Machine learning in trading: theory, models, practice and algo-trading - page 3192
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
Just give me a long trading history, all right?
if you have found your inefficiency - you can only rejoice, but it does not cancel that the market is a sb or almost a sb.
You don't need trading history, you don't even need to prove profitable trading....
It's enough to calculate the probability.
For example, the man in the picture has 4 trades
1) entry point at the best price
2) takek let's say 6 ticks
3) 4 trades in a row
4) in a certain time
If the price is random, how many random attempts did the author of the picture have to make to get these 4 trades?
Do a simulation on random price, or do random trades on real price and calculate the probability of getting such a result.
I think the probability of winning the lottery wouldn't be far off.
Trading history is not needed here, you don't even need to prove profitable trading....
It's enough to calculate the probability
For example, the man in the picture has 4 trades
1) entry point at the best price
2) takek let's say 6 ticks
3) 4 trades in a row
4) in a certain time
If the price is random, how many random attempts did the author of the picture have to make to get these 4 trades?
Do a simulation on random price, or do random trades on real price and calculate the probability of getting this result
That's the point, you can get them randomly as fast or as long as you want, there's no limit to it.
maybe it was four random profitable trades in a losing streak of 1000 attempts.
What are you talking about, it's not serious.
That's the thing, you can get them randomly as fast or as long as you want, there's no limit to it.
Well, he's got a lot of pictures like me, so I believe him.
It's not a one-time thing or a two-time thing.
If we use the analogy of flipping a coin, the man flipped the eagle 30 times in a row...
Yes, this can happen with a 50/50 chance of playing the game for a long time, but it can not happen 5-10 times in a short time or even in a long time.
Well, he's got a lot of pictures like me, so I believe him.
It's not a one-time thing or a two-time thing.
If we use the analogy of flipping a coin, the man flipped the eagle 30 times in a row...
Yes, this can happen with a 50/50 chance of playing the game for a long time, but it can not happen 5-10 times in a short period of time.
and you just don't save pictures with bad results? )
profitable series happen in casinos too, at that moment the player starts to think that he has caught luck by the tail.And you just don't save pictures with bad results? )
I don't save... and with good ones I haven't saved for a long time, it doesn't change anything....
The point is that if the market random to such beautiful pictures could be 1-2, not 20-50.
You can't in a 50/50 random throw out the eagle 30 times in a row and do it 30 times in a limited time.
I don't save... and I haven't saved with the good ones for a long time now, it doesn't change anything....
The point is that if the market random to such beautiful pictures could be 1-2, not 20-50.
there could be any number of any pictures and on random.
The catbusta - by the quantum table is going overboard :)
Maybe try to implement the table at the same time for compatibility?
I have attached a file according to CatBoost's standard - the first column - number of predictor, and the second - split.
But I don't use it. The result does not improve.
By the way, if you set 65000 quanta, as a rule there are not so many different values of a chip and the results absolutely coincide with non-quantised learning.
.
And if I'm not mistaken in catbusta you can switch off quantisation. I could be wrong, haven't used it for 2 years. If not, then by the trick above you can get the result as without quantisation (but 65000 quanta take a long time to create).
can be any number of any pictures and on random.
is the most nonsensical phrase ever.
how do you tell the difference between random and regularity?
the most meaningless phrase ever.
how do you tell the difference between random and regularity?
Welcome to reality.
I have quantisation - I copied a couple of algorithms from catbust and improved them a bit, as it seemed more logical to me.
But I don't use it. The result does not improve.
By the way, if you set 65000 quanta, as a rule, there are not so many different values of a chip and the results absolutely coincide with non-quantised learning.
.
And if I'm not mistaken in catbusta you can switch off quantisation. If not, then by the trick above you can get the result as without quantisation (but 65000 quanta take a long time to create).
My algorithm selects the optimal table for the predictor by empirical and statistical considerations, CatBoost uses the same default settings for all predictors.
65000 quanta - it is too much perhaps, often it is enough 32õ in my experiments.
If you can, it is better to read the markup from a file, it would be interesting to see the effect of a custom table.
I have heard an opinion that one predictor value requires at least 30 observations to make any statistical conclusions - so combining ranges reduces the need for a huge number of observations.