From theory to practice - page 829

 
Renat Akhtyamov:

Nah, the demo's gonna be okay.

On the real one, it's about three times worse.

It won't be 30%, but it'll be 10 but it's up, it's already a result

and if the stops, it will either be more than zero or less than that

50-50 as always.

 
Alexander_K:

What do I know?! Tick sample = const, and by time is variable.

Well, find out. That's for you, not me. The sample has to be commensurate with the duration of the trade. Otherwise "decision making" on it is a sham.

 

Data on the number of collected ticks for different currency pairs for the last 3 days:

Although working with ticks, on my TS, showed, on the whole, satisfactory results, but here is this desynchronization in the number, it surprises and infuriates me.

It turns out that each pair has its own timing, and that shouldn't be the case, of course...

Should we go back to OHLC M1 or what?

Yes, gentlemen, one can go crazy with this Forex.

 
secret:

Well, find out. You're the one who needs to know, not me. The sample must be commensurate with the length of the transaction. Otherwise, "decision-making" on it is a sham.

I did the math.

On average, I have ticks with the frequency of 1 tick = 3 sec. I.e. a sample of 3600 ticks, again on the average, is 3 hours. My trades last on average 1 hour.

Everything would be fine, but there are too many averages when working with ticks... I don't like it very much.

 

There is, after all, some secret to ticks (not to be confused with Bass!).

And switching to OHLC M1 instantly deprives all statisticians of meaningful intraday samples.

Some kind of mystery and unsolvable contradiction here...

 
Alexander_K:

There is, after all, some secret to ticks (not to be confused with Bass!).

And switching to OHLC M1 instantly deprives all statisticians of meaningful intraday samples.

Some kind of mystery and unsolvable contradiction here...

Well, if you only take the closing price from OHLC M1, what is the secret? .... You bind the price to a discrete time, so the continuous non timed tick stream becomes "non-native" to the bar

 
Alexander_K:

That is, a sample of 3,600 ticks, again averaging 3 hours. The trades, on average, last for 1 hour.

This is still relatively tolerable, gut. Although, it would be ideal if the time periods for deciding whether to enter a trade and for its exiting do not overlap

 
Alexander_K:

this unsynchronisation in numbers is what surprises and infuriates me.

It turns out that each pair has its own time, which should not be the case, of course...

It's absolutely normal.

You don't think that, say, the S&P index and the MICEX index should be tick-synchronous, do you? Each asset has its own trades leading to its own ticks.

Similarly, currency pairs. Although they are highly correlated with each other, they are not 100% correlated. There is a significant amount of individuality in each.

 
Igor Makanu:

Well, if you only take the closing price from OHLC M1, what is the secret? .... You bind the price to a discrete time, so the continuous non time based tick stream becomes "not native" to the bar

Once again - when you work with OHLC M1 using TViMS methods, you have to have a meaningful sample. According to Chebyshev it must be no less than 1000 values.

I.e. you have to work in a sliding window = 24 hours (1440 values).

I have been working in this window for half a year(!!!). I had max 2-3 trades per week. The result was +-0% profit. Unbelievable, just catastrophically boring! A normal trader should work in smaller windows. This can only be achieved by working with ticks. And they have another problem - different amount for different pairs and in different brokerage companies. This is bullshit...

 
Alexander_K:

There is, after all, some secret to ticks (not to be confused with Bass!).

And switching to OHLC M1 instantly deprives all statisticians of meaningful intraday samples.

Some kind of mystery and unsolvable contradiction here...

There is no mystery or contradiction.

Want to read ticks every second? You're welcome. Didn't have a tick in a second? Take the previous one, it's still valid.

Don't want zero ticks? Well don't read it if there was no tick in a second.

There won't be any particular difference compared to reading every tick.

But the window in seconds is still more correct.