Am I back-testing correctly?

Hi all

I am currently paper trading a trend following strategy based on the Turtles system. I am entering long on 52-week highs and exiting when moving averages cross, and vice-versa for my short positions. I want to back-test by system, but would rather do it manually than on a computer program as I feel this would be more beneficial to my learning, and I'd have more input over the trades that I take.

I am planning to go back as far as I can in a stock chart and literally just scroll through the history of the price, taking the signals and trades, and making a record on whether or not it was a win/loss. I will also start with a realistic amount of capital and record it's fluctuations throughout, and ultimately whether or not I am in profit or in loss by the end.

My main concern is whether or not this is this an effective/useful exercise and means of backtesting, as when I actually begin trading I won't just be trading one stock from start to finish? I was planning to carry out numerous back-tests in this way, each stock at a time, and see if my results were generally good/bad, but I am wondering whether or not this is an accurate representation as I will be trading multiple stocks at each time when I begin trading in real-life, therefore meaning the number of false signals/losses I could take and drawdown will be higher. Is this just one of the limitations of historical backtesting, or is there a more effective way of doing this?

Any thoughts would be massively appreciated.

Thanks
Sami

syusuf66 said:

Hi all

I am currently paper trading a trend following strategy based on the Turtles system. I am entering long on 52-week highs and exiting when moving averages cross, and vice-versa for my short positions. I want to back-test by system, but would rather do it manually than on a computer program as I feel this would be more beneficial to my learning, and I'd have more input over the trades that I take.

I am planning to go back as far as I can in a stock chart and literally just scroll through the history of the price, taking the signals and trades, and making a record on whether or not it was a win/loss. I will also start with a realistic amount of capital and record it's fluctuations throughout, and ultimately whether or not I am in profit or in loss by the end.

My main concern is whether or not this is this an effective/useful exercise and means of backtesting, as when I actually begin trading I won't just be trading one stock from start to finish? I was planning to carry out numerous back-tests in this way, each stock at a time, and see if my results were generally good/bad, but I am wondering whether or not this is an accurate representation as I will be trading multiple stocks at each time when I begin trading in real-life, therefore meaning the number of false signals/losses I could take and drawdown will be higher. Is this just one of the limitations of historical backtesting, or is there a more effective way of doing this?

Any thoughts would be massively appreciated.

Thanks
Sami

Click to expand...

The way you are suggesting to backtest is very similar to method I use for backtesting and forward testing in fx instruments. I had the same questions at the outset - in real-time I will be monitoring a lot of instruments at the same time, in testing i am only looking at one at a time. After doing a lot of testing in this manner, I would say this was not a problem - and i found that i got testing done more quickly rather than testing five instruments simultaneously.

The ForexTester2 product ($200 one off cost) is very suitable for the type of testing you are looking to do. Provided you have accurate historical data, you can upload this data into the software and then roll it forward one bar at a time - the charts look very similar to MT4 (except that the candles are a little narrower and the gaps in between candles a little bigger).

I found testing in this manner very good for learning, rather than doing an automated-type backtesting exercise. I think I learned a lot more doing it this way.

Regarding the turtles strategy, in the interest of saving you time and potential disappointment - it was originally applied to the futures markets (not the equity markets) and worked well in the 70s and 80s. Some people (including some hedge funds) still trade it very successfully today in a variety of markets, but they have made adjustments to the strategy.

I did a large amount of backtesting using Excel for a dozen currency pairs for 10-15 years on the daily timeframe, and found it produced a barely break-even result after transaction costs.

What you could do is use the original Turtles parameters, use that results as a benchmark and then try to improve on the strategy with refinements.

Hope that helps.

yes, that really does help. Thanks a lot for taking the time to get back to me.

I understand what you mean about the Turtles strategy too, as I have heard many times that it no longer works and I know that it was originally used on futures. However, I have also heard many people have had success with an adjusted version of this strategy too, and I will be using an adjusted version too. I will remain open minded about it though, and just see for myself.

thanks again

I agree that "manual" back-testing is a very useful learning tool. It can help you really understand the system you have, indicators you use, etc. at a level that you will struggle to get going with more automated systems. Back in the old days I actually had to calculate my own indicators!

As for the Turtle system, it's less about the fact that they were trading futures and more about the fact that they were trading a widely varied group of instruments. Read diversity of exposure. That's something you're going to struggle to do in forex because there's so much overlap of risk. You would probably have to work very hard to do it in stocks because of the influence of the broad market.

thanks for the response. I do plan to diversify eventually, but wanted to start off my testing on equities. I have also heard about people who have used systematic trend trading systems similar to those of the Turtles who have mainly traded stocks, and have made decent returns. I'm not saying this is or isn't true, but just going on what I have read.

syusuf66 said:

yes, that really does help. Thanks a lot for taking the time to get back to me.

I understand what you mean about the Turtles strategy too, as I have heard many times that it no longer works and I know that it was originally used on futures. However, I have also heard many people have had success with an adjusted version of this strategy too, and I will be using an adjusted version too. I will remain open minded about it though, and just see for myself.

thanks again

Click to expand...

the turtle strategy..they had numerous strategies
back test those numerous strategies they still work today, unmodified

Personally speaking, I can think back to when and why backtesting has become a chore and a waste of resources for me, so these might be pitfalls to avoid -

* entry/exit rules not comprehensive - unexpected price action making it hard to see why trade on certain day opened or closed

* collecting too much data - decide the questions you need to answer before deciding what information you could gather

* temptation to curve-fit - saying to yourself, well, this was exceptional, so I would not have taken that trade in real life or I would have held the position here - when the trade rules said otherwise


The planning should be where you find you've had to work hardest, the data collection should then be a pleasure.

IMHO, tomorton made some great points! With that in mind/on that topic....
dbphoenix shared some great info on this thread below (post#5 on that page), in regards to testing - I read through the pieces that he shared and thought it was very useful.

http://www.trade2win.com/boards/futures-options/215662-manual-backtesting.html#post2749488