Cone8
 ( 5.88% )
- ago
Monte Carlo-Lab extension - my favorite!
13
1,607
13 Replies

Reply

Bookmark

Sort
Glitch8
 ( 9.28% )
- ago
#1
Looks like this item is in queue for development! Time to dust off the old WL6 and see what Monte Carlo Lab had going on. It's been a long time since I wrote that extension!
2
- ago
#2
I suggest that you add one additional screen where all Equity curves are displayed on a single graph.

Vince
0
Glitch8
 ( 9.28% )
- ago
#3
It'll look like one of those hurricane spaghetti graphs! :D

Actually that's what I did in the Quantacula Studio "Best/Worst Case" Viewer.
0
- ago
#4
Well, it is hurricane season! ;)

The visual display of the variability can help identify those strategies that are very sensitive to small changes in parameters leading to a large number of undesirable potential trades.

Vince
0
Glitch8
 ( 9.28% )
- ago
#5
You’ll be able to multi select the items to plot, in addition to selecting the baseline and best and worst runs. I plotted 1000 runs but the chart gets quite sluggish when plotting that many. Probably stick to a few dozen.
0
- ago
#6
Glitch,

Perhaps you can provide a User-Selected percentage of the total runs (randomly) to be displayed. That way if the User is patient he can have all 1000 runs, or just 10. :)

Vince
0
Glitch8
 ( 9.28% )
- ago
#7
Yes I was thinking along those lines too.
0
- ago
#8
What about this one: If there are more than N curves (say 20) then plot the 5-Percentile, the Median and 95-percentile of all curves in each X-Point.

Will look like a center line with a band around it.
0
- ago
#9
Indeed this type of chart looks like a hurricane spaghetti. In WL6 for the Trade Life visualizer I added an Opacity control to try and improve the looks:

http://www2.wealth-lab.com/WL5WIKI/PVTradeLife.ashx
http://www2.wealth-lab.com/WL5WIKI/GetFile.aspx?File=Community.Visualizers/PVTradeLife.png
0
Glitch8
 ( 9.28% )
- ago
#10
The MC Lab already uses a translucent color.
0
- ago
#11
QUOTE:
What about this one: If there are more than N curves (say 20) then plot the 5-Percentile, the Median and 95-percentile of all curves in each X-Point.


Unfortunately, the distribution is not Gaussian, and it is those negative "fat tails" beyond the 5%/95% lines that are of greatest interest.

Vince
0
- ago
#12
QUOTE:
Yes I was thinking along those lines too.


This is the general approach that is followed in the Data Science community, where there can be millions of data points in a dataset.

Vince
0
- ago
#13
Btw, thanks for the last monte Carlo update, it was previously running slow and would freeze up on me after around 150 runs, the update fixed that! Thanks!
1

Reply

Bookmark

Sort