Machine learning in trading: theory, models, practice and algo-trading - page 1460
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
I'm testing it with 50 bars H1 at the moment and it seems to be better than with 10.
In general it is necessary to try different indices for the forest and observe the result. Maybe I will get to trend indices in time. I also tried Zigzag, but did not get 32% like Alexey did ( Probably he has some kind of special ZZ, or maybe he peeks.
The best results are obtained when there is information about long-term trends, i.e., taking it into account, for example, from the balances to feed short-term transactions
otherwise it always turns out 50-50 for me personally
i need to think what exactly are the patterns in quotes, do you have an answer? i, for example, am pretty sure i do.
i.e. how it differs from random
in general terms it's cycles, no distinct cycles - a general trend. That's all :)))
I have not got the 32%, but it did not work like Alexey's. Apparently he has some special ZZ, or he does peep.
Most likely you have not found optimal ZZ settings on the sample for training, including the optimal TF, or better several TFs.
On Si the settings turned out to be fractal, i.e. I was looking for them only for m1, but maybe if you search for each TF the result will be better.
Also, not the fact that this works for all trading instruments, so try where I did my sample - on the futures Si, and if you see the result, then try other instruments.
After stock trading, my DC quotes look awful - that's just an observation....
...
What is the ratio of buy to sell?
Most likely you have not found optimal ZZ settings on the sample for training, including the optimal TF, or better several TFs.
On Si the settings turned out to be fractal, i.e. I was looking for them only for m1, but maybe if you search for each TF the result will be better.
Also, not the fact that this works for all trading instruments, so try where I made a sample - on the futures Si, and if you see the result, then try other instruments.
After exchange trading, my DC quotes look awful - that's just an observation....
What is the ratio of buy to sell?
That experiment I have already safely closed and forgotten). Now the new features are being tested.
I'm still experimenting on EurUsd M1.On training, as it should be very good, on the test, if the balance became horizontal, then according to SL/TP = 80/120 = 0.67
I successfully closed that experiment and forgot about it). Now the new features are being tested.
No, I asked about something else - the number of buy trades and the number of sell trades, their ratio.
It is enough reliable to judge from the ratio about the presence of fitting, if it significantly differs from 1.0.
No, I asked about something else - the number of buy trades and the number of sell trades, their ratio.
You can fairly reliably judge from the ratio the presence of a fit, if it differs significantly from 1.0.
The best results are obtained when there is information about long-term trends, i.e., taking it into account, for example, from the balances to feed short-term transactions
otherwise it always turns out 50-50 for me personally
i need to think what exactly are the regularities in quotes, do you have an answer? i, for example, am pretty sure i do.
i.e. how it differs from random
in general terms it's cycles, no distinct cycles - a general trend. That's all :))
Finished the forest with 50 bars H1. I.e. this is long-term information for M1.
I have some profit on forward, but sometimes the balance is in the drawdown for 2 or 3 months. Psychologically it will be difficult to stand it, most likely I will switch this kind of forest off as unsuccessful. I want the drawdown to be not more than a week, or even better within a day.
That is, I should look for a solution without long-term features.
I have finished testing the forest with 50 bars H1. I.e. it is long-term information for M1.
I have some profit on forward, but sometimes balance is in the drawdown for 2 - 3 months. Psychologically it will be difficult to stand it, most likely I will switch this kind of forest off as unsuccessful. I want the drawdown to be not more than a week, or even better within a day.
I.e. I have to look for solution without long-term features.
What do you mean with 50 bars? bar prices or what? it turns out not a trend, but incomprehensible in the interpretation of the forest
build trends over time, with different periods (or 1 for starters), subtract them from the price - these will be fics, as many as you want, let's say 500 last differences
on the new data, the trends should continue as a f-from time (backwards or forwards, makes no difference). You can take the time in seconds from the beginning of 1970 or whatever year. That is, you linear (or slightly polynomial) model the time - it will price you. Accordingly, you can calculate the historical variance. If it is exceeded, disconnect the bot (like the model has ceased to adequately show the trend).
There is nothing else to think about (and cannot think about), if it is not some specific strategy such as a breakdown of local levels or return to the mean level in the flat
I've tried a lot of different ways, this one may work miserably
I've tried a lot of different ways, this one might not work so well. I have not got it on python, R may work.
What do you mean with 50 bars? bar prices or what? it turns out not a trend but incomprehensible in the interpretation of the forest
Yes - the difference between the current price and those 50 bars.
I haven't tried all the easier ones yet. I will postpone trends for the future.