Auction Market Value Theory & Analytics

I don't understand ... it's a coding problem? Because I think it' s simple to implement the right calculation in demo mode too.

It's not a problem, just matter of time. Thanks for the suggestuon, I will surely consider it for a next release

Thank you for your response. This must be interesting. Do you have in mind using your ea as a filtering or screening tool to trade manually. JP

Hi, thankyou JPworks.

My goal is to trade in fully automatic way. No manual actions at all.

I am also developing an enhacement to the EA, that is calculate also the Overlay at 10gg; this in order to have a full view of market conditions (overlay at 10,5, 3 days) before issuing an order.

I will share this enhaced version for your comments/contributions, I hope soon.

fabiodgt,
Auction Market Value Theory & Analytics is very tricky. There are so many variables outside the data analysis by in itself to make decision to trade if we have to follow the theory in this wonderful forum. I am looking into your method though I did not fully grasp it. Much appreciated. JP

JPworks,

you are right, so tricky and so wonderful.
I will try to describe the method I follow in the EA.
in steps:

1) calculate bracketing condition for the whole period e.g. 240 30 minutes bars (5 days). Note that the bracketing condition heavily depends on how "thick" price slices are (or TPO price width), and on distribution definition (what is the minimum number of TPO's that you allow in a single distribution? what is the minimum value for TPO?). If the distribution count gives 1, we say we have a bracketing condition
2) then, (in the next release...soon) just double the number of bars. if it was 240, make 580 (10 days). and again calculate bracketing condition. You are looking for a situation where bracketing is both on 10 and 5 days profile. you are looking for a stable market situation.
3) then, look at the last 3 days, and calculate all Direction and Activity parameters. Then compare these parameters with those in the previous three days. you obtain a variation indication both for Direction and for Activity, DIR and ACT. for instance, DIR>0 means the direction variation was up, and ACT>0 means the Activity, such volume or ATR, was getting stronger. you want a strong movement going up (if you are going long, in a similiar way for a short)
4) then, again three days comparison of Value Area variation. you are looking for a Value area variation "up", if you intend to go long.
5) then set a price value for you market order. you set it at upperlimit of 10 day profile. of course you should also checke that this limit is close to the upperlimit for the 5 days profile.

-->in a word, you are looking for a strong 10+5 day resistance to be broken with a sharp last 3 day up movement. stronger the resistance, hopefully bigger the up movement when it's broken

6) if price goes past such upperlimit, still wait at least 25 minutes. you want to enter in the market only if the new price is accepted by the market.

7) ok, now you have a position in the market....of course with take profit, trailing stop, trailing step and stop loss. let it go...unless, (if you want,)you get an opposite market signal ( e.g. a sell one). if you want, if you get a sell signal when you are long in the market with a positive profit, the position can be automatically closed.

Just a moment for a consideration: at what rate do you want to refresh all your data (bracketing, direction, activity...)? Is once a day at 24h the right way? I know that in the books "this is the way", but I am also thinking that nobody could afford a higher refresh rate when calcultaing in a manual or semiautomatic way all this data. Why not at a faster rate? You could get a benefit, take more opportunities exactly when they come, no delay. After all the forex market is always on. I really don't know, I believe that the answer to refresh rate is in testing: test it in both ways with historical data and decide what is better.

Of course, you can do all this not on a single currency (e.g. EURUSD), but in parallel on all the currencies. There is a machine calculating everything for you, so why not?

fabiodgt,
Very interesting. You dealt:
1. Bracketing - 5 and 10 days
2. Activity and direction - from 3 days
3. Attempted direction - I think that is what you mean in point 4 and5, I guess.
Now, how do you consider RI (Rotation Index) into the equation?
You are taking bold effort in the process of solving the complex. I hope you will give room in the future for manual trading as AMVT includes mental analytics too, beside the numbers. I am not sure what Mezvega's input would be. We are curios about your attempts though.
JP

Hi JPworks,

I did not yet consider RI, but there are already many indicators, adding a new one is not a problem.

About manual trading, I believe this is a key point. So my question is: what do you mean with " trading as AMVT includes mental analytics too, beside the numbers"?

I really believe that if there are mental rules, these can always be transferred to a computer.
What a computer cannot do is having a "sentimental decision", but I believe this would be very bad in trading.

So in my opinion, the only limit of a computer in trading is the programmer, that wasn't able enough to put the whole logic in to the program.
On the other hand, if you are able to transfer to a computer program all the logic, a computer will always be better than manual trading.

Faster, no mistakes, no emotions, no illness, never tired or angry or sad....and active 14 hour a day.

If you think this is not correct, could you give me an example of what AMVT does that a computer cannot do when properly programmed?

active 14 hour a day --> active 24 hour a day

))

Hello,

I made some improvements to the EA, and again I'd love to share it with you a demo version in order to have your feedback and suggestions.

The main improvement is that now the bracketing condition is checked:
- in last NUMERO_BARRE 30 mins bars
- AND in last NUMERO_BARRE*MULTI 30 mins bars, with MULTI a real number

an example to explain better: say that NUMERO_BARRE is 240 (5 days) and MULTI is 2.

-> the bracketing will be calculated:
- in last 5 days
- in last 10 days

If there is bracketing in both timeframes, and if the last 5 days are "contained" in the last 10 days (e.g upperlimit 10 days > upperlimit5 days or lowerlimit10 days < lowerlimit 5 days), then if in last 3 days ACT,DIR,COMP and Value Area variation comply with the condition, a virtual Buy or Sell is opened at a level which is the limit (upper or lower) of the long timeframe.

-> in a sentence, the EA is looking for a strong 10+5 day (e.g. with 240 bars and MULTI=2) resistance/support to be broken after a long bracketing period, with a sharp last 3 day up/down movement. Stronger the resistance/sypport, hopefully bigger the up/down movement when it's broken

the new input parameters in the EA are:
MULTI: a real number (e.g. 2.1) that multiplied with NUMERO_BARRE gives the number of bars of the long timeframe, where bracketing is checked and upper/lower limits calculated
multi_delta_perc: a real number (e.g. 0.25) that multiplied with the long period range gives the maximum delta between upperlimit in long period and upperlimit in short period for admitting buy orders (same for sell).


Following a description of the input parameters already existing in the previous version:


NUMERO_BARRE: it's the number of the last 30 minutes bars considered (e.g. 240 is 5 days). Bracketing condition, upperlimit and lowerlimit are calculated also respect to this timeframe
STEP: is the price step in point considered for calculating the overlay profile (e.g. 10 is 1 point)
TICKVALUEURUSD: tick value for EURUSD, has only meaning for testing and when using tickvalue correction (see below PIPVALUE). just leave defalut value
producifile, OutFileName: when testing, an historical report with meaningful overlay data can ben produced. just leave producifile at false
producireport, OutReportName, orareport: when running in demo, a daily report file with overlay data for each of the 28 equities can be produced. just leave producireport false
MIN_TPO: minimum number of TPO for each price for a distribution (e.g.3)
MIN_GAP: minimum number of price steps for a distribution (e.g. 4)
ore_wait_bracketing: oa virtual order will be issued if conditions are met and bracketing condition is true at least since this number of hours (e.g. 48)
NUMERO_CAMBI: number of equities considered. leave at 28;
lot: lot for orders
MAGICNUMBER: metatrader "magic number", unique for each terminal (e.g. 9090)
ORDER_BRACKET; this ea can also work for reverse orders. leave it at false, if somebody is interesed I can provide a description of the related parameters.
STOP_TREND: true (means trend orders are allowed)
NMIN: number of minutes "on hold" before opening a market order after price limit has been reached. for a buy, for instance only if the bid is > pricelimit for this number of minutes a real buy order will ben sent to the market. It helps in avoiding "false price spikes" when entering in the market
DIST: when a condition is met, a virtual order is opened. This is not an order in the market, only appears in Expert tab. the order, for instance for a buy situation, is set at upperlimit level of the 5 days Overlay profile. If DIST is different from zero, there is an offset in points respet to this level (e.g. 100, that is 10 pips above upperlimit)
FSTOPLOSS: all orders have stoploss. the stop loss is calculated as open price - (for buy) a fraction of 5 days Overlay price range. for instance, if FSTPLOSS is 0.5, stop loss is at a mid level between upperlimit and lowerlimit
FTAKEPROFIT. the same idea, for take profit
FTRAILINGSTOP: trailing stop is set, at a level that is a fraction of FTAKEPROFIT. For instance 0.5 means trailing step will start when reaching half of the level respect to take profit
FTRAILINGSTEP: trailing step is set. this is a fraction of trailing stop. for instance 0.5 means that the first trailing stop will be set at a level that is half of trailing level
NGIORNI: a virtual order will be canceled after this number of days, if not yet activated
UsaClose: if true, a buy order will be closed, if in profit, when a sell condition is met. same for sell
REF_ACT1, REF_DIR1: Overlay 3days are compared, last 3 days vs 3 days before. the result is Activity (ACT) ad Direction (DIR) calculation (all the parameters). if DIR>REF_DIR1 AND ACT>REF_ACT1 and Value Area variation is "UP" or "INSIDE",a virtual buy order at upperlimit (+DIST) will be opened. For a sell, if DIR<-REF_DIR1 AND ACT>REF_ACT1 and Value Area variation is "DOWN" or "INSIDE",a virtual sell order at lowerlimit (+DIST) will be opened.

COMP_REF: a compression indicator (number of TPO / Vale Area for the Overlay 3 days). comparison is done, AND the compression must be higher of this value. conditions are similar to aabove, just put COMP istead of ACT and COMP_REF instead of ACT1
DAY_HIGH. for reverse orders, leave false
PIPVALUE: tickvalue correction. if true, stop levels are changed for a factor that is the rate of pipvalue of this equity respect to EURUSD.
RATE: all the overlay data are calculated at this rate (in minutes). the results of the calculation for 3 days overlay is shown in Exper tab, when in demo environment. This value has no meaning when backtesting
TESTING_HIGH_RATE: when backtesting, overlay data can be calculated once a day (if this is false) or eache 30 minutes (if this is true). this affects a lot performance, so if you set at true you should only test with "openprices"

Just copy the .ex4 under MQL4/Experts and fill the set data how you prefere, following above indications.
in the .zip file you can also find a set that I've been using. results of such set in last 2 years with EURUSD adn USDJPY are here:
http://fabiotest.ddns.net/eurusd2y.htm
http://fabiotest.ddns.net/usdjpy2y.htm

I really look forward to receive suggestions/contributions or maybe other test results.

cheers
Fabio

I'm not sure what that has to do with calculating a bracket ?
What is a "multi" and what does it have to do with determining a bracket?

Exactly how are you calculating a bracket? what is the mathematical definition of a bracket? What is the logic you are using to determine if an overlay is bracketing or not?

I don't get it........overlays don't have "value areas"

I don't get it.........the foundation of amvt is that there is no day to day serial correlation.

@mzvega have you considered to move to futures market ? centralized market, a real and stable feed (the one's that comes from our fx mt4 broker is 100% made up by our brokers), real historical data, solid and stable trading platforms that will be surely around next 10 years, can trade currencys too .. ?

Hi mzvega,

thanks for asking, I am happy to try to explain better, see below my answers with >>


The main improvement is that now the bracketing condition is checked:
- in last NUMERO_BARRE 30 mins bars
- AND in last NUMERO_BARRE*MULTI 30 mins bars, with MULTI a real number
an example to explain better: say that NUMERO_BARRE is 240 (5 days) and MULTI is 2.
I'm not sure what that has to do with calculating a bracket ?

>>In order to see if you are in bracketing condition, you have first to calculate the overlay at a certain timeframe; for instance for the 5 days overlay you need 240 30 minute data bars (open, >>close, high, low, volume).
>>You set input parameter NUMERO_BARRE in order to define the number of bars involved in overlay and bracketing calculation.

>>Also, you want to check if there is a bracketing condition with a larger number of history bars, at a higher timeframe. For instance 480 bars a (10 days).

>>So, if you want to tell to the EA that you want to check bracketing condition with 240 bars (5 days) and also with 480 bars (10 days) you should use these input parameters:
>>- NUMERO_BARRE=240
>>-MULTI=2 (so that 2*240=480)

What is a "multi" and what does it have to do with determining a bracket?

>>See above. MULTI is used to multiply the bars in NUMERO_BARRE obtaining a second larger set of history bars, for higher timeframe Overlay.

>>One other example of input parameters and what they mean:
>>a) NUMERO_BARRE=288, MULTI=2.2 --> bracketing is calculated with 288 history bars (6 days) and with 633 history bars (more than 13 days)

>>The EA is looking for a condition when both the "short history set" (e.g 288 in this example) and the "long history set"(e.g. 633 in this example) show the bracketing condition. Bracket is >>checked on a first timeframe and also on a higher timeframe. Of course you can stay "fixed" to 240 bars and 480, but the EA is not limited to such values.


Exactly how are you calculating a bracket? what is the mathematical definition of a bracket? What is the logic you are using to determine if an overlay is bracketing or not?

>> first let me tell you that I checked my results with the MP-Multi-Composite-Profile indicator (in this forum) and all data are perfectly consistent. In order to check this you can just produce a >>report from the EA while backtesting, use for this the "producifile" boolean input parameter to TRUE. You will find the data file under yourterminal/tester/ with filename equal to what you set >>in input parameter "OutFileName" with a line of data for each day. If you launched a test for 100 days, the output file will have 100 lines.

>> so about bracketing calculation/definition: I calcultate max and min in the bar set (e.g. 240), then I divide this price range (max-min) in "slices" of width in points STEP (for instance >>STEP=10 is 1 pip), then I check the "weight" of each price step (that is the number of 30 min bars that cross such step). Bracketing means that the number of distributions is 1. A >>distribution has a minimum number of price steps (MAX_GAP, for instance 4) and has as boudaries upperlimit and lowerlimit, where the weight of price steps out of such values is lower >>than MIN_TPO (for instance, 3). For further details I should open my source code, and you may understand I cannot do it.




Quoting fabiodgt
a compression indicator (number of TPO / Vale Area for the Overlay 3 days).
I don't get it........overlays don't have "value areas


>> In order to calculatate DIR and ACT, and all indicators, I consider the overlay at 3 days. What I call " Value Area" is the price range around POC where there are 70% of TPOs respect to >>the total number of TPOs in 3 days overlay . So, about the compression indicator, I calculate the number of TPOs in the 3days overlay and divide it for the range of "Value Area" defined >>above. The Higher the result, the higher compression.





Quoting fabiodgt
Overlay 3days are compared, last 3 days vs 3 days before
I don't get it.........the foundation of amvt is that there is no day to day serial correlation.

>> this EA doesn't calculate indicators comparing each day, instead I calcultate a) the overlay of the last 3 days and b) the overlay of the previous 3 days. For instance, in order to calculate >>Activity, I compare each activity value (e.g. volume) between these two 3 days overlays. When you run the EA in demo, you can find in the Expert tab all such values (DIR, ACT...) both in >>numerical final result and elementary data.

>>I hope I made it a bit more clear how the EA works, if not please do not hesitate to ask

Your taking one day's "3day overlay" data and comparing it to the previous day's "3day overlay" data

You're still comparing one day's data (3 day overlay) to the previous day's data (3 day overlay)
That's serial…... day to day

(DIR, ACT...) aren't intended to be used to generate trading signals..........but that would defeat your purpose creating an ea, wouldn't it?
AMVA is a discretionary trading method...............

But that's exactly what your testing for..........
Good Luck with your endeavor!

eurnzd

nzdjpy

Thanks for your suggestion, very helpful. And for luck too, that is very important to us all, so my wishes back

I believe that any logic rule, of course not an emotional one, can be automated with an EA. And then tested with historical data to see if it works or not.

What would you suggest instead of using 3 day DIR + ACT + VA variation+ compression?

Yesterday my EA entered long with CADJPY (at 83.638), so far so good . What do you think about it?

My EA yesterday entered long on CADJPY (see link) at 10 day OV upperlimit.

https://api.asm.skype.com/v1/objects...mgpsh_fullsize

at the time of opening a virtual Buy Stop, some days ago, there was bracketing on 10 and 5 day OV. Now they show trend...

Hi mzvega, about NZDJPY it seems my EA has the same idea you have:-)

https://api.asm.skype.com/v1/objects...mgpsh_fullsize

err.....

Believe me. Your EA and her decision to trade that pair cannot be more different then apple and oranges, purely coincidental. It is not so simple to put the reference points to use without proper context. You have to see which reference points is relevant to you (after a lot of self study). The NET figures are over simplifying things but to those that are not aware of how this these NET figures came about, they originate from Jones's run-pause trading methodology which is an intraday trading system using 3-day overlay as minimum context. Those Net figures are a quick guide,not absolute. Still you have to discretionary decide given the reference point you elect to be of importance under various context of the market condition ( testing for trend, testing for end of trend, shape......etc.) in order to make a decision.

I just want to make that clear to those uninitiated who is still exploring this methodology.

As for the EA, i applaud your effort and your skills in programming MT4 which can be invaluable for those who like to pursue AMVT under MT4. But as we all know the potential for continual used of mt4 is questionable as eventually MT5 will take over one of these days; and we are all be left stranded. I started few years ago with the help of Bandung (indicator developer) to initiate the creation of 1st indicator, now known as ver. 2.1 with the help of mzvega and Bandung generously offer to as free). The 1st version was made redundant after Metetrader overhaul the syntax in MT4 after built 509. So as you can see the dependency on metatrader as a platform is not ideal. That is why e.g. mzvega have chosen to pursue AMVT with no such dependencies.

BTW, will you be putting this EA up for sale eventually?

Thanks Evak for your very valuable comment
I don't think my EA id perfect, I instead think it can be improved with valuable comments like this.


About NET, actually I believe you may be right. In detail, what I use at the moment is slightly different, the full (buy) condition is

- OV with NUMERO_BARRE is in bracketing (say 5 days)
- OV with MULTI*NUMERO_BARRE is in bracketing (say 10 days)
- OV "5 days" upperlimit is lower than OV "10 days" upperlimit

AND
(
- DIR> positive value, AND
- ACT> positive value, AND
- "3day OV Value Area change" UP
OR
- Compression > positive value, AND (compression I mean 3 day OV TPO / 3 days OV VA Range)
- Compression change positive, AND
- "3day OV Value Area change" UP
)
If all this is true, the EA issues a virtual buy stop at the 10 days OV Upperlimit;
if, after no more than N days, price reaches this value, and holds above it for at last NMIN minutes (e.g. 25), a buy order is finally issued.

In this moment I am also testing different conditions (e.g RF, VA only....) to see the effect.

I also memorized mzvega comment about serial correlation. Infact I compare 3 days data with 3 days data shifted -1 day, for DIR, ACT, VA, compression...

What I did, in order to check for any improvemement, was a new EA version where you can decide how many days of shift you use when comparing 3 days OV data.

In order to test it, I have set "even" take profit and stop loss (both at 0.25 of 5 day OV range), because I wanted to test a condition similar to tossing a coin, say 50-50 in balanced condition, measuring somehow "how good the EA is in entering a trade"; then I simply counted how many TPs and SL I had when testing.
I tested 7 equities from april 2016 till today, using both 1 or 3 days of shift when comparing 3 days OV data.

these are the results:


...............SHIFT1....SHIFT3
trade ok....55............32
trade ko....27............20
ok%.........67,07%....61,54%



So it really seems to me that my EA performs better using 1 day shift....this is opposite of what I was expeting according to mzvega remark!
Can somebody give an axplanation to this?


If somebody is interested I can post this new EA version...but after such experimental evidence I am still using the old one with shift =1 fixed

About MT4...I have been on mql since a while, I also used MT5 and mql5...finally I decided to stay on MT4, for now, for some reasons:
- MT4 is supported by all brokers, MT5 is not (my broker, Pepperstone, only supports MT4 for instance)
- MT4 is not going to be discontinued, for now
- MT4 has great "companion tools" for testing, for instance I use Tickstory that have very reliable tickdata from Dukascopy. I can easily test 5 years back...
- MT5 has tick history data from the broker, for instance I used it with ICM but I had to give up because data older than 1 months where corrupted....

I will move to MT5 maybe...but for the time being MT4 is great for me, and for a lot of other people too.


last, you ask If I ever considered to sell my EA...actually no, but who wants to discuss it can send me a private message.