Tuesday, October 27, 2015

RESEARCH: Low volatility, high volume end-of-retracements (1H)

It's nearly 2am here. I'll post some interesting backtest results, then go to bed. Be warned, this is pretty complicated.

Just a brief background on tick volume. Since the forex market is decentralised, there's no way of measuring actual trading volume. The next alternative is to use tick volume, which approximates trading volume. 

I've hacked together a custom indicator for MT4, which divides tick volume by a candle's range. So if tick volume for a candle was 200 ticks, and pip range was 10 pips, it would return a value of 20 (200 / 10 = 20). 

Basically this would tell me how much "resistance" a candle faced as it formed. A high value would indicate high resistance, as it meant price required more ticks to move a pip.

Following from my first example, suppose we have another candle with 200 tick volume, but a range of 5 pips. 200 / 5 = 40. This is higher than the 20 we calculated in the first example. This shows that price faced much more "resistance" in this candle. 

With this indicator, I can understand the context of tick volume better. To provide even better understanding, I also added a 6-period SMA to the indicator, giving me the average ticks / pips for the last 6 candles. 

Getting on to the backtest, I examined the USDJPY and AUDUSD on the 1H timeframe, focusing on Tuesday, Wednesday and Thursday, as these days are usually best for daytrading (volatility on Mondays and Fridays tend to be lower). 

I specifically looked for "low volatility" candles which had a range of less than 0.67 ATR(24), AND formed during a retracement of the trend (measured with RSI(24)), AND must have ticks / pips > SMA(6).

Basically, this signal is telling me that the retracement is ending, and we know this because price has slowed down (it has a range < 0.67 ATR(24)), AND we know that more orders have come in, as ticks / pips exceeds SMA(6). Less volatility + more orders = end of retracement. Example is below:

(click to enlarge)

I backtested this signal from September 2013 to September 2015, with an entry at the break of the candle's high or low, in the direction of the trend, and a stop loss at the opposite end. I gathered 359 trades, and obtained the profit factors for the following reward-to-risk ratios:

The profit factors are quite slim. Transaction costs really hurt at this timeframe, eating up about 10% of each trade. That's a significant penalty. 

I put together an equity curve for a reward-to-risk of 1.25, using a starting balance of $25,000, with 2% risk per trade, and 10% transaction cost. 

It's not the best looking curve, but it's pointing in the right direction. Under real trading conditions, I don't think I'd just blindly aim for 1.25R reward with every trade. I'd see if there's any significant S/R or price levels before entering. Looking from the backtest results, rewards between 1R and 2R seem best. 

The next step is to see if this behaviour holds on other pairs. I chose the USDJPY and AUDUSD as their volatility is generally consistent throughout the day. I'll test the NZDUSD, and then the EURUSD.

Until then...


You can find a ticks / pips indicator for MT4 here.

Saturday, October 24, 2015

WEEK IN REVIEW: 19 Oct to 25 Oct 2015

This week has been a bit quiet, but profitable. I only entered 3 trades, winning two and losing one, and reversed most of the losses I made the previous week. 


This trade took awhile to conclude (at a loss). I saw a significant bullish pinbar on the 16th of October, and went long. While it hadn't broken the resistance level of the order block, the large tail showed significant buying. I felt there was a good chance of a breakout, and aimed for a reward of 1R. The trade was eventually stopped out on the 22nd October.

Looking back, was it a good trade? While the setup fitted my trading rules, I think I should've aimed for a bigger reward. Trading into support / resistance isn't really wise, so you want a higher reward to justify the low probability of the trade. I think a reward of 1.5R - 2R would've been better. 

(click to enlarge)


This was a quick trade I took on Thursday 22nd October. I spotted a bearish pinbar, observed that momentum was strongly bearish, and that there was some scope for a tight reward. I got out at 0.5R profit.

(click to enlarge)


Most of this week's profit came from this trade. I saw a low volatility candle on the 22nd of October, and short-term momentum had just flipped to bearish. I wasn't really comfortable and felt this was a lower probability trade, but saw that there was plenty of room for a downward move (next level of buying was between 1.12000 and 1.11000). I aimed for a reward of 2R.

(click to enlarge)


I've reversed most of my losses from last week, so October is now looking profitable again. There's one more week to go. I shouldn't obsess about monthly results, though, as I may only 15-25 trades a month. Hmmm, this is an interesting psychological observation. 

Tuesday, October 20, 2015

WEEK IN REVIEW: 12 Oct to 18 Oct 2015

This post is a few days late.

I ended down last week, with 3 losses and 1 win. At the end of the week, I felt very unhappy and walked away from trading to recollect my emotions.


I had entered this trade the previous week, and held it over the weekend. It got stopped out Tuesday 13 October. Price resumed moving in my favour the next day. Ah well. 

(click to enlarge)


I had spotted a tiny bullish pinbar at Monday's close, and placed my TP just below a congregation of local highs. I was stopped out the next day. The candle's small range may've signaled sellers moving in.

(click to enlarge)


This is quite an exotic pair to trade. I saw a low volatility retracement candle at Tuesday's close, and went short on the break of its low. I took profit some distance above the last major swing low. I think my TP was technically well-placed, and got out with a 1R reward.

(click to enlarge)


This one was a loser. I spotted a low volatility retracement candle on Thursday morning, and went long when it broke higher. Alas, price action for the day was very choppy. The market turned against me shortly after I entered, and hit my stop loss. It then resurged upwards and closed higher for the day. Very choppy.

(click to enlarge)


Last week was psychologically damaging. I finished September strongly positive, but have struggled for October (I'm down about 0.8%-1% for the month). I decided to step away from trading during the weekend, and spent some time yesterday and today "system hopping", which is very bad. Basically I opened up a new chart and felt like experimenting with daytrading with 0.05 sized lots. I think I ended up opening about 10 trades and losing $5. Could this be an outlet? If you feel the need to revenge-trade or to system-hop, do it with a small amount and get it out of your system.

I also spent today going over my journal and doing more backtesting. I found that positive backtests + a profitable journal to be mentally soothing. Keep these handy!

Besides the "dummy" daytrades, I haven't opened any other live trades. I'll get back to proper trading tomorrow. 

Thursday, October 15, 2015

RESEARCH: Oanda Vs Alpari Open Ratios

OANDA and Alpari both offer their open position ratios to the public. Basically, you can use this info to gauge retail sentiment before you open a trade (it's generally best to use retail sentiment as a contrarian indicator, since most retail traders lose). 

The open position ratios of both brokers aren't exactly the same. OANDA indicates that it updates its open position ratios every 20 minutes. I assume that Alpari updates its ratios once a day. 

I checked the percentage of long positions open from both brokers, and came up with the following:

Like I said, the open position ratios of both brokers aren't exactly the same.

So which is most reliable? As OANDA says it updates its ratios every 20 minutes, I'm leaning towards OANDA.

However, if you look on Alpari's page, you'll see that it offers the open position ratios of MANY more currency pairs. OANDA only offers ratios on 14 pairs, whereas Alpari offers ratios on 40+ pairs, including exotics. 

If you're only trading majors or popular crosses, then using OANDA is probably your best bet. For everything else, you have no choice but to use Alpari (if anyone knows of any other brokers offering their open position ratios, please let me know).

Having said that, I wouldn't bother using retail sentiment if it's within 5% of 50% (e.g. 52% long, 48% short). 

And it's just my hunch, but retail sentiment for the more exotic currency pairs will probably be less reliable since less traders will trade them. A handful of traders may be enough to shift retail sentiment dramatically. Like I said, just a hunch.

Related posts

The statistical edge of trading against retail sentiment (from June 2014)
The statistical edge of trading against retail sentiment - part 2 (from July 2014)

Quick update on current trade stats

I've been trading with my current portfolio of systems since late May 2015.

The stats so far (I've normalised the risk level of all my past trades to 2%). 

Trades opened: 89
Win rate: 53,93%
Average Reward-to-risk: 1.06
Profit Factor: 1.17

My backtests show a historical profit factor between 1.10 and 1.30, so I'm performing as expected. 

Tuesday, October 13, 2015

RESEARCH: 4H Engulfing Candles on Tuesdays

I did some interesting research on large engulfing candles in the 4H timeframe. I found that Tuesdays seem to perform quite profitably. I backtested data from September 2012 to September 2015, with 11 currency pairs (AUDCAD, AUDCHF, AUDNZD, AUDUSD, CADCHF, EURGBP, EURJPY, EURUSD, GBPNZD, GBPUSD, NZDJPY, USDJPY). 

I defined a "large" engulfing candle as one that is 1.5 times larger than ATR(6), with the high and low completely engulfing the previous candle, AND closing outside the range of the previous candle. Example is below:

(click to enlarge)

My entry was the break of the engulfing candle's high or low, in direction of the trend (RSI(6) determined the trend). My stop loss was located at the opposite high or low.

The problem with this backtest was that there were relatively few entries. I collected 256 sample trades over the last 3 years. The results are below:

Reward-to-risk of 1 and below showed some profitability. So, for example, if my stop loss was 100 pips, I might want to aim for a reward of less than 100 pips (33 to 75 pips would seem optimal). 

Here's a basic equity curve for a reward-to-risk of 0.33 (the "optimal" reward-to-risk), using a starting a balance of $25.000, and 2% equity risk per trade.

Can this be a system?

Possibly. The backtest shows profitability between reward-to-risk of 1.25 and 0.25, so there's some flexibility in choosing your profit target. The only bad thing is that you can only trade once a week (only on Tuesdays). 

Saturday, October 10, 2015

WEEK IN REVIEW: 5 Oct to 11 Oct 2015

I ended this week at breakeven. I opened four trades, won one with a 1R reward, lost one, and got out at breakeven with two other trades. I entered a fifth trade on Friday, which is still open.


An engulfing candle formed on Friday, which I saw as a very bearish signal. On Monday, I went short at the break of Friday's low, and took profit at 1R a few days later.

(click to enlarge)


The EURCHF has been ranging for awhile, and it looked like it was primed to break out on the downside. On Monday, I went short at Friday's high using a sell limit order, and aimed for a reward of 1.25R.

Price moved close to my profit target before reversing back to my entry. Seeing a very bullish candle, I decided to play it safe and close the trade, getting out at breakeven.

(click to enlarge)


I saw a bearish pinbar on Monday, which signaled a continuation of the recent downward move. I saw some scope for a 1R reward, and went short at the break of Monday's low. An RBA news announcement on Tuesday saw price move against me shortly after triggering my entry, and I got taken out. 

(click to enlarge)


Tuesday produced a bearish pinbar during a downward trend, so I went short on Wednesday at the break of Tuesday's low. Price moved very close to my profit target before reversing to my entry point. I decided to get out at breakeven since price came agonisingly close to my profit target, and knew that I'd become tilted if I turned my near-win into a loss. 

If I had held on, the trade would've won. Was my decision to get out correct? Well, the bullish pinbar had formed near the recent major swing low. I think it was a safe decision. 

(click to enlarge)


I opened this trade on Friday, after seeing a bullish pinbar on Thursday. I wasn't concerned about major price level 185.000 as it's been repeatedly broken in the last month, so I don't think there'll be many orders clustered around here. 

Friday produced a bearish pinbar, but reversals are pretty normal on Fridays as traders exit for the weekend. I decided to hold on over the weekend. 

(click to enlarge)


Don't think there's much I can learn this week. When price comes really close to your profit target and reverses back to your entry point, is it prudent to close the trade and get out at breakeven? I'd say yes. The reversal shows that there is significant support / resistance just before your profit target, and momentum is now against you. 

Tuesday, October 6, 2015

RESEARCH: Monday Morning Breakouts on Asian Pairs

In my previous post, I found that "average-sized" candles tended to signal profitable breakouts on the AUDJPY and USDJPY during Monday morning.

The next step was to check if this pattern continued in other currency pairs, especially Asian ones. I backtested the AUDUSD, NZDUSD, AUDNZD and USDSGD from 2013 to 2015.

Here the results from 405 sample trades (this includes the AUDJPY and USDJPY):

Yep, nothing. The profit factor was less than 1 for all R:R ratios, so the pattern didn't hold. 

Further findings

I used RSI(6) to measure the trend. This indicator tells me the "trendiness" of the last 6 candles. If RSI(6) > 50, then the trend for the last 6 candles is bullish. And if RSI(6) < 50, the last 6 candles have been bearish.

I divided my results further into two groups: trades that opened WITH the trend, and trades that opened AGAINST the trend. 

Results WITH the trend (215 sample trades):

Results AGAINST the trend (190 sample trades):

Interestingly, trading AGAINST the trend yielded pretty good results. It may have something to do with the unique dynamics of Monday opens. However, my gut-feeling is that this backtest has become too finely-tuned. My next step is to step back and look at Monday opens from another angle (perhaps by looking at candlestick patterns). 

Sunday, October 4, 2015

RESEARCH: Monday morning breakouts on the USDJPY and AUDJPY

I just finished backtesting Monday morning breakouts on the USDJPY and AUDJPY. This is based on my findings that 12.5% of the weekly highs and lows of the AUDNZD were found in the first 4 hours of Monday, which is very significant. While the USDJPY and AUDJPY aren't eactly the same as the AUDNZD, they are cheaper Asian pairs to trade, so I chose to backtest them first.

My strategy rules were pretty simple: enter on the first break of the high or low of the first 4H candle on Monday, with a stop loss on the opposite end of the candle. I used RSI(6) to measure the trend, and ATR(6) to measure average candle size. An example setup is below:

I backtested the USDJPY and AUDJPY from January 2013 to September 2015, and gathered 287 trades.

Initial Results

These are the raw results, at various reward-to-risk ratios:

Nothing remarkable, although there seems to be a hint of profitability with smaller reward.


I did a rough optimisation, looking at whether the size of the first 4H candle made a difference. I divided the candles into three groups:

1. Small candles have a range less than 0.75 ATR(6)
2. Average candles have a range between 0.75 ATR(6) and 1.25 ATR(6)
3. Large candles have a range greater than 1.25 ATR(6)

For small candles (94 sample trades):

For average-sized candles (140 sample trades):

For large candles (53 sample trades):


Average-sized candles performed very well, with profitability greatest with smaller reward-to-risk ratios. Large and small candles performed poorly. 

Since I've found that average-sized candles do well, the next step is to test other pairs and see if this pattern repeats itself.

Saturday, October 3, 2015

RESEARCH: Time of Weekly Highs and Lows on the AUDNZD

I spent today examining the AUDNZD and the time of its weekly highs and lows. For this analysis, I focused on the 4H chart, using data from 2013 and 2014. 

What days can we expect to find the weekly high or low? Interestingly, the most common day is Monday, followed by Friday and Wednesday. The least likely day is Tuesday. Results are below:

If we zoom in on the data, we can see what time the weekly high or low occurred (within each 4H block. 00:00 = NY Close).

Of particular interest is that 21.63% of the weekly high or low was found during the first 12 hours of Monday, and 18.27% within the last 12 hours of Friday. This means that 40% of the weekly high or low occurred during the first half of Monday, or last half of Friday. 

Weekly highs and lows also featured prominently during Wednesday and the first half of Thursday. Approximately 29% of weekly highs and lows were found within this timeframe. 


How can we use this information? I suspect that Monday morning breakouts may be profitable. This is something to look at, as it gives you the opportunity to profit from both the daily move on Monday, as well as the move for the whole week. This will be my next project. :)

Tuesdays and Thursdays look best for trading continuations.

Trading reversal signals on Wednesdays may also work. 

Be wary if trading intra-day on Friday, as there's a very likelihood of reversal during the latter half of the day.

Friday, October 2, 2015

WEEK IN REVIEW: 28 Sept to 2 Oct 2015

This post is kinda early as it's Friday night and the markets are still open for awhile longer, but I've closed all my trades so I may as well post my results.

This week has been profitable, 3 wins and 1 loss. Details below.


I'd opened this trade last week, and went short. My profit target was hit on 30 Sept, Wednesday. 

I trailed my stop loss as I got another entry signal on Tuesday 29 Sept. Rather than open another trade, I decided to move my stop loss closer. 

(click to enlarge)


Monday produced a fat bullish pinbar. I aimed for a reward of 0.75R. Tuesday produced another entry signal (bullish pinbar), so I trailed my stop loss up to Tuesday's low. I was stopped out on Wednesday, but because I trailed my stop loss, I only lost half of my initial risk. Overall, EURUSD price action has been quite choppy.

(click to enlarge)


Wednesday produced a bearish pinbar, and I went short on the break of the low, hoping for a reward of 1R. Friday's NFP release moved price in my favour massively, but I intervened in my trade and took profit early at 0.75R. My reason: I wasn't aware of NFP at the time I closed my trade, and thought something unannounced had happened (like a terrorist attack). Lesson: stay aware of major news releases! Silly move. I left money on the table.

(click to enlarge)


I saw a bearish pinbar on Thursday, 1 Oct, and went short on the break of its low. Due to the proximity of the 1100.00 price level, I decided to "scalp" this trade and go for a small reward of 0.5R. It was timed almost perfectly. Price reversed shortly after I hit my profit target.

(click to enlarge)


Stay aware of the timing of major news!

Thursday, October 1, 2015

RESEARCH: Morning breakouts on the AUDJPY part 2

This is part 2 of Morning breakouts on the AUDJPY.

I finished backtesting the rest of 2013, 2014 and early 2015, and collected 412 sample trades. The backtest degraded quite badly. Here are the profit factors for various reward-to-risk ratios.

It's pretty much breakeven.

Wednesday continues to perform very badly. What happens if we omit Wednesday trades? We're left with 316 sample trades, which provided the following results:

The results are better, but profitability is too marginal for my liking. Maybe it can be improved with some discretion, like avoiding trades that are next to support or resistance.

Heh, this is what my typical day is like, testing new ideas and reaching deadends 95% of the time.