Machine learning in trading: theory, models, practice and algo-trading - page 3104
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 the pain of such A/B testing in one video
(don't watch for the particularly impressionable)
Don't you think that in the beginning he talks about a binary example, but in the code he takes a range of numbers from a normal distribution?
And, all such tests depend on a random number generator, you should also take this into account. The nature of the appearance of the numbers may be more complex for each phenomenon, although it may have a normal distribution.
In this approach to the problem, as he shows, it is more reasonable then not to measure "days", but to evaluate the dynamics of arriving at the threshold and the time of being beyond the threshold.
In general, it would be more useful to observe the phenomenon on real data for any conclusions about it.
Don't you think he's talking about a binary example at the beginning, but in the code he's taking exactly the range of numbers from a normal distribution?
And, all such tests depend on a random number generator, so that should also be taken into account. The nature of the appearance of numbers can be more complex for each phenomenon, although it may have a normal distribution.
In such an approach to the problem, as he shows, it is more reasonable not to measure "days", but to estimate the dynamics of arriving at the threshold and the time of being beyond the threshold.
In general, it would be more useful to observe the phenomenon on real data for any conclusions about it.
It must be very difficult when you never know what you're talking about and the context.)
and the point here is that even if you have a trace and a test from the same distribution, you can't always confirm or deny a hypothesis, for example, about the robustness of the model. not to mention when they are from different distributions.
Add to that multiple tests and multiply the statistical significance of such tests by zero.
it must be very difficult when you never know what you're talking about at all? and the context )
and the point here is that even if you have a trace and a test from the same distribution, you can't always confirm or deny a hypothesis, for example, about the robustness of the model. not to mention when they are from different distributions.
Add to this multiple tests and multiply the statistical significance of such tests by zero.
I see. He's talking about the essence, and he's just hanging his ears....
I see. He's talking about the essence, and he's just hanging his ears....
you're missing the point.
you're missing the point.
What a person says there, making far-reaching conclusions, it is already clear, but he does not realise that the result depends on the algorithm of the random number generator.
The only conclusion that is valuable is that computer modelling in a primitive form does not allow you to get closer to real processes without understanding them.
If he divided the distribution into two parts with 0 and 1 and showed that he had an order of magnitude more units, I would be surprised. You see, he says one thing and does another.
I'm just now trying to predict the probability of distribution (I have classification) in the quantum segment for the next 3 months by the actual change of distribution in time, and by tests I managed to raise 15% of accuracy due to these metrics, and I think that this is not the limit.
Well, in general, the author of the video came up with a good "excuse" why he had an incorrect conclusion in the experiment. Yes, it's convenient, but it's not functional. I mean that from his words it is necessary to fix the time/number of observations and make a conclusion on it. In general, I really don't understand him here - what good is it, except for justification to the employer.
What the man is saying there, making far-reaching conclusions, it is already clear, but he does not realise that his result depends on the algorithm of the random number generator.
p-hacking is a known problem, which he tried to explain to nerds who didn't understand the essence.
can we not make tonnes of meaningless letters? it is already clear that there was no understanding.p-hacking is a known problem, which he tried to explain to nerds who didn't get the point.
can we not make tonnes of meaningless letters? it is already clear that there was no understanding.It's a pity that you still don't understand what I'm trying to convey to you.
All the pain of such A/B testing in one video
(don't watch for the particularly impressionable)
It's high time we all moved to the light side - to the matstat!)
The dark side, as always, opposes it) Dark in the sense that it always tries to reduce everything to the dark and unclear - in the extreme version to a certain "flair").
It's about time we all moved to the bright side - to the matstat!)
The dark side, as always, opposes it) Dark in the sense that it always tries to reduce everything to the dark and unclear - in the extreme version to a certain "feeling").