ATC without timeframe(TF)

 

Hi all.

I have an idea to move away from timeframes when creating the TS, so that the TS would not depend on TF, number of last examined bars, etc.

I've found only variants using ticks, but they are not easy to test in Metatrader (the results are very inaccurate)

What other variants, approaches, etc. do you know besides using ticks?

 
Igor Yeremenko:

Hi all.

I have an idea to move away from timeframes when creating the TS, so that the TS would not depend on TF, number of last examined bars, etc.

I've found only variants using ticks, but they are not easy to test in Metatrader (the results are very inaccurate)

What other variants, approaches, etc. do you know, besides using ticks?

The 5 will be extremely accurate. I have a scalper on ticks, doesn't even know about TF.

If I don't have ticks, I will use M1, for example, and work with it. What is the problem, I do not understand.

 
Igor Yeremenko:

Hi all.

I have an idea to move away from timeframes when creating the TS, so that the TS would not depend on TF, number of last examined bars, etc.

I've found only variants using ticks, but they are not easy to test in Metatrader (the results are very inaccurate)

What other variants, approaches, etc. do you know besides using ticks?

I take minute candlesticks and discretize them in my own way, so I don't bother with ticks. There are a lot of ways to modify minute candlesticks if you don't like them, and you will get quite adequate accuracy.
 
Igor Yeremenko:

Hi all.

I have an idea to move away from timeframes when creating the TS, so that the TS would not depend on TF, number of last examined bars, etc.

I've found only variants using ticks, but they are not easy to test in Metatrader (the results are very inaccurate)

What other variants, approaches, etc. do you know, except for using ticks?

If you don't use ticks, you will not have OHLC for different timeframes (candlesticks). All other views are based on ticks, they are primary. But... they are different for different brokerage companies and even for different types of accounts in one brokerage company. Also, like snapshots of real, not retail forex, contain several rates. For a complete idea about real forex the ticks of one brokerage company (account) lacks characteristics that can be obtained only with great work, collecting ticks from dozens of brokerage companies simultaneously. Then we will have both average and spread characteristics, all for the moment.

In addition, there is also an opportunity to catch micro-trends. This is a corporate effect that is not in the flow of ticks in one account, related to the fact, which of dozens of DCs is leading and which is lagging behind. If the last rate changes have already been in one direction for half of the DCs, we can predict that the next ticks from other DCs will change in the same direction. Only this significant work, in my opinion, nobody wants to do.

 
Vladimir:

If you don't use ticks, there will be no OHLC for different timeframes (candlesticks). All other views are based on ticks, they are primary. But... they are different for different brokerage companies and even for different types of accounts in one brokerage company. Also, like snapshots of real, not retail forex, contain several rates. For a complete idea about real forex the ticks of one brokerage company (account) lacks characteristics that can be obtained only with great work, collecting ticks from dozens of brokerage companies simultaneously. Then we will have both average and spread characteristics, all for the moment.

In addition, there is also an opportunity to catch micro-trends. This is a corporate effect that is not in the flow of ticks in one account, related to the fact, which of dozens of DCs is leading and which is lagging behind. If the last rate changes have already been in one direction for half of the DCs, we can predict that the next ticks from other DCs will change in the same direction. Only this significant work, in my opinion, no one wants to do.


I analyzed 4-5 leading brokerage companies on Matlab, directions coincide, levels don't, but it's probably because of different spreads.

 

A timeframe is simply a sampling over time. The refusal of a timeframe is the refusal of sampling and switching to a variable frequency of ticks processing.

There are no technical problems here - in the function OnTick() you just look if it is necessary to take the incoming quote, and if yes, you save it, and then you see if it is time to analyze trade actions and if yes, you do them. This is how Renco Bars work. It is quite simple.

But, timeframes are convenient, I do not understand why they should be rejected, and what are their advantages?

 
Georgiy Merts:

A timeframe is simply a sampling over time. The refusal of a timeframe is the refusal of sampling and switching to a variable frequency of ticks processing.

There are no technical problems here - in the function OnTick() you just look if it is necessary to take the incoming quote, if yes, you save it, and then you see if it is time to analyze trade actions and if yes, you do them. This is how Renco Bars work. It's quite simple.

But, timeframes are convenient, I do not understand why they should be abandoned, and what are the benefits?

Candlesticks were invented about 300 years ago by a Japanese merchant to visualize grain prices on the grain exchange of that time. And for visual representation and manual trading they are very good, despite their antiquity.

A robot doesn't need any candlesticks, because they castrate almost all the information inside the candlestick, leaving only the HL.

 
Alexey Volchanskiy:

Candlesticks were invented 300 years ago by some Japanese merchant to visualise grain prices in the then grain exchange. And for visual representation and manual trading they are very good, despite their antiquity.

A robot doesn't need any candlesticks, for they castrate almost all the information inside the candlestick, leaving only the HL.

Alexei, you are wrong. You yourself work with candlesticks on 1S timeframe (it seems not even with candlesticks, but with only one C price), and you say that "candlesticks are not needed".

The wheel in general was invented thousands of years ago by some troglodyte - and is still in widespread use.

I also didn't explain to you that any signal for representation is quantized and discretized. This is what the timeframe and candlestick representation actually does. All information inside the candlestick is not needed to work. If you do, move to a smaller timeframe.

 
Georgiy Merts:

Alexey, you are wrong. You yourself work with candlesticks on 1S timeframe (not even candlesticks, it seems, but only one C price), and say that "candlesticks are not needed".

The wheel in general was invented thousands of years ago by some troglodyte - and is still in widespread use.

I also didn't explain to you that any signal for representation is quantized and discretized. This is what the timeframe and candlestick representation actually does. All information inside the candlestick is not needed to work. If you do, move to a smaller timeframe.

If it is needed, and especially the order of High and Low in a minute (and other) OHLC (I would remove O and C from OHLC, and replace H and L with E1 and E2 - rate extremums in the order of reaching), then what to do? On 10-second timeframes there would be irregular bar skips, which would spoil both quantisation and signal sampling, they would often be meaningless. Ticks don't have this disadvantage, when they come, that's when they come. No illusions of regularity. And the properties revealed by OHLC analysis may well be meaningless, as may O and C themselves for minutiae.
 
Vladimir:
If we need, and above all, a sequence of High and Low in a minute (and other) OHLC (I would delete O and C from OHLC and replace H and L by E1 and E2 - rate extremums in the order of reaching), then what to do? On 10-second timeframes there would be irregular bar skips, which would spoil both quantisation and signal sampling, they would often be meaningless. Ticks don't have this disadvantage, when they come, that's when they come. No illusions of regularity. And the properties revealed by OHLC analysis may well be meaningless, as may O and C themselves for minutiae.

What do you mean "ticks don't have this disadvantage" ??? Ticks are just as irregular, and also have "bar skips", spoiling both quantisation and discretisation.

The purpose of bars is precisely to exclude redundant information from tick data. Quite a lot of TSs work on ticks, they can do without bars. But, all my tests show that such TCs are too unstable, the most stable are TCs on H1. All that is lower - it is mainly profit of DC.

I once saw "exemplary" reports of one scalper... Seems to be both real and profitable... But it turns out that brokerage companies have five times more profit. They seem to bring profit, but for a trader it is not so big, while for a brokerage company it is very significant.

For me the main profit is on larger timeframes, and it makes no sense to give up the candlestick representation.

 
IMHO. You can build your candles from ticks without reference to time. For example, you can build a candlestick consisting of 10 ticks. You can use a series of such candlesticks for analysis. This can be used for short-term deals and scalping. For medium and long term trades too much tick data will have to be saved for analysis. Therefore, I think the usual candlestick analysis will be quite appropriate.