Whether there is a process whose analysis of one part does not allow predicting the next part. - page 17
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
If profitable trades happen, then it is theoretically possible to pull the system into profit by multiplying the lot.
Mathematically you are wrong:
Consider a random process called price: p[i]=sum{j=0..i; r[j]}, where r[i]~N(0,1) and E(r[i]*r[j])=delta(i-j). Consider a buy-and-hold strategy with profit locking at the end of each period: i.e. at time i a buy occurs, at time (i+1) a profit of r[i+1] is locked in. The probability of profit at time (i;i+1] is 1/2 (based on the kind of process that describes the price). Let us also introduce the capital management strategy m[i]=M(p[i], ..., p[0], r[i], ..., r[0])>=0. Given m[i], the profit at time (i+1) is m[i]*r[i+1]. Then we can write the Ito integral which connects the processes of balance, capital management m[i] and price p[i]. And then you can disprove your statement by calculating the average value of this integral, which is zero, regardless of the choice of process m[i].
1. If the sequence of transaction results (detrended by the payoff value) is not, so to speak, white noise, i.e. if there are correlations between transaction results. In this case, we need to find these correlations and use them (see below)
2. If there are no correlations in the sequence of deals, then it would be good to look for correlations with behavior of price prior to deals or with other factors, at least with the time of day (in other words, filters). Well, here the field for imagination is unlimited, in fact it is possible to create a new TS.
Do not limit yourself to linear dependencies. Analyze the joint distribution of profits on the trade at time (t-1) and the trade at time (t). Or a joint distribution with other factors (this is a generalisation of your point #2).
Mathematically, you are wrong:
Consider a random process called price: p[i]=sum{j=0..i; r[j]}, where r[i]~N(0,1) and E(r[i]*r[j])=delta(i-j). Consider a buy-and-hold strategy with profit locking at the end of each period: i.e. at time i a buy occurs, at time (i+1) a profit of r[i+1] is locked in. The probability of profit at time (i;i+1] is 1/2 (based on the kind of process that describes the price). Let us also introduce the capital management strategy m[i]=M(p[i], ..., p[0], r[i], ..., r[0])>=0. Given m[i], the profit at time (i+1) is m[i]*r[i+1]. Then we can write the Ito integral which connects the processes of balance, capital management m[i] and price p[i]. And then you can disprove your statement by calculating the average value of this integral, which is zero, regardless of the choice of process m[i].
Well... go learn some maths... sorry.
No offence :)
You don't have to restrict yourself to linear dependencies. Analyse the joint profit distribution of the trade at time (t-1) and the trade at time (t). Or the joint distribution with other factors (this is a generalisation of your point #2).
I have so far given up on patterns, at least on higher timeframes (H1 and above). More and more it seems to me that trading on H1 and above is essentially guessing the direction of the news. If consistent patterns exist, it's only intra-hours, on M1-M5 timeframes, that is, patterns of traders' reactions to news that has already come out. That's where you have to dig. After a long fiddling with higher mathematics in Forex (complicated formulas, regressions, Fourier, neural networks, etc.) I have come to my disappointment: it is not suitable for Forex. Pipsing with simple tools is much easier and gives more reliable results.
In contrast, I managed to get more stable results on H1. On M1-M15 the spread, if I may say so, is commensurate with the size of the spread, so there is a stupid struggle of the grid to overcome the load of the spread.
Or do you train the grid without taking the spread into account?
I, on the other hand, have managed to get more consistent results on H1. On M1-M15 the range of patterns, if I may say so, is commensurate with the size of the spread, so there is a dumb grid struggle to overcome the load of the spread.
Or do you teach a grid without taking the spread into account?
On M1-M15 I do not have grids. I do not use them at all. They are slow and of little use. Classic TA works faster and more stable. My interest in nets especially cooled down after I saw this:
https://www.mql5.com/ru/forum/6256
I don't have any nets on M1-M15. And I don't do nets at all anymore. They are slow and of little use. Classic TA is faster and more stable. My interest in nets has especially cooled down after I saw this:
https://www.mql5.com/ru/forum/6256
How did it affect your attitude towards nets?
And, what are these methods that allow you to pipsize more successfully and reliably (can I get a personal note) than working with grids? :)
Mathematicians, pull up here.
So:
Hypothetically there is a pouring strategy at a constant lot, i.e. pips MO is positive, but at a spread of 0.
1. Is it possible to choose such an MM (including martin derivatives), that the system would pour even at 0th spread, and what would such an MM depend on?
2. What would be the formula to calculate the maximum threshold value of spread, at which the system would not be able to fill?
You could probably use the grid itself in the settings to try to get a threshold that will respect the spread through its algorithm.
And it is possible without changing anything (because at zero spread +) to separately incorporate the mechanism for analyzing the spread, in particular movements on ticks, commensurate with the spread, but here we need asc.
I'm going to show a branch where Prival said that he found the value of signal/noise ratio = 1, at H from 1.6-1.8 of spread. i.e. the signal may be separated above the noise.
Maybe take this separately and screw it to the grid with separate logic. Yes the value within the spread (2 spreads) does not make the difference in a simple analysis, but it may be suitable for your case. After all, we already have the logic to make decisions, all we need to do is
reduce commission / payoff on spreads. Although, perhaps it is the opposite of your method, and found this value (and it is already the result) useful signal (as if on the contrary). That is, as in your point 2, he has found the way to determine this threshold ratio.
But in futures the real volume + real spread analysis will give even better results (not in terms of logic of entries but in terms of advantage to get in a profit/loss on entries, that is to make the loss null according to statistics more often)
https://www.mql5.com/ru/forum/115616/page14#154564 second post from the top.
Hello!
Life cannot be predicted in such a way as to make money on that prediction.
But: "You can't sell inspiration, but you can sell a manuscript" :)
If you are firmly planted on nets - look for a fast algorithm, which precedes a slow one. Imho, of course :)
Sounds like a call to analyse the timing of optimisation, which will then need to be combined with the price.