Trading Logic and Market Constants

Have you read H. Rearden?

Such a simple yet profitable concept.

I come back here time to time. Very valuable!

I'm understanding the removal of time as a market constant.

Would you please elaborate the rules in your image?! For instance, what does the "test" under the second point mean? Has the high/low take the high/low of previous bar out or under which condition (how many pips difference) has a high tested the previous days high?

Visit my thread.
You will find the answers there.

You mean "Highest Open / Lowest Open Trade"? Thank you. Seems that I have to read a lot today ...

It's been awhile since I rambled, below I'll talk a bit more about MCs (Market Constants).

Do note that this might seem esoteric, perhaps it is. I have mentioned before my methods won't be publicly available, however my thought process will be for those of you who are interested in looking at something through another's perspective.

Lets begin,
If one takes randomly generated Data (a large number of Data Sets), and compares it to Live Market Data (also a large amount of Data), discrepancies will be found between the two. By observing the differences we can find certain traits that Live Market Data has. Combining these traits with further Market Constants creates greater bias as to:
"what the Market will most probably do when ________ happens."
The Blank is a trigger based off of the Market Constants.

For those of you that wonder why I mention Market Constants so much, it's simply a method that I believe takes away most of the uncertainty from an uncertain event. When we look at any event with this type of logic, we end up at a good point to make further conclusions... take the below example, it may seem long winded, however there is good meaning to it:

Your job is to determine the probability of which way Two people will walk, and for how far, before they turn.

At first glance we know nothing about how Person #1 moves. We know that #1 is bound to move but how far and in what direction is unknown. We can attempt to follow them, we can attempt to say that because they have moved with XYZ pattern before they are bound to do the same now... we can use math, neural networks, moving averages, RSI you name it... but we simply are guessing using past behavior. At this point we know nothing about their character, their nature (Market Constants).

After observing #1 for sometime matters become worse... we find that this Person changes their movement from time to time, they become more or less volatile with their movements, they turn suddenly, they zigzag, and at times take much longer before they make a turn... They are a very uncertain individual however they are and individual who has characteristics, but these characteristics cannot be observed clearly without comparing this Person #1 to to something else.

So we have Person #2 now walking in front of you, We know for a fact that this person makes their turns on a random basis, in a random direction.

So what if we were to take Person #2 for numerous walks, and observed all these walks closely, listing out anything that is constant...

Now if we were to take our very uncertain friend Person #1 out for a walk what can be stated when we compare their walk with Random Person #2's? What characteristics of #1 remain constant? Write them down.

If we did not compare the two and just chose to focus on #1, we would simply be following a very uncertain person that we know very little about, prediction of their movement, and determining accurate probabilities would be futile.

Now this next part is going to make me sound a bit crazy,
When you look at anything in life to determine ALL of it's characteristics it must be compared to something else, for example if an individual was only ever capable experiencing one thing: take the Sun, they would not actually know what it is with great detail. Do they know the characteristics of heat, and light? To understand heat, one must also understand the contrast of cold, to understand light one must also know what darkness is... so a comparison using something of contrast is necessary to determine characteristics in detail.

The Markets are no different, they are an "individual" with some characteristics that can be observed easily, however there are more characteristics that remain hidden unless this "individual" is compared to something with contrast.

Anyways thanks for reading,
I say time and time again I've said that I write for the sake of others... this is true, however the full truth is I write for myself too. There is nobody that I know in real life that can understand this type of thinking, or be patient enough to discuss this type of thought process. It's funny how society can belittle and ignore uniqueness, the stranger something is the more their is too learn.

Lots of Love and have a great week everyone!
Rparm

Are you sure that the information price of the past does not influence in the future price?.
Are you sure that the information price of the past does not influence in the current price?.

If you are not sure you should remove it.

@Rparm: So, in simple words you compare a random walk (or another statistical model of market behavior, e.g. Markov processes) to real markets? But what exactly do you compare: Means? Average changes? Waiting times? There are so many parameters.

Hey AntiCre, you are correct when you say there is a lot to look at. I believe that the more characteristics one compares the more one learns about the nature of the Market.
The characteristics that one observes can be categorized into two groups:
Consistent - characteristics that remain constant regardless of what period of time is used (be it 50 years ago, or yesterday.. these "constants" do not change behavior)
Inconsistent - the nature of these characteristics is self explanatory.

Now when building a trading system the most important factor in determining the consistency of the system is,
the proper application of Markets characteristics that are Consistent, the reason is self explanatory.

One should keep a journal and document all things that are Consistent and Inconsistent, anything Inconsistent should be avoided when trading.

One of the Market Constants that I find crucial to my trading, something that I've mentioned a few times is the Fat Tailed Distribution ("FTD") that exists when looking at Live Market Data, this in itself is a major discovery. Something that has been known for a long time but often brushed aside. There are many people on this forum that know that the Markets are FTD however, one Market Constant is not enough to be profitable once Trading Costs are factored in. By finding other Market Constants that compliment each other one can further increase probability of success when entering/exiting a trade.

Finally, look at the Market Price from numerous angles and try and keep an open mind when comparing it to random Data.

the difference between a market's price and a random walk is quite simple, though fundamental

the price is generally moving from a target to another;

the random walk (assuming it 'random' very well) is just random walk

no matter how alike they look each other, they are so fucking different if you look closer.

but, even though a price is always (rarely fail to) moving from a target from another, only the hawk eyes may tell what exactly are the targets.

and why a price move this way, completely different from a random walk?

it is very understandable if you agree a price is the reflection of collective will of all market participants.

in other words, each market does not behave quite differently from a human being (or a group of them) does.

he/she (they) are generally living with a purpose.

Back after many months on another project.

Thats right... does this high or low have any constants in it?? Is there anyway to recognise them?? Thats just 1 pip move on the other direction

Alex, When you say Market is a product of nature, do you think there is some kind of fibonacci sequence effect/applied on markets?

Hi. It would be so kind to explain this part to me.

"- Hay 24 horas en un día en que una de esas horas contendrá el valor máximo del día. Probablemente, otra hora contenga el valor más bajo (este concepto se puede aplicar a cualquier período de tiempo)."

What I understand is that we must detect the high or low of the day when this is most likely we can enter into an operation.

Thanks.

The unfolding fundamentals are generally a result of past behaviours.

Everything is confined to certain laws.the random part is very real and is why most fail in this endeavour .if you can establish the the parameters of the whole,the atractors within those rules it becomes less random.

A drunk man walking home is a random walk ,it may take him 1,2 5 hours to reach his destination which is in fact a 45 minute walk.he or any observoir will not know his next turn,but he will 9 times out of 10 reach his destination.

These laws are a continuous recalibration but as they converge they become momentarily deterministic for reason of the boundaries of the laws we are bound within.

Yes this may sound like a load of horse shit but just like horse shit it has purpose

Every single thing that has ever happened has some influence on the filter.

[quote = Zelo; 9021682] {quote} Totalmente de acuerdo con CrucialPoint, el cerebro de cada persona está conectado de manera diferente. Cada persona mira y filtra el mundo por su propio sistema de creencias (acumulado desde el día en que nació). Tengo mi propia comprensión de por qué funciona el tiempo. En invierno, ¿obtendré beneficios vendiendo ropa de verano? Y viceversa. El tiempo es temporada. En forex, ¿Alguien es 'sesión asiática' COMPRAR CAD / CHF? El tiempo es sesión. Mercado abierto vs cierre de mercado. El tiempo es horas ahora. El tiempo influye en las personas y las personas influyen en el precio. El mercado se trata de personas, así que no hay certeza. Y como no hay una cosa 100% segura en este mundo (Black Swan), no me gusta la forma en que lo haces en negrita (el tiempo no influye en la sentencia de precios). [/citar]
I agree with your comment, time is important, most of all for that of the fundamentals, time does not say where the manipulation in the fundamentals will be, since at this specific time they accelerate the price towards the direction that the big ones want, Regardless of what the ad comes out, the ad is just an excuse, what they do in this particular time is to get the participants out.
Considering the above, time is one of the FOUR PRINCIPLES so all strategy is derived.
Greetings.

Nice thread... sad to see thread starter leaved FF.

Does anyone know how I can generate a random chart on meta trader? And could someone explain the concept of “fat tail distribution”?

thanks in advance to everyone who will answer.

I will post some other data when I finish compiling it. But I generated some random data along side of Daily Data taken from the G/U.

I created 3 sets of different random data, all generally based on the original set of data in some way. I just did some simple stuff I could do in excel. First way was the crudest way, which resembles price the LEAST amount.

But shown is an original segment of the data, along with its random data that was generated along side it. Is it obvious which one is the original?

First post about it, just includes some basic data.

HOW THE BARS WERE CREATED

There are probably much more sophisticated ways to create random price data, but I’ll offer a general explanation of the 3 types of data I generated. Each data type will have biases associated with it by virtue of how it was created. These biases are really just differences from real market data.

So, each price bar has its OHLC values. 1) These values are connected to their prior bar’s Close, and the current bar’s Open, there is a distance there.

Each price bar has an 2) O-C value, its open to close distance. This is the body of the candle. Its value can be zero to infinity.

Finally, each bar/candle will have a wick on the 3) top and 4) bottom of it. Its value can also be zero to infinity.

These four values where assembled, stacking the wicks onto the tops and bottoms of the body, and translated to standard OHLC values.

I worked out from my data set of real data, what the above 4 values typically were. I.e., it’s max, average, standard deviation etc, and modeled my data from it, so that it would produce various walks. All of the 3 simulants I created initialized at the same price as the REAL data.

Set 1: The real data set

Set 2: Take the min and max of the REAL data set, and select a number randomly in between. I.e., picking the O-C value… if the min was 2pips, and the max was 200 pips, it would randomly select a value in there. This data will have a bias of being evenly distributed, meaning a LARGE range is just as likely as a SMALL range for any part of the bar.

Set 3: This result is really not much different than set 2, but divided by 4. Initial idea I had was that it would make it appear more like the real data set… longer explanation.

Set 4: This is taking the original data set, and creating a NORMAL distribution around the mean of the original data set. I thought this would be the best set(closer to real price), it was not, it was too normally distributed. All bars are the same size. J

Lastly, the direction of each bar(the Open to Close, “+” or “-“) was chosen on each day. Using a random number between 0 and 1. If the random # is >0, the OC value is positive(green bar), vice versa for red.

I have an idea, that it might be reasonable to create a data set that is even less connected to the original data set, by no using ANY parameters from the original data set, and just building bars in the same way, but just setting ranges based on 0 – 100 to set the OC value.



COMPARISON OF THE DATA

There are probably quite a few things I could compare. There are observations that could be made on the micro level of the individual bars and how they open and close in relation to each other. There are broader visual observations of HH/LLs, trends, etc.

To start with, there is the quote,
“95% price will close above or below the open (price will move away from any reference point). 95% are candlestick 5% are dojis. 50% body 25% wick. 80% are breakouts 20% are inside bar 80% are breakouts 20% are outside bars (outside bars are breaks of S/R) GBP/USD - 90% are more than 40points in a London-New York Session”

I think the notable comparisons here, are to compare to the real data, and see how far off the fake data is. I will highlight the real data in red for these first comparisons.

For the dojis, My data is slightly different. My data was for OC distances <= 1 pip, this was 99%. I figure I would zero my data to those stats. If I use a filter of 4 pips, which is small for a Daily GU candle, I get closer to the 95% number. I will continue to use this number(4-pips), but I will show both here.

You can also see there are ZERO dojis in Set_4, which I thought initially would be the most similar to price.


“95% are candlestick 5% are dojis”

For the bodies, I get an average body size of 46% for the REAL data, close to 50%. The other data sets are shown in the image. This is generally one of the biases built into the data… and that is price from the simulated data is more wicky, in that TIME bars in real data will have less wicks involved when momentum is under way, particularly at the tops in an uptrend. My fake data builds bars with a higher occurrence of top wicks in uptrends. But this is a known bias in the way the data was created I suppose?


“50% body 25% wick”


For the candle types, these can be HH, LL, OB, IB. Similar to the doji stats, mine are a bit different. You can see I generally have 73% breakouts, 27% everything else(real data). This relationship holds true across the fake data also. Here, there seems to be a higher propensity for a bar to poke out of only one side of the bar only. The percentages do not vary by much.


“80% are breakouts 20% are inside bar 80% are breakouts 20% are outside bars”

For the range, it wouldn’t make much sense to compare the ranges of the fake data as it will vary wildly, but 92% of the samples had a range of greater than 58 pips(whole 24 hour period).

“90% are more than 40points in a London-New York Session”


Summary, it appears based on the quote above, set_3 appears to be most like the Real Data set. Set_4 the least like it. Which is more useful, I dunno?


OTHER DATA COMPARISONS


Changes in volatility, I have a thought that the fake data will vary less. It will be more uniform in nature. I think REAL data will spend more time in contracted ranges, and periods of volatility expansion.

There are probably better ways to test this, but I’ll just display a 12-period Std deviation(2 wks) of the bar’s returns. Lining it all up on top of each other. I was hoping that it would show different in changes of contraction and expansion. SEE FIRST PIC

I’m wondering if the measure of volatility(orange line) looks different on real vs fake data? I thought I noticed both a periodicity and a smoothness to the real data’s volatility line. To look at this, I zoomed out on the data bit, lined them all up and attached a second picture(zoomed out pic attached). Maybe the fake data is more oscilatory, but I dunno? SEE SECOND PIC

Something seems different about it I think? I have MANY more modifications of this, but I won’t bore you with them here. The zoomed out pic is in the attachments below the post. There is probably a better way to measure this and this isn’t telling me much.







Data put on a zero line. Looking at the bars(first image), all of which have their opens lined up on the zero line. Shown below. This is a smaller segment for viewability.

Just me looking at it, the “wicky” aspect of it can be seen, particularly on the down bars. Looking at the up bars, the set of real data seems to have a few more outliers, where the fake data is generally evenly distributed.




How likely are trends, and how long do they persist in the various data sets? For this I’ll compare the likely hood of multiple bars that close in the same direction. So if two bars close in the same direction, I am calling that the START of a trend. Below is an image of the rationale for tracking this metric, I think the image explains it all.




The scheme I am showing above for counting bars that may be part of a close-to-close trend… Looking at the chart above, considering a TREND existing when the count goes 1 or higher. I will also show the case for 2 or higher afterwards.

>>How likely are trends?

In other words, how likely is the close mixed or directional, -1, 0 or +1 as I have defined a trend above? Shown below are the distributions between -1,0 and 1. Here, among all 4 sets, 0 occurs pretty much dead @50%, meaning as I have defined the trend, half the time is spent in chop. Reminder, this is a trend count of ±1.

Also of note, is that the up vs down directions have a similar difference between the two, approx. 27% vs 23%.



As mentioned earlier, I said I would mention how many times the persistence of a trend count of 2 or higher. The image above shows the difference between 0 and 1.

If I combine bar counts of 0 and 1(trend count of ±2), that gives values for sets 1-4 as: 73.88%, 74.56%, 74.38% and 78.13%, as time spent in chop.

>>How long do they persist in the various data sets?
The above looked at how likely a trend exists vs chop. That was between 50-75% across all data sets depending on how the trend is counted.

This part is looking at how LONG, or how many bars each of the trends lasts. Not much here, other than it being less likely for trends to persist for very long.




Post 1/2

Post 2/2 Continued...

Up vs Down bars. Pretty simple.





Path Walked, for each data set. Below is an example of each data set, constructed into a candle. I.e., imagine those images above showing the candle stick charts(first image I posted), constructed into a bar.




The size of the 2nd bar is very volatile, skewing the scaling of the other three. So also attached is an image w/the second bar removed. Not much difference. They all START at the same price.



But just LOOKING at the candles representing the 4 different data sets, there is not much difference visually/subjectively.

End of the day, this is what we a trading, and looking at the PA at this far zoomed out, they look the same. I’ve shown some stats for the micro differences… but looking at the big picture, they can all be turned into a candle. Is there anything visual that distinguishes these candles from one another? What distinguishes a real candle, from a pseudo-random candle?