Machine learning in trading: theory, models, practice and algo-trading - page 441
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
A new version of R 3.4 was recently releasedhttps://cran.r-project.org/doc/manuals/r-release/NEWS.html
I like this one.
For now, the compiler will not compile code containing explicit calls to browser(): this is to support single stepping from the browser() call.
JIT compilation can be disabled for the rest of the session using compiler::enableJIT(0) or by setting environment variable R_ENABLE_JIT to 0.
enableJIT(3)
The majority of the "experts" in IR are actually tool collectors, spending all their time searching for and discussing new features, bibles or plugins for trendy IR applications, while solving real-world problems is put off for later.
spend all their time looking for and discussing new features, bibles, or plugins for trendy MO apps.
There's no way without it. The Perseptron from the 80's won't pull the forex. But modern gbm will. It is most likely that with the development of machine learning models forex is getting more complicated, and lagging even by 1 step can have bad consequences. If you go one step ahead in comparison with modern MO - it will be a grail for forex.
So what, I thought and decided to ignore all stereotypes and not to go back to my main belief. It does not matter how the model goes as long as it goes in the plus. The fact that I have about 9 degrees of freedom of orientation TC. Sometimes none of them leads to profit. But most often one of the degrees necessarily leads to a profit. Well, that's the way it is. If we rotate TS, then some of the degrees will earn something. Well, never mind.... So I decided to keep the spread for myself. There's no point in throwing it away. And voila, a break-even strategy is ready, with the proper risk... Of course, but in this case, the spread is OURS! Or even a little more. Here's today's report. The point is that the two NS work, which have not seen the data since 07.01. Ie since the beginning of the week period outside the sample. And the most interesting thing is that they work with pauses, what may be more reliable using the tester with its flaws. But pay attention to the trades, we take the SPRED, or even a little more.
The risk is really overestimated due to the fact that two pending orders are set relative to the signal, one at 0.00050 and the other at 0.00100. If the arrows are directed both ways, the minichannel is set to buy and sell, but with a positive profit of about 0.00100 points. It is not rare that both of them trigger, and as a result we earn the spread. Some form of CaryTrader. This is just for today. And the network is already running for the fourth day. Show?????
Okay, I will not bore you. And let me remind you that all work is done in pending orders. And you know why... Because the SPRED is ours!!!!! Not a drop to the enemy!!!! :-)
300% in four days.... Approximately, with a maximum drawdown of no more than 20% The question is how this strategy will feel during the next week......
Or at least tomorrow..... But the fact that since 07.01 they have not seen it is a fact. And the output I do not peek, I watch the purity of data collection.A new version of R 3.4 was recently releasedhttps://cran.r-project.org/doc/manuals/r-release/NEWS.html
I like this one.
You used to be able to call in codeenableJIT(3)
Okay, I will not bore you. And let me remind you that all work is done in pending orders. And you know why... Because the SPRED is ours!!!!! Not a drop to the enemy!!!! :-)
300% in four days.... Approximately, with a maximum drawdown of no more than 20% The question is in another, how this strategy will feel during the next week......
Or at least tomorrow..... But the fact that since 07.01 they have not seen it is a fact. And the output I do not look, I watch the purity of data collection.Tester data any will show. You put it on real or demo.
Or do not look at the behavior of the strategy in real time, and valking forward testing do for six months to a year ... So train on data till March and check it in March, then train on data till April and check it in April and so on. In real time, it is too long to wait to evaluate the performance of an EA.
The majority of the "specialists" in MO are actually tool collectors, spending all their time looking for and discussing new features, bibles or plugins for trendy MO applications, while solving the real problems is put off for later.
There is no way without it. The Perseptron from the 80s won't handle forex. But a modern gbm can already do it. Most likely, with the development of machine learning models, forex becomes more complex, and lagging even by 1 step can have bad consequences. If you go one step ahead compared to modern MO - it will be a grail for forex.
GBM is faster than MLP, no doubt about it, but I wouldn't comment on MLP, it's not so important in the market - features are more important, data is even more important.
I've only recently taken up MO... I have only recently become engaged in the MO... But the same behavior was observed with conventional Expert Advisors... I tried to use them, but they did not bring me anything useful. Nothing is found
Exactly! I think everyone had a similar period and more than one, I remember copying codes from kodobase, hoping to someday look through all the turkeys and owls, then collecting classifiers...
Now I'm sure that collecting high-level algorithms is an empty exercise, it's impossible to "quickly try" someone else's algorithm, it's a fake, one MLP is not possible to try for a hundred lifetimes, millions of nuances and pitfalls, the fact that you "tried" at the weekend means little, the person who wrote and rewrote this algorithm many times will get very different quality results than you and even if you take a fancy GBM :)
Either you know exactly what you need, and then write yourself faster, or suffer nonsense, something like TV before people flipping channels that would find something interesting, very tedious and inefficient work.