Preliminary backtest on 'double-top' and 'double-bottom' candles

This is still at an early stage, so it may not work on other pairs, but I just finished testing a 'double-top' and 'double-bottom' trading system on the USDCHF. I identified short-term support and resistance on the daily chart whenever two consecutive candles have very similar highs or lows (within 10 pips). A snapshot is below:

I found if the second candle had a range between 0.5 and 0.75 ATR(14), there's a good chance of a reversal. I excluded candles below 0.5 ATR(14) since these candles would already be traded with my Hermes low volatility system, and I wanted to avoid overlapping data between two different trading systems.

Candles with a range of over 0.75 ATR(14) don't seem to work so well. I think this is because the reversal is already under way if the second candle is large, and if you decide to trade it, you've missed out on the meat of the move. 

Anyway, with regards to the screenshot, I placed my stop loss at the top of the second candle, and my entry on the break of the bottom of the second candle. For long positions, just reverse the SL and entry. Here's a summary of profit factors for various R:R ratios: Total number of trades during the time period was 207.

1.75 reward:risk looked ideal. Here's the equity curve with an initial balance of $10,000, 2% risk per trade, starting from Jan 2001 to Jan 2014, using a 1.75 R:R ratio. 

It's a good start, but I know how these backtests can easily fall apart. Stay tuned...

Week in Review: 17th February 2014 to 23rd February 2014

Made a bit of profit this week. I won six out of eight trades, with three trades left open over the weekend. However, those three open trades are currently negative, and I expect one of them to hit its stop loss on Monday. I only left that particular trade open out of discipline (need to stick to your system's rules, no matter what). 

Here's a snapshot of two of the trades that I took this week.

USDCAD - Daily Chart

This worked out quite well. The 18th of Feb saw low volatility, so I expected the next day to have some good movement and put in a pending long and short on the break of the high or low. 

Because my reward is only 0.5 * risk, I managed to avoid getting hurt by the initial bearish 'fakeout', as seen below. My short was triggered when the low of the 18th Feb broke, and I took profit at 0.5R before price turned back upwards.

As price rose and broke the high of the 18th Feb, it triggered my long and I took profit at 0.5R.

System Development

I just created the Ebisu trading system this week. It's based on large pinbars that appear on the daily charts and backtesting shows it to be profitable over the last 12 years. based on 18 currency pairs .  

At the moment I'm backtesting a new system based on short-term support and resistance on the daily charts. Will post more details if the backtest looks good. 

Presenting Ebisu large daily pinbar system

Finally finished backtesting large daily pinbars. The end result is good enough to trade, so the next stage is trading it live in a forward-test. I've decided to call this system Ebisu, named after the Japanese god of luck and sea merchants. 

Entry Signal

We look for a pinbar with a range greater than 1.5 * ATR(14). The open and close of the pinbar must occur within the top or bottom 50% of the candle.

Example:

Entry and Exits

We enter on the break of the pinbar's short wick, with a stop loss at the opposite end. We take profit at 0.5:1 reward:risk (or 50% the size of our stop loss). Yeah, this is an inverted R:R ratio, but the backtest showed this ratio to be optimal.

Win Rate and Profit Factor

Using a reward of 0.5R, the backtest win rate was 76.6%. The profit factor was 1.62, using my broker's costs.

Pairs Tested

Tested 18 pairs from 2001 to 2013.

Historical Equity Curve

This is an equity curve from 2001 to end of 2013, with a starting balance of $10,000, risking 2% per trade.

Notes

This is a relatively rare signal, occurring perhaps once every 1-2 months. I only found  244 signals between 2001-2013 across 18 pairs, and many of these signals appeared on different pairs during the same day. Using this system alone won't make you rich due to the infrequency of signals, but when an exceptionally large pinbar does appear, this would seem to be a good way of exploiting it.

Preliminary results on 'exceptionally large' daily pinbars

I just finished backtesting exceptionally large daily pinbars on five pairs: the USDCHF, GBPUSD, EURGBP, NZDUSD and USDCAD. I defined 'exceptionally large' as any pinbar that had a range higher than 1.5 * ATR(14). 

From the results, it looks as if a reward:risk ratio of 0.67:1 seems optimal, although all R:R ratios look great. The following is a snapshot of the profit factors for each pair at 0.67:1 reward:risk. 

While the profit factors look good, you should notice the lack of actual trades in the backtest. From 2001 to the end of 2013, I only found 68 pinbars that had a range greater than 1.5 ATR(14) on the daily charts. This was from five pairs. Based on these figures, we can expect around five signals per year, or one signal every two and a half months. I still have 13 pairs to backtest, so if we trade 18+ pairs, we may get a signal once or twice a month, which is better. I hope to get the backtest completed by next weekend. 

Week in Review: 10th February 2014 to 16th February 2014

It's been an exceptional week. I opened 15 trades and won 12, lost one and have two trades which will remain open over the weekend, so overall it has been very profitable. This kind of performance isn't normal, though. A snapshot of the losing trade is below.

EURGBP - Daily Chart

I went long on the break of the high of the low volatility candle. This trade came close to hitting my profit target, but fell away when it was within 3 pips, and eventually took out my stop loss.

When my stop loss got hit, I went short and won the next trade. In trading, you shouldn't have a bearish or bullish bias. You need to be prepared to change your mind if you see a signal telling you to trade in the opposite direction. 

System Development

I realised that I've been spending alot of time trying to design something for the short-term 5M and 15M timeframes, but have nothing to show for my efforts. I spent yesterday going back to the daily timeframe and felt like I was back on familiar ground. I've focused my backtesting efforts on testing extremely large daily pinbars. I still got alot to do, so if I find anything positive, I'll post it on the blog.  

Six months on...

I just realised that I've spent over six months trying to develop some sort of daytrading system and haven't succeeded yet. I've learnt alot, and discovered some nuances that don't exist on the daily or weekly timeframes. However, I haven't been able to translate these discoveries into profitable strategies. This is the toughest project I've encountered to date.

Week in Review: 3rd February 2014 to 9th February 2014

Saw some action this week. Five trades were triggered. I won four and lost one, capping my third consecutive profitable week. Example of one of the winning trades is below.

AUDCAD - Daily Chart

This was based on my Hermes low volatility system. Spotted low volatility the day before the RBA's interest rate decision. On the actual day, the RBA announced that it was no longer cutting interest rates for the foreseeable future. The RBA's announcement proved bullish for the AUDCAD and I went long on the break of the previous day's high.

I could've handled the bullish rise much better if I used a bit of discretion here. The upward move was based on fundamentals, not purely technicals, so all types of traders were buying the AUD. In any market, you will have different types of traders with their own philosophies, methodologies, outlook etc. If you can find a situation where an extreme diversity of traders suddenly find themselves agreeing with each other and taking the same side of the market, you'll find large movements like the one shown above. It's a good low-risk, high-reward situation. I should keep better track of news releases and see if a move is driven by technicals or fundamentals (or both). If the move is mainly driven by fundamentals, then the move will more likely be bigger in scope.

System Development

Not much to report on here. I'm still testing pinbars during the Asian session. Large, prominent pinbars seem to be doing okay, but the problem is the lack of signals. Over a five month period, there were only 41 pinbars with a size greater than 1.5 ATR(12). And this is on the 5 minute chart. If trading manually, you may have to stare at a chart for perhaps 20 hours before you see a signal. That's not a lifestyle I want. Still doing more testing, though. 

Week in Review: 27th January to 2nd February 2014

It's been a quiet week. Two trades were only triggered. I won the first trade, while the second trade is still open. A snapshot of the winning trade is below.

NZDUSD - Weekly Chart

This is based on my Cernenus system. A pinbar formed on the NZDUSD weekly timeframe last week. I went short on the break of the low this week and took profit at 0.25:1 reward:risk. It's a low R:R ratio, but according to my backtest, this is very optimal over 10+ years.

I didn't like shorting the NZDUSD since the overnight carry cost is huge due to New Zealand having the highest interest rates in the developed world. 

SYSTEM DEVELOPMENT

Over the last week, I spent alot of time backtesting low volatility breakouts during the London and London / NY sessions but the results were too marginal for my liking. I just couldn't make it work to a satisfactory level. The only useful finding was that taking positions around the London open and during the London / NY overlap were more profitable than during any other hours I tested. 

I'm now testing pinbars on the EURUSD 5M chart during the Asian trading session. I gained this idea after reading some research from FXCM, which showed that range-friendly strategies work well during the Asian session, especially when using an oscillator like RSI. I intend to filter my pinbars with an RSI indicator so that I only short a pinbar when RSI is overbought, and go long when RSI is oversold. I also look for significant pinbars which have a range greater than ATR(12).

I only tested 106 sample trades from 2010, yielding the following equity curve (initial balance: $10,000, risk: 2% per trade, reward:risk 1.25:1)

106 trades is much too small to have a verdict, but on the back of FXCM's research, I'm very eager to continue my testing. Will post further details as my backtest progresses.