Machine learning in trading: theory, models, practice and algo-trading - page 1293
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
Yes, 100% that after the flat there will be a trend. There is nothing to predict.
Well sort of yes, the question is when and how much, and in general there is not even uniqueness how to form targets for learning to recognize a trend and a flat
aaaa
If there are no ticks, it seems that the market is flat, 100%.
If there are a lot of ticks, then it is not a flatNo, the volumes and moreover the density of ticks are not linearly connected with the probability of trend<->fleet, more volatile areas may seem to be locally trending, but in fact they are not.
Well sort of yes, the question is when and how much, and in general there is not even unambiguity how to form targets for learning to recognize a trend and a flat
You don't need MO for that. It's just done by regular indicators.
I accidentally came across an investment company where NS trades. Judging by the description - self-learning neuronet.
Here is a printout of December trading on 6 instruments https://ita-lab.ru/sites/default/files/dekabr.pdf
You can see that they are trading on M1. There are 3500 trades per month, about 150 per day.
TP and SL are not fixed as they are always closed with different pips. Options of closing - either MM or if the direction from the new forecast does not coincide.
I have counted 86 losing trades for the first 200 trades, i.e. 43% mistake in determining the direction.
As a result, we can say that even with 43% error we can work to beat the spread and stay in the black.
You don't need an MO to do this. This is simply done by ordinary indicators.
Perhaps you misunderstood what I'm talking about, I'm talking about predicting market conditions, trendy or flat, indicators cannot do it, in some places it may seem to be true, in some places not, but on average - random and pareidolia. The target is of course done by the "indicator", but it is also not a trivial task, what exactly and why, what "by eye" seems to be a correct indicator for this purpose is not the fact that the best option.
PS: In general in the market "just" does not happen, when it seems that "just" is a bad sign, or yourself... Or someone wants...
Maybe you misunderstood what I'm talking about, I'm talking about predicting market conditions, trendy or flat, it cannot be done by indicators, somewhere it may seem to be true, somewhere not, but on average - random and pareidolia. This target is of course done by an "indicator", but it is also not a trivial task, what exactly and why, what "by eye" seems to be a correct indicator for this purpose is not the fact that it is the optimal variant.
When you are dealing with bees, you can't tell anything in advance. (c) Nothing can be predicted in the market.) Decisions can only be made in the process (i.e. your trend or flat should have already started, the process should go), but even in this case the decision will be probabilistic.
Yes, it's simple, you see with your indicators the beginning of a trend or a flat. Whether this condition will continue or not is another question.) And it's more a question of statistics than prediction.
When you're dealing with bees, you can't say anything in advance. (c) You can't predict anything in the market.) You can only make a decision in the process (i.e., your trend or flat should have already started, the process should be going), but even in this case the decision will be probabilistic.
Yes, it's simple, you see with your indicators the beginning of a trend or a flat. Whether this condition will continue or not is another question.) And this question is more about statistics than forecasting.
Well, the question is how persistent is the state of a trend/flit, how much less it "wobbles" compared to predicting the direction, it seems to me very noisy with it but I'm not sure I mark the right target for such predictions.
Well, the question is how persistent (stable) state of the trend/flat, how much it less "wobbles" compared to the forecast direction, I have a very noisy with it, but I'm not sure that I mark the right target for such forecasts.
And if not to be wise and look at the trend from the higher timeframe as the body of the candle, and the flat as its absence, it becomes clear that its prediction is essentially the same as the prediction of the direction.
Well, that's why it also sucks as a direction, I said that we are not talking about any "90%"
I accidentally came across an investment company where NS trades. Judging by the description - self-learning neuronet.
Here is a printout of December trading on 6 instruments https://ita-lab.ru/sites/default/files/dekabr.pdf
I can see that they are trading on M1. There are 3500 trades per month, about 150 per day.
TP and SL are not fixed, as they are always closed with different results in pips. Options for closing - either MM, or when the direction from the new forecast does not coincide.
I have counted 86 losing trades for the first 200 trades, i.e. 43% error in determining the direction.
As a result we can say that even with 43% error you can work to beat the spread and stay in the plus.
There's a report for less than a month
Even a year or two is not convincing either.
With a satisfactory test of 10 years or more, you can go to the real
I counted 86 losing trades for the first 200 trades, i.e. 43% error in determining the direction.
As a result, I can say that it is possible to work with 43% of error in order to decrease spread and stay in profit.
It is not possible, but necessary.
I wrote somewhere the other day that the system's zero return should be no more than 30-40% of successful trades. Otherwise the system is good for nothing.
Here, I found it.)
Forum on trading, automated trading systems and trading strategies testing
From Theory to Practice
Yuriy Asaulenko, 2019.02.04 10:56
I should use statistics. And the system should be such, that at 30-40% of successful trades it would be in profit at zero, and better in profit). On this basis, and the system should be designed, and not from a flashlight to draw averages and dispersions).
Actually, it is better to close not on fixed TP and SL, but on the analysis of the situation on the market and in the trade.