Machine learning in trading: theory, models, practice and algo-trading - page 3261

New comment
 
In many cases you can shrink the array by removing neighbouring rows, which are almost always highly correlated. By a factor of 5 at least.
 
Maxim Dmitrievsky #:
In many cases you can shrink the array by removing neighbouring rows, which are almost always highly correlated. By a factor of 5 at least.

It strongly depends on the attributes. For example, the attribute (MaxLastN - MinLastN) changes abruptly.

 

My hypothesis of the PD/SP levels

There are many different participants in the market, but they can be divided into professional and non-professional participants.

Professional participants (stock exchange, market makers) are those who have information about the market.

Non-professional - all others who can only build probabilistic models.


So what is a level in my subjective understanding? It's a certain correct sequence of events.


In highly liquid markets, liquidity is always plentiful, there's always a robot in the stack waiting for your action to become your counter agent in a reverse trade.

As Lekha Maitrade used to say - "just get into the market any way you can, they've been waiting for you there for a long time".


So let's assume that a trader or a group of traders in the market made a purchase at a specific price, respectively a counter agent made a sale at the same price.

And then a series of events should happen:

If the price went down and the trader didn't bury the position.

What does that mean? It means that the trader has switched on the "over-sitting" mode. Waiting mode. like I' m going to close in the BU when the price comes back up.

Isn't that right? Remember yourself, we've all done it, it's hard for us humans to accept our mistakes.


What happens next? What happens next?

Price comes to a level:

1) The trader sells moving the price down.

2) The market maker protects the price by not letting it rise, thus protecting this level.

Thus all participants of trade (professionals and non-professionals) at this particular price act in one direction, as if together....


So levels exist, they cannot exist a priori, because there is a price....


Here is an illustration of my theory of levels on the example of today's trading of my robots, my robots here act as non-professional traders.

Notice how the price bounces off the robots' trades (selling bounces up, buying bounces down).


I'm interested in constructive criticism of the hypothesis ...


What does it have to do with MO?... Actually, it is direct, what understanding of the market is such and such signs... nobody has ever built a successful algorithm on stochastics and other curveballs.

 
mytarmailS #:

notice how the price bounces off the robots' trades (selling bounces up, buying bounces down).

Interested in constructive criticism of the hypothesis....

Trade on SB, there the price will also bounce in the opposite direction from your robot trades.

Reverse the trades and you will see the same situation again. The essence of the hypothesis is in the "own eye".

 
fxsaber #:

Trade on the SB, there the price will also bounce in the opposite direction of your robot trades.

Reverse the trades and you will see the same situation again. The essence of the hypothesis is in "my eye".

I will have to gather my strength and write a test comparing the statistics of level development on the SB and on prices.
 
Maxim Dmitrievsky #:

What estimate can be made that curveballs, on average, cross the zero line back and forth as often as possible?


Maybe the Sum of Crossings!?!? )))) unexpectedly)

Or look at stationarity tests
 
mytarmailS #:
Maybe the Sum of Intersections!?!? )))) unexpectedly)

Or look at stationarity tests

expectedly

 

ridiculous MO strategy from this post

we continue testing in real time, 4th day of trading...

15 trading systems traded simultaneously, exclusively on euras so far.

 
mytarmailS #:

My hypothesis of PD/SP levels

There are many different players in the market, but they can all be divided into professional and non-professional.

Professional participants (stock exchange, market makers) are those who have information about the market.

Non-professional - all others who can only build probabilistic models.


So what is a level in my subjective understanding? It's some kind of correct sequence of events.


In highly liquid markets, liquidity is always plentiful, there is always a robot in the stack waiting for your action to become your counter agent in a reverse trade.

As Lekha Maitrade used to say - "just get into the market any way you can, they've been waiting for you there for a long time".


So let's assume that a trader or a group of traders in the market made a purchase at a specific price, respectively a counter agent made a sale at the same price.

And then there should be a series of events:

If the price went down and the trader did not bury the position.

What does that mean? It means that the trader has switched on the "over-sitting" mode. or "wait and see" mode. like I' m going to close in the BU when the price comes back up.

Isn't that right? Remember yourself, we've all done it, it's hard for us humans to accept our mistakes.


Next... What happens next?

Price comes to a level:

1)The trader sells moving the price downwards

2) Market Maker protects the price by not letting it rise thus protecting this level

Thus all participants of trade (professionals and non-professionals) at this particular price act in one direction, as if together....


So the levels exist, they cannot exist a priori, because there is a price....


Here is an illustration of my theory of levels on the example of today's trading of my robots, my robots here act as non-professional traders.

Notice how the price bounces off the robots' trades (selling bounces up, buying bounces down).


I would be interested in constructive criticism of the hypothesis....


What does it have to do with the MO?, yes, in fact, directly, what understanding of the market such and such signs... no one has ever built a successful algorithm on stochastics and other curves.

A level becomes a level when it is over, and in the future, in which trading decisions are made, there will be another level, i.e. the level has no predictive ability.

But in reality it is even worse.

Building TS on levels at the stage of training, testing can give excellent results because of "looking ahead", as we teach on ready-made levels, and when we start counting the predictor, there will be no level - see above. Therefore, the prediction error on different there ALE OOS can give excellent results, but if we start chasing step by step outside the training files, doing predictor calculations, the classification error will be arbitrary.

 
mytarmailS #:

Ridiculous MO strategy from this post

Strategies can be the most unsightly, the only thing that matters is that the average balance increment is greater than zero.


generate 100 conditional strategies with average above zero.

f <- function() cumsum(rnorm(500,mean = 0.01))
s <- sapply(1:100,\(x) f())


the profit on individual strategies is below zero and the strategies themselves look terrible.

par(mfrow=c(4,4),mar=c(2,2,2,2))
for(i in 1:16){
          plot(s[,i], main=paste("стратегия номер",i),t="l") 
          abline(h=0,col=4,lty=2)}


but together these strategies generate stable profits.

sc <- rowMeans(s)
layout(1)
plot(sc,t="l", main="все стратегии")

========================================

That's how it is... by rethinking everything a little bit and giving up quality (one good TS) you can move to quantity and as a result come to quality.


=========================================

The only question is how to select strategies with non-zero expectations.

Everything else is solvable

1...325432553256325732583259326032613262326332643265326632673268...3526
New comment
Reason: