(binned per thread starter’s request) MM (Money Maker) Detective | Page 174 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Money management: Capital needed per 0.07 USD per microlot 0 replies

The Volume Detective 411 replies

Attachments: (binned per thread starter’s request) MM (Money Maker) Detective
Exit Attachments

(binned per thread starter’s request) MM (Money Maker) Detective

  • #3,461
  • Jun 22, 2019 10:17am Jun 22, 2019 10:17am
Using Spikehunter and going through the charts I noticed that the buy and sell signals this indi produces are way more accurate in a ranging market then a trending one, has anyone also noticed this?
  • #3,462
  • Jun 22, 2019 11:14am Jun 22, 2019 11:14am
Quoting LongBTC
Using Spikehunter and going through the charts I noticed that the buy and sell signals this indi produces are way more accurate in a ranging market then a trending one, has anyone also noticed this?
@LongBTC -- I haven't been paying attention to that, but I will now.

And, you know, that probably makes sense. Since we are hunting for spikes, then in a trending market it might be that "spikes" are less inclined to reverse -- which would result in more false positives for. Certain envelope-types might work better than others when the market is trending -- e.g. envelopes whose bands are further apart. These types of envelopes would have fewer signals (i.e. fewer opportunities to trade) but, also, fewer false positives.

I'm currently adding a couple more envelope-types and then perhaps it might become clearer that certain envelopes work better / worse under different market conditions.
  • #3,463
  • Jun 22, 2019 11:16am Jun 22, 2019 11:16am
Quoting LongBTC
Using Spikehunter and going through the charts I noticed that the buy and sell signals this indi produces are way more accurate in a ranging market then a trending one, has anyone also noticed this?
That's to be expected. As with all band indicators, they are a reflection of the current expected deviation (cycle degree). A sloping band is a reflection of increased volatility and so to 'see the forest for the trees' you need to use a wider band (possibly higher TF) such that it will enclose the increased price deviation whereupon the market will once again appear to be 'ranging'. All markets are ranging and trending simultaneously but it's only within the ranges that we can make sense of the price stream.
  • #3,464
  • Jun 22, 2019 11:55am Jun 22, 2019 11:55am
Quoting skyway
{quote} AUDCAD is an example {image}
@skyway -- a few comments, see below...

Attached Image (click to enlarge)
Click to Enlarge Name: skyway-01.png Size: 29 KB


1. Time frame may make a difference. My preferred time frame is M15 because it fits within the ADR. As to why that would make a difference, I don't know. At this point it just seems an intuitive choice.

2. The envelope in the screen shot is using one of the new envelopes that I'm adding. As you can see, it appears to miss those two false positives. BTW, the chart does not show SpikeHunter -- only the centered TMA.

3. In the middle subwindow I attached the momentum indicator used by SpikeHunter. It would have identified and ignored the second false positive.

4. Both false positives did initially reverse. Using an indicator for the Exit condition -- like the RSX indicator I've mentioned previously -- you can see that an Exit condition would have been triggered a few bars later. We might have still won a few pips or, in the worst case, the loss would have been mostly mitigated. The main advantage of coding an Exit condition is that if your TP isn't reached, then at least there is a way out before the SL is hit.
  • #3,465
  • Jun 22, 2019 12:13pm Jun 22, 2019 12:13pm
Quoting skyway
{quote} AUDCAD is an example {image}
@skyway -- another point. Looking at your chart above one might ask why not use two envelopes, a "relaxed" one and a "strict"? And then only accept spikes that poke through both.

That might still be a viable solution, but one I wouldn't care for. The reason is because too many trade opportunities would be missed. It is good to avoid false positives, but not to the extent that your strategy becomes paralyzed and never trades. I mean, I could turn off my computer and that would be an effective way to avoid false positives. It all comes down to trade-offs. We'll never win 100%, but crafting the right combination of indicators should get us something profitable.
  • #3,466
  • Jun 22, 2019 12:18pm Jun 22, 2019 12:18pm
Quoting skyway
I overlay your spikehunter with mladen tma bands m15 and m5 timeframe using same parameters {image} {image}
@skyway -- I take it that you like mladen's tma bands? I do too. But they do repaint, but only subtly. Since we are hunting for price extremes, then the subtle repainting may not be an issue. And if the mladen tma bands dramatically cut down on the number of false positives then it surely is worth it.

I'm currently learning towards @Swingman's LazyTMA bands and/or mladen tma bands.
  • #3,467
  • Edited 5:55pm Jun 22, 2019 12:44pm | Edited 5:55pm
Quoting skyway
{quote} Look at this combination, mladen tma bands, your spikehunter and adr, smLazy signals {image}
@skyway -- as usual, a few comments...

1. Regarding ADR -- I used to use an "ADR Touch" algorithm based on @rosalieone's observations awhile back (it now seems like ages ago). But the problem is those ADR lines don't stabilize until around 0.5 - 0.75 fractional ADR is reached. That is, they keep shifting their vertical position during the first half of the day and are not a good signal. In hindsight they can be used to confirm that our strategy is accurately picking-out reversals, but I'm not comfortable using them as part of a trading signal.

2. Regarding fractals -- they require 3-5 bars to form. Since the trading window is very narrow for reversals, I think that is too long to wait. @Nih98 had suggested using the data from lower time frames to create fractals and I have tried that -- but it didn't work well in my hands. The bars from lower time frames make their own fractal patterns and I don't think those necessarily can be applied to a higher time frame.

3. I like how the mladen tma bands are looking. I also like @Swingman's LazyTMA. If we could come to some consensus that their subtle repainting is not an issue then it makes Entries sooooo much simpler. And combined with a momentum filter, then also very accurate.
  • #3,468
  • Jun 22, 2019 1:43pm Jun 22, 2019 1:43pm
Quoting robots4me
{quote} @skyway -- another point. Looking at your chart above one might ask why not use two envelopes, a "relaxed" one and a "strict"? And then only accept spikes that poke through both. That might still be a viable solution, but one I wouldn't care for. The reason is because too many trade opportunities would be missed. It is good to avoid false positives, but not to the extent that your strategy becomes paralyzed and never trades. I mean, I could turn off my computer and that would be an effective way to avoid false positives. It all comes down...
Or you could assign trade " conviction strength" tags to signals so that ie a grade #3 would mean a 1/3 size entry, grade #2 a 1/2 and grade #1 a full weight entry size etc.

Also, if you were already in a #3 level entry that went on to trigger a higher quality signal, then that could be an add-on point.

This is actually a pretty standard S.O.P as you are then stacking into a move when it continues to move in your direction, while effectively keeping any wrong move as a smaller loss.

Just a thought.
  • #3,469
  • Jun 22, 2019 2:07pm Jun 22, 2019 2:07pm
Quoting robots4me
{quote} @skyway -- I take it that you like mladen's tma bands? I do too. But they do repaint, but only subtly. Since we are hunting for price extremes, then the subtle repainting may not be an issue. And if the mladen tma bands dramatically cut down on the number of false positives then it surely is worth it. I'm currently learning towards @Swingman's LazyTMA bands and/or mladen tma bands.
@R4M...I also found a NRP version of the Mladen bands...but have not tested/used it yet.

Attached File(s)
File Type: mq4 TMA+CG mladen NRP.mq4   19 KB | 385 downloads


Nice weekend...
I never lose - either I win or I learn...
  • #3,470
  • Jun 22, 2019 2:14pm Jun 22, 2019 2:14pm
Quoting lemercenaire
{quote} Or you could assign trade " conviction strength" tags to signals so that ie a grade #3 would mean a 1/3 size entry, grade #2 a 1/2 and grade #1 a full weight entry size etc. Also, if you were already in a #3 level entry that went on to trigger a higher quality signal, then that could be an add-on point. This is actually a pretty standard S.O.P as you are then stacking into a move when it continues to move in your direction, while effectively keeping any wrong move as a smaller loss. Just a thought.
@lemercenaire -- sure, great idea. And that could certainly be incorporated. Currently we are working on identifying Entry signals. Once we've decided on how best to detect Entries then ranking them could be next.

Another thing -- you mention a metric called "conviction strength". In words, describe a rule that measures "conviction strength".
  • #3,471
  • Jun 22, 2019 2:15pm Jun 22, 2019 2:15pm
Quoting jdva
{quote} @R4M...I also found a NRP version of the Mladen bands...but have not tested/used it yet. {file} Nice weekend...
@jdva -- excellent, thanks. I will compare this one with the version I found.
  • #3,472
  • Jun 22, 2019 2:41pm Jun 22, 2019 2:41pm
Quoting robots4me
{quote} @lemercenaire -- sure, great idea. And that could certainly be incorporated. Currently we are working on identifying Entry signals. Once we've decided on how best to detect Entries then ranking them could be next. Another thing -- you mention a metric called "conviction strength". In words, describe a rule that measures "conviction strength".
Lol. People with computer-brains

By Conviction Strength, I simply mean how confident you feel that it's going to work. As we are saying that a certain break of the envelope (ie in the range style band) is going to be more likely to bring success, then that is Grade #1 etc.

The amount of grades will depend on how may different band styles you would use, it may only be two if we'd have a ranging entry and a trending entry for instance.

If price pokes it's head above the first style of band, it gets a grade #3 tag, with a trade position size of 1/3...
If it then goes on to the second style of band, grade #2, position size of 1/2...
Then if we see it finally have enough conviction to go to the third style of band, it gets graded #1 and position size - full.

Mmmmm...as I write that, it means you are going in 1/3 size, then adding 1/2 then full-size. That's obviously not right, it is adding a 1/3 size at each new level to end up with a full sized position. I am sure you realised that but it bears saying.
  • #3,473
  • Jun 22, 2019 3:06pm Jun 22, 2019 3:06pm
Quoting lemercenaire
{quote} Lol. People with computer-brains By Conviction Strength, I simply mean how confident you feel that it's going to work. As we are saying that a certain break of the envelope (ie in the range style band) is going to be more likely to bring success, then that is Grade #1 etc. The amount of grades will depend on how may different band styles you would use, it may only be two if we'd have a ranging entry and a trending entry for instance. If price pokes it's head above the first style of band, it gets a grade #3...
@lemercenaire -- okay, we're getting closer. Yes -- different bands get different grades. And how do we know the granularity of the band separation? And how do we know how to grade them? An indicator can not make discretionary decisions. The rules have to be exact. If we are to have multiple bands with different confidence levels then that will require a heck of a lot of testing -- including different pairs and time frames.

My intent is not to give you a hard time about it -- just that rules have to be very simple in order to code them. An indicator / EA will never be as good as a discretionary trader. But their advantage is automation. And it is up to the trader whether that trade-off is worth it.

For me it's a no-brainer since I'm content with "good enough". But for someone who requires 98%-100%, then indicators and automation will be disappointing.
  • #3,474
  • Jun 22, 2019 3:21pm Jun 22, 2019 3:21pm
Quoting robots4me
{quote} @lemercenaire -- okay, we're getting closer. Yes -- different bands get different grades. And how do we know the granularity of the band separation? And how do we know how to grade them? An indicator can not make discretionary decisions. The rules have to be exact. If we are to have multiple bands with different confidence levels then that will require a heck of a lot of testing -- including different pairs and time frames. My intent is not to give you a hard time about it -- just that rules have to be very simple in order to code them....
Oh, it's ok, I get that .

My take on it came from the comment about more successful entries coming in ranging markets, so going with that, I guess that a slope-angle detector would have to be involved and then an entry triggered just as it is now, by price breaking whatever band (or at least reversing back through it). No discretion in that side of things.

However, yes, it's at the point where the two aspects meet that will cause you problems!

Ah well, just spittballin', the idea can go back in the cupboard, lol.
  • #3,475
  • Jun 22, 2019 6:33pm Jun 22, 2019 6:33pm
Range Detector.

I have a Range Detector coded into an EA these are the rulers, but they could be changed to work with Bands

Range Detector = True/False

Range_Detector_Period_Zone = 200 (This means the EA will check 200 candles from where the MA's crossed (this will be the reference point) to make sure the Range Detector rules are followed before anymore trades are allowed within the Range Detector Zone, once price moves out of the Range Detector Period Zone then the EA would work like it does at present)

MA3_Pips_ Above_Reference_Point_Buy = 20 (MA3 MUST have been at some point a minimum of 20 pips above the reference point (the MA cross) for a Buy trade to be triggered if MA's 1&2 move back above MA3 in a Bullish direction when inside the Range Detector Period Zone)

MA3_Pips_Below_Reference_Point_Buy = 100 (MA3 MUST have been at some point a minimum of 100 pips below the reference point (the MA cross) for a Buy trade to be triggered if MA's 1&2 move back above MA3 in a Bullish direction when inside the Range Detector Period Zone)

MA3_Pips_Below_Reference_Point_Sell = 20 (MA3 MUST have been at some point a minimum of 20 pips below the reference point (the MA cross) for a Sell trade to be triggered if MA's 1&2 move back below MA3 in a Bearish direction when inside the Range Detector Zone)

MA3_Pips_Above_Reference_Point_Sell = 100 (MA3 MUST have been at some point a minimum of 100 pips above the reference point (the MA cross) for a Sell trade to be triggered if MA's 1&2 move back below MA3 in a Bearish direction when inside the Range Detector Zone)

The Range Detector Period Zone is ONLY activated by a Buy or Sell trade being triggered and the reference point is always the MA cross which happened just before the trade was triggered, if a new trade is triggered within the Range Detector Period Zone, then the Range Detector Period Zone reference point will change to the new MA cross position which triggered by the new trade. Once the Range Detector Zone Period has ended (in this example it is 200 candles) then the EA will revert back to opening trades based on the normal MA crosses like it does based on the current code, but as soon as a new trade is triggered then the Range Detector Period Zone rules will apply.

Example charts of how these rules apply to MA's
Attached Image(s) (click to enlarge)
Click to Enlarge Name: Range Detector.JPG Size: 243 KB
Click to Enlarge Name: Chart_showing_both_MA3_Pips_Above_Reference_Point_Sell_=_100_andMA3_Pips_Above_Reference_Point_B.png Size: 205 KB
Click to Enlarge Name: Chart_showing_MA3_Pips_Below_Reference_Point_Buy_=_100_rules_being_met.png Size: 207 KB
Click to Enlarge Name: Chart_showing_MA3_Pips_Below_Reference_Point_Sell_=_20_rules_being_met (1).png Size: 175 KB
Some people say I'm lucky, but the more I practice the luckier I get
  • #3,476
  • Edited 8:21pm Jun 22, 2019 7:49pm | Edited 8:21pm
https://www.forexfactory.com/showthread.php?p=12345083#post12345083

This podcast announcement is meaningless.

Why ?

It will be regurgitating the same old repeated charts that's already posted on his thread.

My opinion, you may disagree.
  • #3,477
  • Jun 22, 2019 8:35pm Jun 22, 2019 8:35pm
Quoting Robot Trader
Range Detector. I have a Range Detector coded into an EA these are the rulers, but they could be changed to work with Bands Range Detector = True/False Range_Detector_Period_Zone = 200 (This means the EA will check 200 candles from where the MA's crossed (this will be the reference point) to make sure the Range Detector rules are followed before anymore trades are allowed within the Range Detector Zone, once price moves out of the Range Detector Period Zone then the EA would work like it does at present) MA3_Pips_ Above_Reference_Point_Buy = 20...
@Robot Trader -- thanks. And it is a lot to digest. It's way over my head and much too complicated for me at this point.

But from glancing at your screen shots and explanations it reveals an approach that I wish to avoid and I'm glad to have the opportunity now to address this issue. So, @Robot Trader, it's good that you brought this up.

You'll notice there are rules saying to do this or that when a certain number is reached. I don't want to go that route -- and I'll explain why.

How many times have you heard people say their EA used to work okay but then it went bad? Really? The EA went bad? An EA is an algebraic formula. Algebra doesn't go bad. What happens is the price patterns have changed. When someone develops an EA they optimize the settings using their current broker data -- that's how the Range Detector developers came up with 20. But that number doesn't hold under all conditions, all pairs -- rather, it needs to be continuously updated. If people knew how to re-calibrate their EA's settings using their most recent broker data then there is a good chance their EAs would continue to perform well. The developers of most EAs typically use settings that are over-curve fitted for the data patterns at the time the EA was developed. Usually, but not always, EAs fail due to over-curve fitting.

I want to avoid hard-coded settings. The indicators we're using behind the scenes have default values that have stood the test of time -- so, SpikeHunter will not be exposing any settings -- even the StopLoss. I'm thinking of using ADR to calculate the stop loss -- since a stop loss that works for USDCAD may not be appropriate for GBPNZD. I'm looking to make SpikeHunter market and pair agnostic -- in other words, it performs roughly the same across all pairs and market conditions. The trade-off will be it won't be the greatest EA for GBPUSD or USDJPY, but I'm hoping it will perform "good enough" across all pairs under all market conditions.
  • #3,478
  • Jun 22, 2019 8:38pm Jun 22, 2019 8:38pm
Quoting robots4me
{quote} @Robot Trader -- thanks. And it is a lot to digest. It's way over my head and much too complicated for me at this point. But from glancing at your screen shots and explanations it reveals an approach that I wish to avoid and I'm glad to have the opportunity now to address this issue. So, @Robot Trader, it's good that you brought this up. You'll notice there are rules saying to do this or that when a certain number is reached. I don't want to go that route -- and I'll explain why. How many times have you heard people say their EA used to...
i learnt some thing new
people leave, but this thread will go on for ever
  • #3,479
  • Jun 22, 2019 8:49pm Jun 22, 2019 8:49pm
Quoting lemercenaire
{quote} Oh, it's ok, I get that . My take on it came from the comment about more successful entries coming in ranging markets, so going with that, I guess that a slope-angle detector would have to be involved and then an entry triggered just as it is now, by price breaking whatever band (or at least reversing back through it). No discretion in that side of things. However, yes, it's at the point where the two aspects meet that will cause you problems! Ah well, just spittballin', the idea can go back in the cupboard, lol.
Band levels depend on time frames. Stick with highs/lows, same across all time frames.
My Threads: Trading is as simple as 1-2-3, Highest Open / Lowest Open Trade
  • #3,480
  • Jun 22, 2019 9:01pm Jun 22, 2019 9:01pm
Quoting robots4me
{quote} @Robot Trader -- thanks. And it is a lot to digest. It's way over my head and much too complicated for me at this point. But from glancing at your screen shots and explanations it reveals an approach that I wish to avoid and I'm glad to have the opportunity now to address this issue. So, @Robot Trader, it's good that you brought this up. You'll notice there are rules saying to do this or that when a certain number is reached. I don't want to go that route -- and I'll explain why. How many times have you heard people say their EA used to...
Hi R4M,

I understand fully your comments and the charts I showed were from 2012 when I developed the EA and they are not curved fitted and they never stop working on any currency pair.

Yes the values for the parameters will change depending on the TF, but that is it.

The Range Detector monitors movement of PA if there is no movement then we are in a range, if PA starts to move then we are out of the Range and Trending. The EA constantly looks for NEW reference points and thus can never be classed has being curve fitting, because these reference points are Dynamically following PA in any/every market where PA in moving.

I am not the sharpest tool in the box, so it took me a while, but not 40,000 hours, to watch PA and come up with these parameters and they always work.

Regards

RT
Some people say I'm lucky, but the more I practice the luckier I get
1 trader viewing now