Machine learning in trading: theory, models, practice and algo-trading - page 423
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
All right, I'll give you an introduction. Or as they say in such cases, learn the basics. Volumes are received in real time. For each bar, you know the traded volume, delta, maximum volume in the bar at a certain price, etc. You are confused with the volume at the end of the day, which is used to determine the market context.
Well, yes. You kind of brought up the point that if the output peeks into the input, you get unrealistically beautiful numbers in training, but not in real trading.
You mean the opposite again? ))) It should be if the entrance peeks atthe exit.
Do not invert your thoughts - this may confuse beginners. And correct the posts where it happened.
So? It may be a daily bar or an hourly one - what difference does it make? You know the traded volume only at the end of this bar. And you also know the closing price.
Especially when I work ONLY on formed bars, but the advantage of volume is that it can be traded up 2-3 bars before the movement begins. Read the BSA analysis, and go to the Cluster Dultu. You'll learn a lot. Not to mention the fact that in the CD you can see the weakness or strength of the buyers or sellers. Please read the above material, and then we'll talk...
Do you mean the opposite again? ))) It should be if the entrance peeksout the exit.
Do not invert your thoughts - it can confuse newcomers. And correct the posts, where it happened.
It is quite possible that I was mistaken. I apologize, but will not correct anything. I have no time, and do not want to. With all due respect :-)
Especially when I work ONLY on formed bars, but the advantage of volume is that it can be traded up 2-3 bars before the movement begins. Read the BSA analysis, and go to the Cluster Dultu. You'll learn a lot. Not to mention the fact that in the QD you can see the weakness or strength of the buyers or sellers. Please read the material above, and then we'll talk...
It's not really about the analysis. So what's the point of all this here? First of all, you need to tone down your arrogance. If we are already talking about the analysis, I personally recognize it only on the daily market. All that is below - just noise, nothing more. That's why I started to talk about the daily volume and daily closing. If you are fond of intraday piping and try to pull some kind of analysis out of your ears, I am sorry.
It's not like there's a branch on the analysis. So what's the point of all this here? For starters, you need to tone down your arrogance. If we are already talking about the analysis, I personally recognize it only on the daily market. All that is below - just noise, nothing more. That's why I started to talk about the daily volume and daily closing. And if you are fond of intraday pipsing and trying to pull some kind of analysis out of your ears, then excuse me.
You know I didn't mean to offend you at all, seriously!!! I was up all night and now I'm really pissed off. That's why my tone seemed rude to you. In fact, I'm kind and fluffy. It's just that the subject of volumes is already eaten up pretty good. And with regard to my model of cause-and-effect relationship I will say the following. When I was speaking about the volume trade, I wasn't referring to the volume of a specific bar, but to all kinds of volumetric patterns that can be formed over several bars and subsequently push the price in the right direction. The identification of the major players and their positions. The concept of point volume in a cluster. Read KD, they say quite interesting things there....
And on the subject of tinkering, I'm surprised this issue even came up here.... It's like sawing off the bough you're sitting on. Deceiving yourself......
You're right, however the problem is not as obvious as it may seem, as you can see from the articles of popular users of this forum, who use ZZ and similar indicators as a target and are happy with 70-90% accuracy of predictions.
I confess that I myself was ill with this disease not so long ago. I must confess that I myself was ill with this problem not so long ago. No wise books mention it, no forums discuss it, while trader's mind wants to believe in miracles. I was completely convinced only when I ran a training set on a random 10-100k samples which cannot strongly deviate from 50+-1% forecast.
Do you mean the opposite again? ))) Must be if the entrance peeks outthe exit
Don't invert your thoughts - it might confuse newbies. And correct the posts where it happened.
Everyone knows about classical peeking (input sees output), but Mikhail correctly wrote, you need to and THE OUTPUT DIDN'T SEE THE INPUT otherwise the model will simply predict not the output but that part of the input which is contained in the output, i.e. those 70-90% have nothing to do with the output, it's actually predicting the input which was NOT explicitly mixed with the output via ZZ interpolation or similar.
I'm telling you, it's a pretty subtle issue, not really obvious...
Everyone knows about the classical peek-ahead (the input sees the output), but Mikhail wrote correctly, you need the and EXIT DIDN'T SEE THE ENTRANCE Otherwise model will just predict not the output but a part of the input which is in the output, that is those 70-90% don't relate to the output, it's actually predicting the input which was NOT explicitly mixed with the output via ZZ interpolation or a similar inductor.
I'm telling you, it's a rather subtle question, not quite obvious...
It's kind of confusing what you're writing.
1) the exit should be built with a peek-ahead.
2) all entrances should be without peeking ahead
If these 2 rules are followed, the problem is that EXIT DID NOT SEE THE ENTRANCE and inverse input peepsoutput simply won't arise, since inputs and outputs are not related.
And if you could find an entrance without peeping forward which is connected with an exit (which with peeping), the grail is in your pocket).
Peeping, swiping.
Leave those human characteristics for other areas of study.
We have the original series of OHLCV quotes. All other, most diverse variables are generated from them by various functional transformations. To train the model we need to present in addition to the independent variables(predictors) - the target(i.e. the expected future value). This target value is always from the future. Regardless of the problem you are solving - regression or classification.
Let me try to translate this sentence"you need to andTHE OUTPUT DIDN'T SEE THE INPUT.You mean that the target should have nothing to do with the quotes?
Give me an example of such a target, just curious.
Chaos of mind...