The problem of trend algorithms – the large movements

The purpose of trend algorithms is obvious-to follow the trend, fix the profit. It seems that everything is easy and simple, but in fact we have a number of unsolvable problems. Let’s talk about trading trends by levels, that is, when the robot forms a level and opens a deal at a breakdown

The picture shows an example – at the beginning of the chart sideways, a series of losses, then a strong movement, fixed — did not turn over, and only after 3000 points shorted. Case just for an example — it doesn't matter that the expiration was in this place-abstract

We don't know:

1 when the trend started and whether it actually started

2 trend direction

3 the amount of pullback before the trend continues and whether it will continue after the pullback

4 size of the initial stop.

Therefore, we will always / almost always be in the market, and if the movement happens, move along with the market, or we will skip some of the trends and jump into an already moving train, losing some or all of the possible profit, depending on which station we entered and left.

In global trends, there may be 1-2 transactions not every year, while trading, you can incur large losses or re-buy 1-2 times or improve your position, then trading local trends, everything is a little more complicated.

And at the moment it's not about the choice of the model how to trade or how to catch trends and scenarios — if you still managed to catch the large move (keep in mind the context of the large local motion rather than global)

Why is this a problem? First of all, after a large movement, volatility increases, and the price flies up and down during consolidation, and can form a bunch of false signals, and in principle can easily lose everything earned on a large movement. There is another problem, you need to determine whether there will be a pullback, whether there will be a continuation of the movement, in which direction to stand or be out of the market?

In my experience, I tried a lot of options: to be out of the market, and the calculation of entry points after a large movement changed, and canceled signals to continue the movement and canceled reverse stops, and so on.

In particular situations, of course, problems are solved, but on a long history, you can see that there is no clear answer according to statistics. Sometimes after a large move down the market can immediately play all the fall, and we will be out of the market, or sometimes there is an even greater movement down, sometimes sideways. In general, somewhere we lose, somewhere we don't

As a result, I work according to this scheme:

  • Оne algorithm after a large movement — stops trading for 3 days and then forms a new level, and does not use the calculated levels before. That is, a strictly new level different from the historical one.
  • The same algorithm erases past levels after movement, and opens trades only after forming a new level, which also takes into account the current situation of volatility.
  • The robot does not look at the size of the movement and always trades
  • The robot always trades, but reduces the volume after a large movement
  • The robot always trades, but reduces the volume for 3 days or until the signals that cancel this restriction are received
  • The robot has reached a sufficient profit and stops trading in this futures (there were also cases when after a good movement there was nothing left because the market was sideways for a long time and the robot along with it)
  • Ignores counter input for 3 days
  • Ignores entry in the same direction for 3 days
  • Algorithms with shorter signals, waiting for the go-ahead from the algorithm with a long signal ( that is, some of the algorithms have worked out their movements that consider it large from 3-5%, and then look at the algorithm with slow parameters, and if it is still sitting in the transaction, then small algorithms will not open counter transactions) naturally, only part of the algorithms.

I'm not sure if I described all the situations… I don't keep records, I change algorithms in the process, then sometimes I remember for a long time what I changed where and how.

If you meet people on the Internet who talk about running 100-200 robots (I've personally seen this), it's not always about the fact that it's a lot of different algorithms, sometimes it's just varieties of 1-2 robots, on different models, or on different tickers. Add to my situations that these algorithms also have different scenarios with different parameters.

I don't want to exaggerate. Share your experience — how do you work with the consequence of a strong movement?

Submitted November 05, 2020 at 05:00AM by Saro_M_V

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s