Indicator based on historical data analitics.

[Darek Pietraszewski]

I would like to ask for your opinion. I am currently analyzing historical data and using it to predict future customer behavior (non-FX industry). This is a so-called big data analysis. I am thinking about a scalping indicator that would be based on historical market behavior. Creating a history of the EMA shape (e.g. 9) and RSI (10 or 14) - let's say the last 5 years - and forecasting the shape of the EMA on the next 5 candles in a 5-minute time frame based on historical data/ shape of EMA and/or RSI. What do you think about such an indicator, has anyone seen anything similar? I assume that large institutions and speculators use the same patterns and repeat them, which is reflected in strategies such as SMC.

[Ryan L Johnson]

Darek1812:
I would like to ask for your opinion. I am currently analyzing historical data and using it to predict future customer behavior (non-FX industry). This is a so-called big data analysis. I am thinking about a scalping indicator that would be based on historical market behavior. Creating a history of the EMA shape (e.g. 9) and RSI (10 or 14) - let's say the last 5 years - and forecasting the shape of the EMA on the next 5 candles in a 5-minute time frame based on historical data/ shape of EMA and/or RSI. What do you think about such an indicator, has anyone seen anything similar? I assume that large institutions and speculators use the same patterns and repeat them, which is reflected in strategies such as SMC.

My initial thoughts are:

• There are "predictive" indicators out there already. They shoot futuristic line studies out ahead of current price:  Free download of the 'Fourier extrapolation of price' indicator by 'gpwr' for MetaTrader 5 in the MQL5 Code Base, 2010.07.05
• I don't find them very useful because, truth be told, I have no idea how to trade them. Price rarely follows the prediction to a T, so then what?
• Similarly, how exactly does one go about quantifying a 5 year shape of a line study? At some point, it has to be reduced to "black & white" code.
• Feel free to call me Captain Obvious, but I would rather simply look for statistically profitable interactions between prices, the EMA, and the RSI based on more typical lookbacks.
• I don't bother with scalping because I openly concede that I can't compete with the fatcats in Chicago. They're set up with FPGA chipped supercomputers right next to the exchange on which I trade. I'm definitely not, so I swing trade.
• I like where you said "patterns" and "repeat." If you can identify a repeating shape (pattern), power to you. The name of the game is statistics.
• I have to guess that "professional" predictive analytics probably involves more than an EMA and an RSI. I envision a team of quants, sitting in front of server-workstations, pushing the limits of machine learning/AI, in an office within walking distance to the exchange data center.

Well, that's my 2 cents. Now, I'm interested to hear your thoughts.

[Lorentzos Roussos]

Darek1812:
I would like to ask for your opinion. I am currently analyzing historical data and using it to predict future customer behavior (non-FX industry). This is a so-called big data analysis. I am thinking about a scalping indicator that would be based on historical market behavior. Creating a history of the EMA shape (e.g. 9) and RSI (10 or 14) - let's say the last 5 years - and forecasting the shape of the EMA on the next 5 candles in a 5-minute time frame based on historical data/ shape of EMA and/or RSI. What do you think about such an indicator, has anyone seen anything similar? I assume that large institutions and speculators use the same patterns and repeat them, which is reflected in strategies such as SMC.

Neural nets

Collect the data , you wont have to worry about the size of the window as the NN will "decide" on what to use just give it a decent amount of data.

Make sure it cannot cheat by memorizing specifics.

Normalize the data so that they look similar across different assets.

Train the network on a portion of your data.

Test it on the remaining portion of your data when it appears to have "learned" so that you check if it learned or memorized (like in school exams)

Embed it into the indicator if its good

[Conor Mcnamara]

I don't know if you like to accept the quote often repeated by brokers and platforms "Past results is not indicative of future performance".

[Michael Charles Schefe]

Conor Mcnamara #:
I don't know if you like to accept the quote often repeated by brokers and platforms "Past results is not indicative of future performance".

yes, but then, that is more about them trying to shed all legal liability and satisfying their legal responsibility to notify traders to do their due dilligence.

The last thing that brokers are: is nice enuf to tell me that I shouldnt give them a commission and open a trade, to save my "hip pocket" -- out of their nice natures.

[Ryan L Johnson]

Michael Charles Schefe #:

yes, but then, that is more about them trying to shed all legal liability and satisfying their legal responsibility to notify traders to do their due dilligence.

The last thing that brokers are: is nice enuf to tell me that I shouldnt give them a commission and open a trade, to save my "hip pocket" -- out of their nice natures.

Hear, hear!👏

Just imagine if this were factually true... repainting indicators only, gambling EA's only, and no backtester in MT5. And oh yeah... businesses would have a "nospectus" instead of a prospectus.

I have similar disclaimers on my own products.

[Conor Mcnamara]

Michael Charles Schefe #:

yes, but then, that is more about them trying to shed all legal liability and satisfying their legal responsibility to notify traders to do their due dilligence.

The last thing that brokers are: is nice enuf to tell me that I shouldnt give them a commission and open a trade, to save my "hip pocket" -- out of their nice natures.

But when you look at the state of US inflation always rising, obviously this quote holds more meaning. The past will not be the future.  Patterns in the past are history and the history doesn't repeat, it may appear to sometimes, but overall, not with accuracy. Both gold and bitcoin have seen all time highs in 2024

[Ryan L Johnson]

Conor Mcnamara #:

But when you look at the state of US inflation always rising, obviously this quote holds more meaning. The past will not be the future.  Patterns in the past are history and the history doesn't repeat, it may appear to sometimes, but overall, not with accuracy. Both gold and bitcoin have seen all time highs in 2024

I alluded to this too in Post #1 when I suspected problems in quantifying the shape of 5 years of price data, and I suggested more typical lookbacks. In this way, a good strategy would go into buy only mode and capture the intermediate upswings.

[Stanislav Korotky]

Darek1812:
I would like to ask for your opinion. I am currently analyzing historical data and using it to predict future customer behavior (non-FX industry). This is a so-called big data analysis. I am thinking about a scalping indicator that would be based on historical market behavior. Creating a history of the EMA shape (e.g. 9) and RSI (10 or 14) - let's say the last 5 years - and forecasting the shape of the EMA on the next 5 candles in a 5-minute time frame based on historical data/ shape of EMA and/or RSI. What do you think about such an indicator, has anyone seen anything similar? I assume that large institutions and speculators use the same patterns and repeat them, which is reflected in strategies such as SMC.

You probably want to look at this (actually many) one.