Hoggums
Senior member
- Messages
- 2,176
- Likes
- 878
This exercise is mostly for my own benefit but I thought it would be worth sharing the results with everyone. I've played around with automatic trading in the past and failed to find a system which actually worked (well). Some of them got quite complex using trigonometry to identify trend lines and resistance points but they all failed - usually working well for 95% of the time (sometimes for years) only to have all profits wiped out within a bad week of trading. I never traded any of them for real.
So I thought I would go back to basics, looking at simple indicators such as MA's, RSI, bollanger, parabolic SAR etc and seeing how well they work on their own and in combinations. Hopefully it'll throw up some interesting observations and perhaps start a discussion about how to avoid the arch enemy of any automatic system.. runaway drawdown.
I'm not expecting to find a Holy Grail system here - but hopefully it'll be useful to others in confirming/trashing their beliefs in auto-systems & indicators and I'm hoping I'll find a firm foundation to build a more robust system on, or at least know where not to start!
I will only be looking at price action as I do not have access to volume information. I've written my own application to do the analysis which can cope with the vast amount of tick data needed to test a multi-year period.
To start off with I'll look at individual indicators on their own to see how they perform. I'll try them on a range of timeframes - 1min, 5min, 15min, 1 hour, 4 hour and 1 day bars, to see which perform best. I'm going to try to avoid the trap of optimising too much by sticking with fairly standard parameters for the indicators. I'll supply the results in a spreadsheet with the trades that were triggered so you can perform your own analysis/money management on them if you wish.
Later, and hopefully with the suggestions from members of T2W, I'll test combinations of indicators and introduce simple rules to create some trading strategies, and post the results of the backtests in this thread. My analysis will include simply pips won as well as more complete picture with simple money management techniques - from agressive to conservative, analysis of margin requirements, averaging in/out techniques, stops and limits, interest charges, slippage etc - suggestions will be welcome as this thread progresses.
For the moment I'm exclusively testing on EUR/USD data from Jan 2001-Jun 2007, I've downloaded tick data from gaincapital - I can't vouch for it's accuracy, there is a small gap of 2 weeks in the data in 2004 because what gain supplied was nonsense and I've spent a while sorting out other issues with the data, but as this is a backtest a some inaccuracy in the data is acceptable. I can download data for most of the major ccy pairs but I currently have no access to index future or commodity tick data. If anyone would like analysis of this data at a future date I would be happy to do this, provided they could supply me with the data in a text file format. (e.g. csv, txt format - or anything that can be opened in notepad).
I'm not prepared to test your own trading system to validate it for you unless it is both valid to the flow of this thread and you don't mind revealing the strategy and results in full on this thread.
First Test!!
I'm going to look at moving averages and exponential moving averages, for 20,50 and 200 periods on 1min/5min etc.. bars. Simple rule - if the price is above the (e)Ma - you buy, if it's below you sell. It should give us a reference point for all other tests. Afterwards I'll look at (E)Ma crossovers to see if they offer an improvement.
First set of results using Mas coming up shortly....
So I thought I would go back to basics, looking at simple indicators such as MA's, RSI, bollanger, parabolic SAR etc and seeing how well they work on their own and in combinations. Hopefully it'll throw up some interesting observations and perhaps start a discussion about how to avoid the arch enemy of any automatic system.. runaway drawdown.
I'm not expecting to find a Holy Grail system here - but hopefully it'll be useful to others in confirming/trashing their beliefs in auto-systems & indicators and I'm hoping I'll find a firm foundation to build a more robust system on, or at least know where not to start!
I will only be looking at price action as I do not have access to volume information. I've written my own application to do the analysis which can cope with the vast amount of tick data needed to test a multi-year period.
To start off with I'll look at individual indicators on their own to see how they perform. I'll try them on a range of timeframes - 1min, 5min, 15min, 1 hour, 4 hour and 1 day bars, to see which perform best. I'm going to try to avoid the trap of optimising too much by sticking with fairly standard parameters for the indicators. I'll supply the results in a spreadsheet with the trades that were triggered so you can perform your own analysis/money management on them if you wish.
Later, and hopefully with the suggestions from members of T2W, I'll test combinations of indicators and introduce simple rules to create some trading strategies, and post the results of the backtests in this thread. My analysis will include simply pips won as well as more complete picture with simple money management techniques - from agressive to conservative, analysis of margin requirements, averaging in/out techniques, stops and limits, interest charges, slippage etc - suggestions will be welcome as this thread progresses.
For the moment I'm exclusively testing on EUR/USD data from Jan 2001-Jun 2007, I've downloaded tick data from gaincapital - I can't vouch for it's accuracy, there is a small gap of 2 weeks in the data in 2004 because what gain supplied was nonsense and I've spent a while sorting out other issues with the data, but as this is a backtest a some inaccuracy in the data is acceptable. I can download data for most of the major ccy pairs but I currently have no access to index future or commodity tick data. If anyone would like analysis of this data at a future date I would be happy to do this, provided they could supply me with the data in a text file format. (e.g. csv, txt format - or anything that can be opened in notepad).
I'm not prepared to test your own trading system to validate it for you unless it is both valid to the flow of this thread and you don't mind revealing the strategy and results in full on this thread.
First Test!!
I'm going to look at moving averages and exponential moving averages, for 20,50 and 200 periods on 1min/5min etc.. bars. Simple rule - if the price is above the (e)Ma - you buy, if it's below you sell. It should give us a reference point for all other tests. Afterwards I'll look at (E)Ma crossovers to see if they offer an improvement.
First set of results using Mas coming up shortly....