Algorithmic ''centrifuge'' - page 19

New comment
 
Andrey Khatimlianskii:

The ideal entry point is an extremum of the ZZ with a spread+point knee size.

Not only entry, but controlling the movement in general

 
Andrey Khatimlianskii:

The ideal entry point is the ZZ extremum with a knee size of "spread + point".

That's right. Obviously this is a bad option.
Then we will use the optimal points (let's call it that), where the knee size will be 10 or more times the spread.
 
Nikolai Semko:

Peter, a riddle for you:
You have three indicators.

The first two indicators generate signals for opening and closing trades.
Both indicators are profitable if tested over the last 10 years.
Both indicators have no parameters as they are self-adjusting, i.e. they do not need to be "optimised" and they only use the minute TF or ticks.
The first indicator showed 55% of profitable trades (if we assume that each trade ends in profit or loss of the same size taking into account the spread) and the second one showed 58% of profitable trades.
Analyzing the behaviour of these first two indicators on the history, we can conclude that the first indicator shows its effectiveness at flat segments of the history, while it has a negative performance on the trend, and the second indicator, on the contrary.

The third indicator determines the current market condition as flat or trending with a 66% probability of correct determination.

Question: how to create the most effective trading system, based on these three indicators:

  1. use only the second indicator, because it has greater effectiveness
  2. use the first two indicators, independently of each other, as they are both profitable and have a more even balance curve, compensating the losses of each other, depending on the flat or trend
  3. using the third indicator, determine what is a trend or flat at the moment, and depending on this, use only one of these indicators at the moment
  4. something else
When you get the right answer, you will realize the absurdity of this thread.

To remind you, the topic of this thread is automation of strategy building. It is desirable that the assembled strategy is profitable, but this is a secondary objective.
 
Andrey Khatimlianskii:

The ideal entry point is the ZZ extremum with a spread+point knee size.

Eh, pips :)

 

Perfect entry points do not exist in isolation from strategy.

Depending on how the strategy is formulated - i.e. what is the best outcome of the game in the model that this strategy uses (in game theory terms)

there will be different ideal inputs. The above pips model is an attempt to maximise profit over all time. in pips. in isolation from risk, stop-loss etc.

Therefore one can only look for ideal points for an already prepared strategy, i.e. formulated one. For this one it makes sense to look for the ideal.

Everything else is empty.

The Zig-Zag is given as the points closest to the ideal entries, if the strategy - to take the maximum trend with min risks on this timeframe with this dimension of the zigzag (the larger the dimension the more small fluctuations are eliminated).

If we bother, we can also enter those breaks of the zigzag, which are the second and larger in the current trend.

But if we add profit/time here, then it is a wider family of strategies, and there is no sense to search for the ideal points for them in general.

 
Nikolai Semko:

Peter, a mystery to you:

Nikolai, what is the right answer? I wonder.

 
Aliaksandr Hryshyn:

Aliaksandr, I think you also created a thread on a similar topic - strategy search automation, can you tell us about the results of your research?

 
Реter Konow:
To remind you, the topic of this thread is automation of strategy building. It is desirable that the assembled strategy is profitable, but, this is a secondary task.
The truth is that the secondary task is not solved by this primary task. And automation for the sake of automation, if it does not ultimately lead to a profitable strategy, is absurd.
 
Aleksei Stepanenko:

Nikolai, what is the right answer? I wonder.

The use of only the 3rd indicator and the first two to the rubbish bin.
 
Aleksey Mavrin:

Perfect entry points do not exist in isolation from strategy.

Depending on how the strategy is formulated - i.e. what is the best outcome of the game in the model that this strategy uses (in game theory terms)

there will be different ideal inputs. The above pips model is an attempt to maximise profit over all time. in pips. in isolation from risk, stop-loss etc.

Therefore one can only look for ideal points for an already prepared strategy, i.e. formulated one. For this one it makes sense to look for the ideal.

Everything else is empty.

The Zig-Zag is given as the points closest to the ideal entries, if the strategy - to take the maximum trend with min risks on this timeframe with this dimension of the zigzag (the larger the dimension the more small fluctuations are eliminated).

If we bother, we can also enter those breaks of the zigzag, which are the second and larger in the current trend.

But if profit/time is also added here, then we are talking about a wider family of strategies and there is no point in searching for ideal points for them in general.

You have to give up looking for perfect points.
There are things on which we can earn - the trend and flat dynamics, as well as peculiarities and patterns of the session. There are countless entry and exit points on history within different sections, and the maximum profit on them is a result of artificial fitting to the history. Adjustment for maximal profit in specific segments should not be a deal benchmark.
1...121314151617181920212223
New comment