Support Board
Date/Time: Sat, 23 Nov 2024 20:45:16 +0000
Post From: Offering To The Community: Forex: Currency Decorrelation (and Currency Strength)
| [2018-11-15 03:25:18] |
| bjohnson777 (Brett Johnson) - Posts: 284 |
|
I understand. I like trend and SAR lines, too. The problem boils down to multiple floating numbers (currencies, already mentioned) and the math used for decorrelation. SC has decorrelated currencies that have been locked down to a common point some years ago, but they've diverged significantly since then. You can have a look at EUR-X and JPY-X as examples. Those might be a workaround for drawing chart lines. The other lockdown problem is with scaling to each other. It's really hard to tell which one is moving more relative to the other. It may look like one currency is worth more than another when it really isn't. To be fair to the SC devs, this is a very common problem in other decorrelation programs, too. There really isn't a good solution. I made the choice to follow the relative path to get around these problems at the cost of the lock down. I default to 300 bars back as that will show overall movement (or lack of it) fairly well for a reasonably recent time period keeping all the currencies on screen (as in, not having 1000 point vertical distances between each of them). If you scroll the chart back a little, you can see where all the currencies cross at that point. I also made the choice to try and keep all the decorrelated currencies properly scaled to each other in the chosen main currency (or a different chosen home currency). 2 years ago when I wrote this, I didn't see another like it. That may have changed by now. When I was studying FOREX back then, the lack of this information was driving me batty. Your mentioned experiments won't work because anytime the main chart redraws/reloads, it will start from the 300 bars back and continue from there. That's why you're seeing vertical shifts on the chart. The data on the disk or spreadsheet would then show a large bar gap where none actually exists. That could make for some very confusing and possibly smashed indicators. With my health problems, I don't have the math skills to try and figure out something around this problem. I haven't had the time to look recently, but I'm doubting anyone else has a solution, either. Even if there is one, it may be too much for a chart study running in a limited execution environment to handle. My decorrelation study running with my currency strength indicators are already a heavy load. These are my immediate thoughts on the issue. |