Optimization of 3 or more parameters

[ehbeehefak]

lso, if I understand well, for each fast -slow sma value, you are running a strategy
and extract the sharp ratio, that you plot ?
Now you want to also include another variable the stop loss?
Am I understanding correctly what you are trying to do?

Yes, I want to run a strategy for every fast-sma,slow-sma,stoploss-distance and takeprofit-distance value. Plotting the strategy results of fast-sma and slow-sma is achievable, but as you said, what if I add the stoploss-distance to the mix?

Sharpe ratio was just an example, it can just as be Net Profit, Net Profit/Max Drawdown and the similar statistics.

 

[smallStops]

You need first to formalise your problem in a "math" way

fast-sma ( valuei) = fast-sma + increment_fastsma
Fsma(i) = Fsma0 + i*coeffF
Ssma(j)= Ssma0 +j*coeffS
stops(k)= stops0 + k*coeffK

For each you get the profit = Function ( i,j,k).
Now, you want to maximise the profit, while minimizing k, this on the "field" (i,,j).

Is this what you want to do?

 

[ehbeehefak]

This is one suggestion, as you are using matlab.
few minutes that I copy this.

Actually, I am using R, I found that image on google just to clarify what I meant.

 

[ehbeehefak]

You need first to formalise your problem in a "math" way

fast-sma ( valuei) = fast-sma + increment_fastsma
Fsma(i) = Fsma0 + i*coeffF
Ssma(j)= Ssma0 +j*coeffS
stops(k)= stops0 + k*coeffK

For each you get the profit = Function ( i,j,k).
Now, you want to maximise the profit, while minimizing k, this on the "field" (i,,j).

Is this what you want to do?

Yes, kind of. Minimizing k is not needed though. Stops can be far away if that is the most optimal. Also, maximizing function(i,j,k) is not adequate, as it might return me parameters that might represent a "peak" value (which means the parameter combination is very sensitive to market changes). I want to choose "plateau" values (parameter combination is more robust compared to "peak" combination). Here is a picture to explain further:
http://www.multicharts.com/img/8323_3d_graf.png

 

[smallStops]

In that case, your could plot a set of 3D "wires"
i,j, stop values curve ( the chart named profitvaluek)

 

[ehbeehefak]

In that case, your could plot a set of 3D "wires"
i,j, stop values curve ( the chart named profitvaluek)

That can results in hundreds of plots, but seems it is the only way to do this.

 

[smallStops]

No : you plot the different "wires" on the same 3D plot.
You can start with 5 profits values , all on the same chart.

 

[ehbeehefak]

No : you plot the different "wires" on the same 3D plot.
You can start with 5 profits values , all on the same chart.

Oh, ok I see. So I will plot 5 of these (https://www.wavemetrics.com/products/igorpro/creatinggraphs/3dandvolume/surfacepix/wireframe1.jpg). Seems it is going to be too cluttered, but giving them different colors might improve it a little. Thanks.

Also, although this solves the problem for 3-parameter strategies, it won't work with 4-parameter strategies.

 

[smallStops]

Try with one profit target. And plot its wire of stops on (i,j).
Different wire colors is a very good way to highlight it.

Actually, do you mind putting such a 3D plot for a 50 and 100sma,
20 and 50 sma, as well as 100sma and 200 sma.
I am curious to check how different these are.

 

[ehbeehefak]

Try with one profit target. And plot its wire of stops on (i,j).
Different wire colors is a very good way to highlight it.

Actually, do you mind putting such a 3D plot for a 50 and 100sma,
20 and 50 sma, as well as 100sma and 200 sma.
I am curious to check how different these are.

I don't have any experience with wireframe plots in R but will try to do that for you. I have only done heatmaps in R, but look into it when I get home from university.