stevehopwoodforex.com
https://www.stevehopwoodforex.com/phpBB3/
Print view

Boulder's Grind
https://www.stevehopwoodforex.com/phpBB3/viewtopic.php?t=4743		 Page 5 of 6
Author:  Boulder [ Thu Jun 09, 2016 5:26 pm ]
Post subject:  Boulder's Grind
Here's GBPJPY today.
GBPJPY.JPG
Can you see how often BG would've worked out? :) Now of course it would incur some losses too. That's inevitable. But with good calibration, modest lotsizes, strict pip targets, tight market exposure, etc. we would come out net on top I believe. I've PM'd Steve and he assured me he would work on it whenever is convenient. Greatly looking forward to it :clap: :party: Cheers!

Aram
Author:  SteveHopwood [ Thu Jun 09, 2016 6:17 pm ]
Post subject:  Boulder's Grind
Boulder » Thu Jun 09, 2016 5:26 pm wrote:Here's GBPJPY today.


Can you see how often BG would've worked out? :) Now of course it would incur some losses too. That's inevitable. But with good calibration, modest lotsizes, strict pip targets, tight market exposure, etc. we would come out net on top I believe. I've PM'd Steve and he assured me he would work on it whenever is convenient. Greatly looking forward to it :clap: :party: Cheers!

Aram
Yes, he did and he will. Eventually.

Thing is, I am not entirely sure that your pm and your post are talking about the same thingy. I sort of understand your pm. I haven't really got a clue what your post here was describing.

Guys, don't get the wrong impression from a week like last week. It was half-term here in the UK, so I did not have to go out to schools to teach the piano. Instead, I could spend the week programming.

This week was back to school. Today was hot for us here in the UK. People in other parts of the world experiencing our temperatures would still be donning a third sweater, but I do not live there. I live in the UK. It was hot for us. I do not do heat at all well. Cold, fine. Heat? Forget it.

I spent the afternoon flogging through the heat to get to the school where I teach in The Shame of the Black Hole of Calcutta. Then I flogged back through the heat and arrived home knackered. I have the same motivation to plough through posts/pm's with incomprehensibly long paragraphs as an athematic Ant would have of climbing Everest whilst carrying a bag of heavy shopping.

Is what you are after a new ea? Is it an extension of what we already have? Please send me the big-print, short-sentence version of the answer to that question.

:xm:
Author:  Boulder [ Thu Jun 09, 2016 6:21 pm ]
Post subject:  Boulder's Grind
SteveHopwood » Thu Jun 09, 2016 8:17 pm wrote:
Boulder » Thu Jun 09, 2016 5:26 pm wrote:Here's GBPJPY today.


Can you see how often BG would've worked out? :) Now of course it would incur some losses too. That's inevitable. But with good calibration, modest lotsizes, strict pip targets, tight market exposure, etc. we would come out net on top I believe. I've PM'd Steve and he assured me he would work on it whenever is convenient. Greatly looking forward to it :clap: :party: Cheers!

Aram
Yes, he did and he will. Eventually.

Thing is, I am not entirely sure that your pm and your post are talking about the same thingy. I sort of understand your pm. I haven't really got a clue what your post here was describing.

Guys, don't get the wrong impression from a week like last week. It was half-term here in the UK, so I did not have to go out to schools to teach the piano. Instead, I could spend the week programming.

This week was back to school. Today was hot for us here in the UK. People in other parts of the world experiencing our temperatures would still be donning a third sweater, but I do not live there. I live in the UK. It was hot for us. I do not do heat at all well. Cold, fine. Heat? Forget it.

I spent the afternoon flogging through the heat to get to the school where I teach in The Shame of the Black Hold of Calcutta. Then I flogged back through the heat and arrived home knackered. I have the same motivation to plough through posts/pm's with incomprehensibly long paragraphs as an athematic Ant would have of climbing Everest whilst carrying a bag of heavy shopping.

Is what you are after a new ea? Is it an extension of what we already have? Please send me the big-print, short-sentence version of the answer to that question.

:xm:
Coming up Steve :good:
Author:  Boulder [ Thu Jun 09, 2016 7:45 pm ]
Post subject:  Boulder's Grind
Guys I apologise if my post was a bit too overloaded and not clear enough. To give you better idea what the EA will do specifically, these are the specs I sent Steve:

A. Main mode of trading = the stairs

Take a long if (analogously for shorts):
  • 1. 3 rising histograms

    2. Diff_1 between [U, V] & Diff_2 between [X,Y].

    3. Define Diff2-to-Diff1 ratio. Example: we specify Ratio = 2. When Diff_2 is larger than 2 times Diff_1, then don't take the trade. This is a safety mechanism against stop hunts.

    4. Don't enter a trade if the spread during the last X minutes went over Y pips. Suspend and resume trading after Z minutes.

    5. Allow only X long trades in sequence. Then suspend trading until a non-stairs pattern occurs. Example: we specify X=2, then imagine 5 falling candles and all the Diffs between them qualify for rule 1, 2, 3 and 4 at the same time. If this rule isn't there; the EA will take the 1st short after candles 1/2/3, 2nd short after candle 2/3/4 and 3rd short after candles 3/4/5. We want the EA to stop trading after the 2nd trade (even though qualified patterns occur). When at least 1 non-qualifiying pattern occurs, is when we enable trading again. So: only enable trading if the previous 1st, 2nd, 3rd and 4th condition are NOT met at least once. Again this is a safety mechanism.

    6. TP = e.g. 5 pips (preferably different from the secondary trades' TP)

    7. SL = 35 to avoid disaster (preferably different from the secondary trades' SL)

    8. If 6. and 7. didn't happen after G minutes; abandon trade (preferably different from secondary trades' CandleLifeTime)
B. Secondary mode of trading = momentum

Exactly the same as things are now. Only:
  • 1. TP problem remains to be solved
    2. the same spread-detection mechanism as in point 4 above should be added.
------------------
The best thing is for the EA to be able to run on both modes or either mode separately. When both are on, we just specify the Diffs in such a way that they won't interfere (we must do this either way :)). Example:
  • Primary mode:
    • Diff_1 = (0.4 ; 1.8)
      Diff_2 = (0.4 ; 1.8)
      => one TP amount
  • Secondary mode:
    • TradeLevel 1 = 2.4
      TradeLevel 2 = 2.8
      TradeLevel 3 = 3.2
      => incremental pip amounts if we specify it so.
Take care, Aram
Author:  Boulder [ Sat Jun 11, 2016 2:22 am ]
Post subject:  Boulder's Grind
I want to shed some more light on the stairs-pattern. To recapitulate; it is based on the Wilder's DMI histograms. A classic, text-book stairs-pattern looks like this highlighted in orange:
first.png
In order to capture it mathematically, we must first define Diff as the difference in absolute terms between a current DMI histogram value and its preceding DMI histogram value. A nice set of stairs has occurred when we have:

1. 3 consecutive rising (falling) DMI histograms
2. Diff between bar 1 and 2 greater than 0
3. Diff between bar 2 and 3 greater than 0

If those 3 rules are met, go long (short).

Now the Riemann sums reminded me of the method whereby one systematically divides an area in different partitions and sums them up to get an approximation for that area. The more partitions you have, the better an approximation you'll end up with. Have too few and you'll end up grossly overestimating the area. Now analogously I applied this to waves. Break a wave down with the help of Wilder's DMI in 'meaningful' partitions and try to extract its properties in terms of patterns. The stairs represent the transition of a top to the bottom of a wave and vice versa, like so:
second.png
As you can see, in green I've highlighted a potential reversal pattern. I have yet to find out whether these reversal patterns are robust and can be trusted/used. Let's concentrate again on the stairs. Have a look at GBPJPY today. This time I use timeframe 15 minutes and Wilder's DMI also put on 15 minutes.
Screen Shot 2016-06-11 at 03.26.40.png
Lets keep it simple and make abstraction of momentum trades and safety mechanisms I described in an earlier post. Enter shorts and longs precisely using those 3 rules at the top. Every time a new 15 min bar permanently prints, review the previous preceding 2 bars and make a decision. The relation between the 2 diffs doesn't matter at the moment. Just make sure to go long on the open of the 4th bar after 3 rising histograms bars. Similarly go short on the open of the 4th bar after 3 falling bars. Finally, abandon the trade after 30 minutes in the case of not hitting your TP or SL. We set TP at 15 pips and SL at 45 for instance. I drew a line of 15 pips in the screenshot for scale referencing. This is one of the configurations I want to test.

On the few occasions a SL would get hit, we'd need 3 wins to make up for it. A break-even or small loss/win is more likely - due to merely staying in for 30 minutes per trade - when your TP doesn't get hit. The big profits come when we have a swing and thus a sequence of stairs.

That was the base case of the main mode of trading. The numbers and statistics of how this scenario in isolation plays out are yet to be seen and to be tested with various configurations. Using a minimum Diff for example should enhance the performance by ensuring the EA to get in a trade when there's enough volatility going on.

Now we also enhance the probabilities of profitability by enabling momentum trading on top of the stairs-only trading. These are simply defined as 2 DMI histogram bars having a Diff bigger than a certain threshold value. In casu I'm using 5. So when a Diff > 5 is confirmed, momentum is taking place. You enter a long/short in the continuation direction immediately at the open of the next bar and hold the trade for 30 minute max. TP potential is between 20 to 30 pips. SL 40 to 60. I've highlighted the momentum trades with green.
Screen Shot 2016-06-11 at 03.24.13.png
Now unfortunately there are instances such as these w.r.t. momentum trades. They happened this week.
bad 1.png
bad 2.png
The first one contains 2 momentum trades; a SL hit and a break-even. The second picture contains 1 momentum trade which also has its SL hit. This is why it's important to have safety-mechanisms in place. The spread detection mechanism would've kept you out of these trades. You don't enter a trade when the spread went over e.g. 3 in the last 15 minutes. When it does, we suspend trading for another 15 - 30 minutes. Looking at these last 2 pictures also begs the question whether defining a super-range could be useful. That is, whether momentum trades with a Diff higher than a certain maximum value should be foregone. Have to research this further. Btw please do appreciate the beautiful stairs in both pictures which would've resulted in profit-takings.

Also understand that whether BG is profitable and to be trusted will only be seen after many many trades and sufficient testing. It will take both wins and losses. The question is whether over the long run the numbers will play out.

Cheers,

Aram

EDIT:
It is also important to understand that you'll have to use feasible TP and SL combinations. A TP = 10 and SL = 50 will require BG to take make 5 winning trades (main mode only that is) in order to break even. Set TP = 25 and you only need 2. Fortunately we're using this on GBPJPY which does allow you to employ higher TP values. If the statistics allow for it, BG should be profitable over time.
Author:  Boulder [ Sun Jun 12, 2016 12:05 am ]
Post subject:  Boulder's Grind
Rene pointed out a very important issue in a PM which is absolutely worth elaborating on. With his approval, you can read his question:
renexxxx wrote: In your template you use DMILength=55 and DMILevel=4, whereas Steve uses (OOTB) DMILength=14 and DMILevel=20. These two sets of parameters produce totally different histograms and I would expect different trade results as well.

So, the question is: what are the recommended settings for these two parameters?
My response:
Boulder wrote: Yes you're right we haven't yet elaborated on the DMILength matter. For some reason I've always been using DMILength = 55. Trying different values around that number gave me the impression that's it's rather a matter of scale. Seeing as we use Diff as the key variable, it shouldn't really pose big problems using different values. You would of course have to recalibrate the Diff-levels relative to the DMILength you're using.

Now what I may have missed is this: DMILength 55 generally requires smaller Diff minimum values like 0.10 or 0.20 to better ensure trading with volatility. When taking ratios by dividing by small values like this (e.g. for that stop hunt mechanism I talked about), it can be become quite inconvenient as the ratio values become rather uncontrollably large. Therefore it seems that shorter DMILengths should be better suited as they bring out more clear stair-shape forms. The shorter DMILengths will therefore require larger minimum Diff-values which in turn will enhance the use of the Diff2-to-Diff1 ratio. This is actually a major point. Thanks for bringing it up Rene! So in light of the above, I'd recommend staying with DMILength 14. With regards to the DMILevel, you can just ignore this as we don't use it in any way.
The implications of using a too high DMILength become apparent in the following scenario. For those interested in an even shorter timeframe; throw GBPJPY on a M1 chart. DMILength = 14 and DMI TF = M1. For reference put in a 2nd indi with DMILength = 55. It should become clear that DMILength = 14 is superior relative to 55. The Diff's are larger, clearer and more outspoken. This makes it easier to use the Diff2-to-Diff1 ratio for that stop-hunt mechanism. Now I'm not sure as to how effective this will be. Sufficient testing will tell. The rationale for this came to mind after quite some chart time and goes as follows:

When we take the steps (pun intended), we want to ensure that the steps are easy to step on. They should be congruent and rather symmetrical. Now we can allow deviations, but not too much. How do we do this?

Diff_1 and Diff_2 can't be:
  • too small; hence a minimum value to ensure volatility
  • too large; hence a maximum value to ensure not falling prey to relatively hard retracements/stop-hunts
  • deviating too much relative to each other; hence a maximum on the Diff2-to-Diff1 ratio. The perfect reference ratio = 1.
"Why not use a minimum for that ratio?", you may wonder. Because it is redundant. A picture will explain much better. Consider a base case of 3 falling histograms with values 9 - 6 - 3. Diff_1 = Diff_2 = 3. Here the Diff2-to-Diff1 = 1. This only happens in the perfect scenario.
Untitled-1.jpg
The dashed red line represents the perfect Diff_2 from which a perfect Diff2-to-Diff1 ratio (=1) will follow. The purple dashed line represents a Diff_2 that'll be bigger than Diff_1, in function of the angle between the purple dashed line and the red dashed line. The blue dashed line represents a very tiny Diff_2. There's no point in putting a minimum constraint here, for the minimum value we have in place will take care of this.

Having these constraints on will attempt to prevent BG to enter just after a stop-hunt and/or at point where price is likely to retrace/reverse. Look here for example:
Screen Shot 2016-06-11 at 13.05.40.png
Aim to stay in a trade no longer than 2 to 3 minutes.
  • Purple boxes: nice steps from which we could extract for example 3 to 5 pips.
  • Teal box: although the steps are symmetrical in shape, their absolute Diffs are too big. The risk of retracement is too high. We stay out. These also could figure as stop-hunts.
  • Orange box: a classic stop-hunt. Diff_2 is much bigger relative to Diff_1 (also in absolute sense). We stay out.
Having these in place may prove to be beneficial as they attempt to enhance the quality of the steps/trades. Also you won't be overloading your broker's servers. Furthermore we also have another optional constraint built in which will not allow more than X trades in the same direction until the stairs-pattern gets broken at least once. This also to help not going in when price is about to retrace. It'll all come down to finding the right inputs and combinations of constraints. It may for instance very well be that over time the constraints limit BG's potential to exploit an entire run downwards/upwards just because that run wasn't over nicely formed stairs. All of this yet remains to be seen. I'll be providing configurations and set files of course and be extensively testing them. Personally I'm keen to see how BG will do on the M5 and M15. Think I'll place my bets on those timeframes :). I feel like there's more to be discovered. I'll keep you posted.

Cheers,

Aram
Author:  SteveHopwood [ Sun Jun 12, 2016 1:53 pm ]
Post subject:  Boulder's Grind
Life is too short to try to incorporate Primary trading into the original version. The original version is now "Boulder's Grind Momentum" and there is a new one, "Boulder's Grind Primary".

I have added a spread filter to both.

Sparse explanation in post 1.

:xm:
Author:  Boulder [ Sun Jun 12, 2016 5:30 pm ]
Post subject:  Boulder's Grind
:clap: :clap:

Ok I have quite a few things set up now.

Primary trading:
Base M5 set 1:
  • DMI Configs:
    • DMITimeFrame = 5 minutes
    • DMILength = 14
    • DMILevel = 0
  • minimum Diff_1 = 0.5 and Diff_2 = 0.5
  • MaxRatio Diff_1/Diff_2 = 9999
  • TP = 15
  • SL = 50
  • MaxSpread = 3.3
  • MinutesToPause = 10 mins
  • TradeLife = 15 mins
  • EnforceNewSequence = false
  • EnforceAlternateTradeTypes = false
M5 set 2: Set 1 apart from TP = 25
M5 set 3: Set 1 apart from EnforceSequence = true
M5 set 4: Set 1 apart from EnforceAlternateTradeTypes = true
M5 set 5: Set 1 apart from EnforceSequence = EnforceAlternateTradeTypes = 1
M5 set 6: Set 1 apart from MaxRatio = 6
M5 set 7: Set 1 apart from MaxRatio = 12

M15:
M15 set 1: M5 set 1 apart from Diff_1 = 0.6 and Diff_2 = 0.6, TP = 25 and TradeLife = 30
M15 set 2: M5 set 1 apart from Diff_1 = 0.6 and Diff_2 = 0.6, TP = 35 and TradeLife = 30
M15 set 3: M5 set 1 apart from Diff_1 = 0.6 and Diff_2 = 0.6, TP = 45 and TradeLife = 30

It'd be good to have these M15's tested with tradelife = 45. All of them on MaxSpread = 3.3 and MinutesToPause = 15.

M30:
M30 set 1: M5 set 1 apart from Diff_1 = 1 and Diff_2 = 1, TP = 30, SL = 90, TradeLife = 90, MaxSpread = 3.5 and MinutesToPause = 30

I'm attaching my base set file for M5 as well. Feel free to use different settings as this will speed up the learning curve and be much more interesting for comparison later - especially concerning that stophunt mechanism. Also don't forget to check that you've got Wilder's DMI Timeframe and TradingTimeFrame set up correctly. Other than that have fun folks :).

Cheers

Aram

EDIT
Be sure to adjust you trading times as well. These work based on local time. I've got mine setup to kick in everyday during roughly mid-Asian session until the end of New York.
Author:  SteveHopwood [ Sun Jun 12, 2016 6:08 pm ]
Post subject:  Boulder's Grind
I posted a wrong version of the Momentum ea in post 3. Re-download if your version does not have the MaxSpread and MinutesToPause inputs in your "Wilder-related inputs section".

:xm:
Author:  manfloy [ Mon Jun 13, 2016 2:08 pm ]
Post subject:  Boulder's Grind
Thank you Steve :good:
All times are UTC Page 5 of 6

Powered by phpBB® Forum Software © phpBB Limited