High & Low 8 Hours (GBP/JPY)

Quoting tjsun80

Disliked

Todday's high low:

high - 213.54
low - 212.63

Ignored

Quoting joyrider

Disliked

Hi tj,

Can you also post for EU, GU & EJ later? Thanks. BTW, is your gj based on 8 hr or 5 hr? The 5hr has more potential.

Ignored

Quoting shauny

Disliked

Hi,

Just to reconfirm the breakout timing this system uses. I read from the beginning till page 16. I saw someone posting 1700 GMT to 0100GMT. And the Nicotina uses 1800 GMT to 0200 GMT. Which is the correct timing??

Ignored

Quoting tjsun80

Disliked

Today's high low:

high - 213.54
low - 212.63

Ignored

Quoting tjsun80

Disliked

sell stop is triggered.

Ignored

Quoting sweety

Disliked

+100 on sell order

Ignored

Quoting sweety

Disliked

+100 on sell order

Ignored

Quoting ilanr

Disliked

First of all, huge thanks to Nicotina for sharing, to saaedb for the EA, and to all the constructive contributors to this thread for suggesting ideas. Esp. to cesarnc and pino2, they were absolutely correct in suggesting the 100 p. SL. In my experience, using anything less than 80+ p. SL for GBPJPY us suicidal.

I did some optimization on this strategy. I don't use MQL, so I used the data from MQ and a programming language I know. The data were 15 min prices from Jan 1999 to Nov 2007.

The bad news is that, as noted previously on this thread, 2007 was the only good year for this system in its original form. The good news is that it can be easily adjusted to work consistently well. The optimized rules are:
1. Identify the minimal and the maximal price of GBPJPY between 1900 and 0300 GMT.
2. Don't trade if the range (max-min) was less than 32 pips or greater than 173 pips.
3. If the price touched the max or the min, order a buy limit 17 pips below the max and a sell limit 17 pips above the min. The orders are valid until one of them is triggered, then the second one is cancelled. If no order is triggered by 1030 GMT, both are cancelled.
4. Set SL at 115 pips from max/min, set TP at 112 pips from max/min. All these levels refer to the bid prices, account for your broker's spread when opening a long or closing a short.
5. No re-entry if SL or TP hit. If neither of these happened, close the position on 1900 GMT of the third day (i.e., 72 hrs after the 1900 in para. 1).

If you'd have blindly used these rules every day for the last 9 years, you'd earn, on average, 47 pips per trade, winning 67% of the trades (considering for a 7 pip spread, not considering for swaps and slippage), about 170 trades per year. There are some differences between weekdays: Mon-Wed are about 50 p., Thu is 45 p., Fri 35 p. Contrary to what some people may think, the NonFarm Fridays are better than others (53 pips with 73% win ratio). 2007 would still be the best year (70 pips per trade), but other years would fare not bad either (from 31 p. in 2005 to 62 p. in 2000).
There were 4 losing months, 3 of them with 100-200 pips lost, one was worse (-570 p. in Sep 2003). 3-4 consecutive losses are common, 5-6 in a row can sometimes occur, the worst strike (Sep 2003) was 9! So, be careful and use proper MM. And, of course, it should be borne in mind that, like any optimized system, its future results will be worse than those described above.

Some more good news. Many people like GBPJPY these days because of the big pips it yields, but the following example shows that there might be better options. I optimized Nicotina's rules for GBPUSD, which is strongly correlated with GJ, in the following way:

1. Identify the minimal and the maximal price of GBPUSD between 1900 and 0600 GMT.
2. Don't trade if the range (max-min) was less than 19 pips or greater than 155 pips.
3. If the price touched the max or the min, order a buy limit 11 pips below the max and a sell limit 11 pips above the min. The orders are valid until one of them is triggered, then the second one is cancelled. If no order is triggered by 1630 GMT, both are cancelled.
4. Set SL at the opposite border (at max for short, at min for long positions). Set TP at 140 pips from max/min. All these levels refer to the bid prices, account for your broker's spread when opening a long or closing a short.
5. No re-entry if SL or TP hit. If neither of these happened, close the position manually at 1900 GMT of the same day.

This set of rules would bring you, on average, 32 pips per trade (67% win ratio). Yes, 32 p. are only 68% of the 47 p. GJ can give you, but with GU, you risk on average (3 pip spread included) 28 pips, which are only 23% of what you risk on GJ! Given this, and considering for the shorter average time that it takes GU to bring you profit or loss, your return on investment per day is 76% higher. I'll add to this that GU profits are very stable across weekdays and years: Only 1 losing month since 1999 with -100 pips… GU breakout is also insensitive to trend, while if GJ breakouts against the main trend, it yields half the pips compared to a trend-following breakout. However, mind that GU can hit SL 4-5 times in a row, with 150-200 pips of cumulative loss, so proper MM is important here as everywhere else.
I'm sure it's possible to improve this system further…

Good luck to all.

Ignored

Quoting ilanr

Disliked

First of all, huge thanks to Nicotina for sharing, to saaedb for the EA, and to all the constructive contributors to this thread for suggesting ideas. Esp. to cesarnc and pino2, they were absolutely correct in suggesting the 100 p. SL. In my experience, using anything less than 80+ p. SL for GBPJPY us suicidal.

I did some optimization on this strategy. I don't use MQL, so I used the data from MQ and a programming language I know. The data were 15 min prices from Jan 1999 to Nov 2007.

The bad news is that, as noted previously on this thread, 2007 was the only good year for this system in its original form. The good news is that it can be easily adjusted to work consistently well. The optimized rules are:
1. Identify the minimal and the maximal price of GBPJPY between 1900 and 0300 GMT.
2. Don't trade if the range (max-min) was less than 32 pips or greater than 173 pips.
3. If the price touched the max or the min, order a buy limit 17 pips below the max and a sell limit 17 pips above the min. The orders are valid until one of them is triggered, then the second one is cancelled. If no order is triggered by 1030 GMT, both are cancelled.
4. Set SL at 115 pips from max/min, set TP at 112 pips from max/min. All these levels refer to the bid prices, account for your broker's spread when opening a long or closing a short.
5. No re-entry if SL or TP hit. If neither of these happened, close the position on 1900 GMT of the third day (i.e., 72 hrs after the 1900 in para. 1).

If you'd have blindly used these rules every day for the last 9 years, you'd earn, on average, 47 pips per trade, winning 67% of the trades (considering for a 7 pip spread, not considering for swaps and slippage), about 170 trades per year. There are some differences between weekdays: Mon-Wed are about 50 p., Thu is 45 p., Fri 35 p. Contrary to what some people may think, the NonFarm Fridays are better than others (53 pips with 73% win ratio). 2007 would still be the best year (70 pips per trade), but other years would fare not bad either (from 31 p. in 2005 to 62 p. in 2000).
There were 4 losing months, 3 of them with 100-200 pips lost, one was worse (-570 p. in Sep 2003). 3-4 consecutive losses are common, 5-6 in a row can sometimes occur, the worst strike (Sep 2003) was 9! So, be careful and use proper MM. And, of course, it should be borne in mind that, like any optimized system, its future results will be worse than those described above.

Some more good news. Many people like GBPJPY these days because of the big pips it yields, but the following example shows that there might be better options. I optimized Nicotina's rules for GBPUSD, which is strongly correlated with GJ, in the following way:

1. Identify the minimal and the maximal price of GBPUSD between 1900 and 0600 GMT.
2. Don't trade if the range (max-min) was less than 19 pips or greater than 155 pips.
3. If the price touched the max or the min, order a buy limit 11 pips below the max and a sell limit 11 pips above the min. The orders are valid until one of them is triggered, then the second one is cancelled. If no order is triggered by 1630 GMT, both are cancelled.
4. Set SL at the opposite border (at max for short, at min for long positions). Set TP at 140 pips from max/min. All these levels refer to the bid prices, account for your broker's spread when opening a long or closing a short.
5. No re-entry if SL or TP hit. If neither of these happened, close the position manually at 1900 GMT of the same day.

This set of rules would bring you, on average, 32 pips per trade (67% win ratio). Yes, 32 p. are only 68% of the 47 p. GJ can give you, but with GU, you risk on average (3 pip spread included) 28 pips, which are only 30% of what you risk on GJ! Given this, and considering for the shorter average time that it takes GU to bring you profit or loss, your return on investment per day is 76% higher. I'll add to this that GU profits are very stable across weekdays and years: Only 1 losing month since 1999 with -100 pips… GU breakout is also insensitive to trend, while if GJ breakouts against the main trend, it yields half the pips compared to a trend-following breakout. However, mind that GU can hit SL 4-5 times in a row, with 150-200 pips of cumulative loss, so proper MM is important here as everywhere else.
I'm sure it's possible to improve this system further…
Good luck to all.

Ignored

Quoting ilanr

Disliked

...

Ignored

Quoting tdion

Disliked

Now we're talking...

Do you mind posting the 01-07 backtests for GU and GJ that you described?

Ignored

Quoting billbss

Disliked

ilinar,
Great work!

I have a question.

You indicated that you used 15M data.
Scrolling back through 15M charts, I find it impossible, on many days, to determine whether a trade was triggered or not.
On GU, a 15M bar can have a range of 20 or 30 pips. On GJ it can be 40 or 50 pips.

Many times it is impossible to determine if the H/L was touched first or if the retrace point was touched first, never to be touched again.

How did you deal with this in your backtesting?

Thanks

Ignored

Quoting ilanr

Disliked

Billbss, you are absolutely correct: there is no way to know. Using 15m data is, for me, a compromise between accuracy and feasibility of running thousands of optimisation iterations. Evidently, at least part of the trades were not triggered. At the same time, I don't think that those that weren't triggered are more or less profitable than those that were, so the overall conclusion re. profitability should be valid. In my opinion, it is especially true when we compare systems or system versions using the same data set. What do you think?

Ignored

Quoting billbss

Disliked

backtesting the parameters you have arrived at using 15M data, with lower TF data (5M or 1M) might reveal any major discrepancies in the results.

Ignored