteed winner. Unfortunately, it doesn’t necessarily work like that (trust me, I wish it did).

From what I’ve seen, some of the simplest strategies end up being the most profitable. This is easily explained since most indicators are displaying the same thing at pretty much the same time. Therefore, more does not necessarily mean better. Sometimes it just means redundant.

In closing, I will leave you with this. Almost every technical indicator out there is based on the same 6 properties of any given market. These properties are the ask price and volume, bid price and volume, and the last price and volume. It is truly amazing that thousands of these indicators have been created from only 6 pieces of information.

‘Til the next time, have a great weekend and be sure to KISS (keep it simple stupid)

Hi, my name is Kevin and I’m the lead developer here at the TnT Group.

One of the major services we offer here is the development and backtest of automated trade strategies/systems (ATS). These are systems that watch the market for specific criteria and then execute trades automatically. The concept seems simple enough but is riddled with potential complications and pitfalls. In my future posts, I hope to talk about some tricks to developing your own ATS but since this is an introduction, I will start at the very beginning. What is an ATS?

When defining an ATS, I always look for 3 things:

• Indicator
• Entrance Criteria
• Exit Criteria

ATS indicators can be anything. They can be as simple as using moving averages on prices or as complicated as using Bayesian networks on news events. The goal of any indicator is to take all of the data available in the market and boil it down into a manageable subset that the computer can watch.

Entrance and exit criteria are equally important. You can ask any trader out there and they will tell you how critical timing is. These criteria determine when the ATS looks to get in and out of trades. Small changes to either of these values can lead to HUGE swings in your PnL.

Each of these things must be tested to determine the best combination of parameters to be successful.

The ultimate purpose of any ATS is to emulate the actions of a successful trader. However, there are many things that traders do that computers just can’t. Remember, computers are stupid and can only do EXACTLY what they are told. Always keep that in mind when developing your own ATS.

Til the next time, Happy Trading!

TnT Group… helping Skynet become self-aware since 2010