Prediction Wizard

Some of the questions that comes up related to scaling in / out of positions:

Can we make definite answer for all trading situations  for  this question?

What is needed to  make the decision generally  about  this  very important topic in trading?

Can we have some general guidelines?

Can we have general solution for all trading setups?

Can we estimate the benefits of  a good solution?

We saw discussions about this topic on trader forums and found the  situation, that  traders generally not aware of the most important aspects of this, and some  never does scaling, some trader does it in one way, some in another way, and again others call that counter intuitive…

To try to answer the above  questions  we created some examples.

Those examples will  refer to the  levels, presented on the following  chart:

LRef is the reference level, where we are now, when we are about to make the decision about scaling.

L1, L2, L3   Our Hypothetical  first, second and third profit targets.

To make life and our solution simple we created only three levels to differentiate profit objectives.

Can we say something about  scaling without knowing, what are the probabilities, that we reach a specific profit target?  No, unfortunately math is no wizard without  food, without those numbers.

So assume that we are diligent, and collected statistics about the probabilities, that the asset in question will reach the L1, L2, L3 levels.

These probabilities are P1  to reach L1 level, P2 to reach L2 level and P3 to reach L3 level.

Having our predefined profit targets and the probability of reaching those  profit targets we  created  a very simple  mathematical  model to give us the  solutions.

We  just need to enter these numbers into our model and it gives us the solution.

Obviously we have less and less probability to reach higher and higher levels.

The difference between the L1, L2, L3 levels and our reference level LRef can be anything, like

percentage changes or dollar changes, the final answer will be the same  independently from the units we apply.

For demonstration and for further examinations we entered  two sets of numbers, and  got the output  for those inputs from the model:

With these numbers  we got that on a long term basis,  optimally we need to play this setting  in a way such that:

Having a 100% position opened at LRef level, we need to sell our first position, which is about   85 – 100% of the total position at  L1 levels, and need to sell the remaining  0 – 15% position at L2  level.

The worst play would be to sell all positions at L2 level.

The difference between the worst  scaling out and the best scaling out for the long term results is about 100%

In this case we used two positions to scale out, in a way such that Pos1 > Pos2

The second sets of input data for our model application is in the following table:

With these numbers  we got that on a long-term basis,  optimally we need to play this setting  in a way such that:

Having a 100% position opened at LRef level, we need to sell our first position, which is about   0 – 25% of the total position at  L1 levels, and need to sell the remaining  75 – 100% position at L2  level.

The worst play would be to sell all positions at L3 level.

The difference between the worst  scaling out and the best scaling out for the long term results is about 30%

In this case we used two positions to scale out, in a way such that Pos1 < Pos2

Based on this and other testing we can  make some conclusions:

Scaling in  and scaling out of positions  is a complex  methodology, and the optimal  play of it depends on the situation, which could be described by the levels, relative to our reference levels, and the probabilities, reaching those predefined levels.

As those probabilities shifting between our predefined levels, the scaling will also need to be shifted.

When we scale out of existing  positions using two, three…positions, sometimes the  Pos1Size > Pos2Size > Pos3Size  will give optimal solution (Where Pos2Size, Pos3Size might be zero)  other  times the Pos1Size < Pos2Size < Pos3Size.

So sometimes we need to close the smallest position first, other times we might need to sell the biggest position first to reach optimal  profit levels.

It is because of these complexities,  that one individual, intuitive trader  can  hardly adapt to apply the optimal solution in different kind of environments, different kind of setups, where probabilities and ranges are different.

So there is no general solution,  the optimal solutions will be dependent on the predefined levels and the probabilities, that we can reach those levels. Each and every setup is different  even the same setup in different time-frame might require different scaling strategy  if we really want to optimize it.

We have to get those probabilities from  research, depending on the time-frame we are about to play.

Those L1, L2, L3 levels  could be used generally on any time frame, but for example if applied on daily historical data, those can represent the R1, R2, R3 resistance levels.

What can we say without knowing those predefined levels and the corresponding probabilities?

We always need to think forward, and  need to think not only the potential risk levels, that we are taking but equally importantly  the possible / potential profit levels that we are about to shoot for and the possible probabilities that we reach those levels and select our scaling strategy accordingly.

As the above example also demonstrate it,  the maximum total profit not necessarily comes with the maximum Win / Loss ratio. So  once we have  an optimal solution for a setup,  we need to stick with it.

When we consider scaling into positions:

Assuming that the high probability profit target  are really far from our position opening level, we can  scale into  positions,  but it is also very important  where are  those levels, where we open the second, third.. position, as the probability that even the lastly opened position will be a winner  is decreasing.  So if we scale in too late, we might not play optimally, and if we scale in too early, that might not  be optimal play  either.

To translate this into some practical situations, and simplify it for those who like it that way:

During  range markets, when we have smaller profit potentials  than scaling into positions might not be  ideal, but when we have trending days, when relatively high probability targets  might be at high-enough levels (At a good distance from our entry levels.) for nice profits it is  better strategy to increase our total position as the prices reach higher levels (in case of an uptrend) and enter multiple positions.

As we could see the difference between the total results of  the best position scaling, and the worst position scaling (or no scaling)  could be 100% and in some cases even much bigger than that. It is dependent on the situation.  But the first step to get close to optimal  solution is to collect data about the setups we are about to play and  have statistics  about probabilities reaching  our  predefined profit target levels.

In our next entries we try  to highight these ideas with some more practical examples.