Strategy backtest pt.10 - copper trading strategy

In my last last portfolio review I was quite optimistic about being on the right path with my copper trading strategy. Just one week later the strategy revealed its weakness again by creating two lossy trades that ruined all profits gathered before. The overall result after 10 weeks of applying the strategy was thus disappointing:

• without trading mistakes: +2,33% 
• actually realized in my portfolio: -9,33% 

The following chart shows the performance in the course of time. The red line shows the "theoretical" development of my equity without trading mistakes. The blue line shows the actual development including the trading mistakes:

Due to these results I stopped applying the strategy but didn't want to give up that quickly. Instead I adjusted my backtesting tool, so that I can backtest this kind of trading system. My main focus was on finding better and fixed entry signals rather than relying on my chart analysis skills.

Adjusting my backtesting tool

In order to backtest my copper trading strategy I first had to adjust my backtesting tool. So far it was purely signal based - trades were opened when certain indications were found and closed on other signals or TP/SL limits. So the period for which an opened trade is running varies and solely depends on the market movements, signals and trading limits. The opening time was either in the evening (market closing rate) or in the morning (via trade order that was set before market opening).

Now, my copper trading strategy is based on a daily time frame. Trade orders are set in the morning, depending on opening rate. When a trade is opened but not closed by SL/TP it is closed in the evening (closing rate). On some days, no trade is opened and the order is deleted in the evening. For more details on my copper trading strategy, please refer to this older post.

Thanks to the flexibility of my backtesting tool, I could adjust it quickly to the different trading style. However, since I do not have intraday data available for my backtest, calculations can be inaccurate when either SL or TP can be possibly hit. With intraday data I could judge wich limit is hit first. With daily data that is not possible. In my previous backtests I always assumed the worstcase, that is that SL is hit before TP. In order to be more accurate I now calculate both, bestcase and worstcase.

Another inaccuracy I am always facing with my backtesting tool is the conformity of the data. For the copper backtest I took the market data from investing.com. They do not mach the market data of eToro, where I applied the strategy. I already asked for a possibility of extracting the market data from eToro. Obviously they are working on an interface for that but they couldn't tell me when it will be available. I am looking forward to that.

Backtesting results

Since my focus was on finding rules for opening the trades, I kept the offset from the opening rate and the TP/SL limits constant for all investigated trading setups. The only variable was the entry signal itself. In a first shot I concentrated on rather simple signals:

1. Buy on red, sell on green candle
2. Buy on red candle when rate > EMA20, sell on green candle when rate < EMA20
3. Buy on green candle when rate > EMA20, sell on red candle when rate < EMA20
4. Buy on red candle when rate < EMA20, sell on green candle when rate > EMA20
5. Buy when closing rate increased, sell when closing rate decreased
6. Buy when closing rate decreased, sell when closing rate increased
7. Buy when closing rate decreased and a red candle appeared, sell when closing rate increased and green candle appeared

The following chart shows the backtesting results compared to the theoretical performance of my strategy when I applied it on eToro:

Unfortunately none of the tested trading setups performed better than my chart analysis based approached that I used in the past. No, it is even worse: None of the setups managed to generate a profit. The best out of those bad setups is #3. After 10 weeks it has neither a loss nor a profit because not a single trade was opened.

Conclusion

The results presented above clearly show that a simple entry signal approach is not working for this type of day trading strategy. In comparison my chart analysis provided better indications. That's probably because it considers a wider range of signals like trendlines, candle stick patterns and also some fundamental aspects and general market sentiments. It appears difficult to put all that into fixed and automated trading rules. Nevertheless I will keep working on an improved strategy. Until I find that, I will for now not apply my copper strategy any further.

Please let me know what you think about my approach and if you have ideas for improving my strategy!

***

Related posts:

Portfolio Review 18.02.2017
copper trading strategy
pt. 9: "past performance is not an indication of future results"
pt. 8: using trade orders
pt. 7: stop loss limits
pt. 6: delay factor
pt. 5: market trend
pt. 4: short trades
pt. 3: stop losses
pt. 2: long trade indicators
performance index DPI
backtesting tool

***

Important remark: The results presented above and throughout my blog are no recommendation for your trading! I only share my personal findings and opinions to give ideas and let my followers and copiers know what I am currently working on. I can not guarantee the correctness of my tool and my presented results. Furthermore past performance is not an indication for future results. Only trade with money you can accept to loose!