- This topic has 270 replies, 23 voices, and was last updated 9 years ago by
.
- Posts
- Data needs to be “normalized” before such comparison. G.By “normalized” do you mean “weighted”?As in each currency needs to be weighted by its market share (i.e. significance).One way to do this is via Money Supply.Or here is a post from @gg53 years ago.Which the the preferred method?
Something I found deep in the net posted from GG:
Individual currency weight – How to: (info gathered from the internet- percent market share)
EURUSD = 28% USDJPY = 14% GBPUSD = 9% AUDUSD = 6% USDCAD = 5% USDCHF = 4% EURJPY = 3% EURGBP = 3% OTHER = 28%
Those numbers doesn’t change much – but you can update them monthly (or whenever…)
For current example – Let’s ignore the “Other” and deal with the 72% (above sum). Now we take all the pairs that include the USD = 66%. Divide in 2 (to extract the USD only) = 33%. Since the 33% is only 72% of the market, USD = 33/72*100 = ~ 45%.
Meaning – roughly 45% of the market volume consists of USD. (Sorry if I caused some headaches…)
We repeat the above calculation for each of the other 7 currencies.
Now go to the indicator (FXCorrelator in NVP case) and multiply each currency with its “factor”, i.e.: USD = USD*0.45
Just make shure that the total “factors” are equal to 1 or 100%….
Weight or Factors have nothing to do with proper DELTA calculations.
All those are “nice to have” or increase accuracy and meaning, but the BASIC is each and every DELTA proper calculation.
G.
Here is another analysis I did myself at the beginning of the year.
Maybe it’s better if you link the source of your analysis so everyone can learn something…
Nothing special. Just took the results from the Bank for International Settlements Triennial Bank Survey. Refer to the link.
http://www.bis.org/publ/rpfx13fx.pdf
What's it all about? It's all about money.
Data needs to be “normalized” before such comparison. G.By “normalized” do you mean “weighted”?As in each currency needs to be weighted by its market share (i.e. significance).One way to do this is via Money Supply.Or here is a post from @gg53 years ago.Which the the preferred method?Something I found deep in the net posted from GG:
Individual currency weight – How to: (info gathered from the internet- percent market share)
EURUSD = 28% USDJPY = 14% GBPUSD = 9% AUDUSD = 6% USDCAD = 5% USDCHF = 4% EURJPY = 3% EURGBP = 3% OTHER = 28%
Those numbers doesn’t change much – but you can update them monthly (or whenever…)
For current example – Let’s ignore the “Other” and deal with the 72% (above sum). Now we take all the pairs that include the USD = 66%. Divide in 2 (to extract the USD only) = 33%. Since the 33% is only 72% of the market, USD = 33/72*100 = ~ 45%.
Meaning – roughly 45% of the market volume consists of USD. (Sorry if I caused some headaches…)
We repeat the above calculation for each of the other 7 currencies.
Now go to the indicator (FXCorrelator in NVP case) and multiply each currency with its “factor”, i.e.: USD = USD*0.45
Just make shure that the total “factors” are equal to 1 or 100%….
Weight or Factors have nothing to do with proper DELTA calculations. All those are “nice to have” or increase accuracy and meaning, but the BASIC is each and every DELTA proper calculation. G.
We seem to be jumping back and forth.
I would like to try and remove any uncertainty.
Can you please confirm I have the signal processing technique concept correct:
1) First derivative
Create a “strength” indicator (built using “traditional” techniques )
2) Second derivative
From this “strength” indicator value we apply an ROC calculation. The ROC value we are interested in is the Delta from the previous bar. We are interested in the RELATIVE change, therefore resetting the ROC isnt necessary.
3) We then weight the ROC delta signal according to the market share e.g. USD = 50% or 0.5 weighting.
Is that the basic idea?
What's it all about? It's all about money.
USD is THE common denominator. It exist in each and every major currency-pair. If you want a TRUE, effective, and most importantly LOGICAL “reset” or “anchor” – that’s the one. G.
After items 1 – 3 in my post above have been mastered….. should the next step be (?) :
4) Create synthetic currency.
Since we dont all have access to the USDX we should use the Market Share calculated in item 3 above to create a synthetic currency of the USD vs Other Major’s.
When viewing just the top 8 currencies, USD is 50% of the market share and The Rest (other Majors) make up the other 50%. The same way the USDX is calculated we create a synthetic currency.
We then use this synthetic currency and apply the Zig Zag indicator.
We reset the ROC at the swing points on the Synthetic Currency.
The Synthetic will also help your trading greatly in general trend direction.
What's it all about? It's all about money.
Data needs to be “normalized” before such comparison. G.By “normalized” do you mean “weighted”?As in each currency needs to be weighted by its market share (i.e. significance).One way to do this is via Money Supply.Or here is a post from @gg53 years ago.Which the the preferred method?Something I found deep in the net posted from GG:
Individual currency weight – How to: (info gathered from the internet- percent market share)
EURUSD = 28% USDJPY = 14% GBPUSD = 9% AUDUSD = 6% USDCAD = 5% USDCHF = 4% EURJPY = 3% EURGBP = 3% OTHER = 28%
Those numbers doesn’t change much – but you can update them monthly (or whenever…)
For current example – Let’s ignore the “Other” and deal with the 72% (above sum). Now we take all the pairs that include the USD = 66%. Divide in 2 (to extract the USD only) = 33%. Since the 33% is only 72% of the market, USD = 33/72*100 = ~ 45%.
Meaning – roughly 45% of the market volume consists of USD. (Sorry if I caused some headaches…)
We repeat the above calculation for each of the other 7 currencies.
Now go to the indicator (FXCorrelator in NVP case) and multiply each currency with its “factor”, i.e.: USD = USD*0.45
Just make shure that the total “factors” are equal to 1 or 100%….
Weight or Factors have nothing to do with proper DELTA calculations. All those are “nice to have” or increase accuracy and meaning, but the BASIC is each and every DELTA proper calculation. G.
We seem to be jumping back and forth. I would like to try and remove any uncertainty. Can you please confirm I have the signal processing technique concept correct: 1) First derivative Create a “strength” indicator (built using “traditional” techniques ) 2) Second derivative From this “strength” indicator value we apply an ROC calculation. The ROC value we are interested in is the Delta from the previous bar. We are interested in the RELATIVE change, therefore resetting the ROC isnt necessary. 3) We then weight the ROC delta signal according to the market share e.g. USD = 50% or 0.5 weighting. Is that the basic idea?
In “my” words:
1. Currency Strength indicator is a RANKING indicator, showing the currently strongest from top to bottom.
2. The Delta between current and previous bar, as described by you, is only half “Ok”… You can’t simply compare DELTA of one pair to DELTA on another pair – without some kind of “normalization”. I already given the “interest rates” example and several others.
3. ROC or “speed of movement” should be displayed as text along side the indicator ranking and direction lines (optional).
4. “Reset” or “anchoring” – optional.
The main task, IMHO, is to solve the #2 issue first.
G.
USD is THE common denominator. It exist in each and every major currency-pair. If you want a TRUE, effective, and most importantly LOGICAL “reset” or “anchor” – that’s the one. G.
After items 1 – 3 in my post above have been mastered….. should the next step be (?) : 4) Create synthetic currency. Since we dont all have access to the USDX we should use the Market Share calculated in item 3 above to create a synthetic currency of the USD vs Other Major’s. When viewing just the top 8 currencies, USD is 50% of the market share and The Rest (other Majors) make up the other 50%. The same way the USDX is calculated we create a synthetic currency. We then use this synthetic currency and apply the Zig Zag indicator. We reset the ROC at the swing points on the Synthetic Currency. The Synthetic will also help your trading greatly in general trend direction.
You DO have access to the USDx. there are plenty free ones on the NET, and it’s only one line of code.
G.
USDx (or USD index, based on USD trade ballance with major countries):
Usdx=50.14348112 * MathPow(iClose(“EURUSD”,0,i),-0.576) * MathPow(iClose(“USDJPY”,0,i),0.136) * MathPow(iClose(“GBPUSD”,0,i),-0.119) * MathPow(iClose(“USDCAD”,0,i),0.091) * MathPow(iClose(“USDSEK”,0,i),0.042) * MathPow(iClose(“USDCHF”,0,i),0.036);
G.
USDx (or USD index, based on USD trade ballance with major countries): Usdx=50.14348112 * MathPow(iClose(“EURUSD”,0,i),-0.576) * MathPow(iClose(“USDJPY”,0,i),0.136) * MathPow(iClose(“GBPUSD”,0,i),-0.119) * MathPow(iClose(“USDCAD”,0,i),0.091) * MathPow(iClose(“USDSEK”,0,i),0.042) * MathPow(iClose(“USDCHF”,0,i),0.036);
Hi G.
Thanks a lot for providing this detail!
At least I’m getting an idea now what caused the confusion about 1st vs. 2nd derivative of prices in previous posts. We were talking about completely different base arithmetics.
Got to check out now the implications for my indicator.
Thanks again,
J.
A good trader is a realist who wants to grab a chunk from the body of a trend, leaving top- and bottom-fishing to people on an ego trip. (Dr. Alexander Elder)
USDx (or USD index, based on USD trade ballance with major countries): Usdx=50.14348112 * MathPow(iClose(“EURUSD”,0,i),-0.576) * MathPow(iClose(“USDJPY”,0,i),0.136) * MathPow(iClose(“GBPUSD”,0,i),-0.119) * MathPow(iClose(“USDCAD”,0,i),0.091) * MathPow(iClose(“USDSEK”,0,i),0.042) * MathPow(iClose(“USDCHF”,0,i),0.036);
Hi G. Thanks a lot for providing this detail! At least I’m getting an idea now what caused the confusion about 1st vs. 2nd derivative of prices in previous posts. We were talking about completely different base arithmetics. Got to check out now the implications for my indicator. Thanks again, J.
I think that you are going to try or test it on Weight/Factor issue (based on the USDx code logic, MathPow()…).
I believe that the main issue in one step before that: on how to properly calculate and “normalize” the DELTA.
G.
A quick note:
Some background info can be found over there: https://www.theice.com/publicdocs/futures_us/ICE_Dollar_Index_FAQ.pdf
A description of the corresponding EUR index: https://www.theice.com/publicdocs/rulebooks/futures_us/24_ICE_Futures_EURO_Index.pdf
A good trader is a realist who wants to grab a chunk from the body of a trend, leaving top- and bottom-fishing to people on an ego trip. (Dr. Alexander Elder)
I think that you are going to try or test it on Weight/Factor issue (based on the USDx code logic, MathPow()…). I believe that the main issue in one step before that: on how to properly calculate and “normalize” the DELTA.
Yes, you’re right! Did some quick tests last night, and came to the conclusion that this will require a proper normalization when calculating crosses. Need some more time to work it over!
A good trader is a realist who wants to grab a chunk from the body of a trend, leaving top- and bottom-fishing to people on an ego trip. (Dr. Alexander Elder)
Normalization brainstorm:
IMO, methods like Stoch or Fisher should be eliminated per se. These would distort the index lines.
So an easy and obvious option would be to display the difference in percent of the current index value as compared to some fixed reference point of the same index line. This reference point could be:
(A) a point in the past X bars away (floating) -> with X chosen reasonably high could be robust enough to work with, adding minor distortion
(B) a fixed date & time in the past (non-floating) -> possibly the most stable one of these four solutions. Date & time could be chosen reflecting the trading horizon.
(C) an averaged point in the past X bars away (floating) -> more robust regarding the influence of past fluctuations as compared to (A)
(D) current average of the index line -> resulting in MACD-like oscillations: very dynamic, but prune to short term distortions.
Just a brainstorm for now. Out of these options (better ones may exist) I would prefer (B) for a first try. Regarding that reference point our recent discussion about resetting the indicator comes to my mind.
s.
A good trader is a realist who wants to grab a chunk from the body of a trend, leaving top- and bottom-fishing to people on an ego trip. (Dr. Alexander Elder)
Normalization brainstorm: IMO, methods like Stoch or Fisher should be eliminated per se. These would distort the index lines. So an easy and obvious option would be to display the difference in percent of the current index value as compared to some fixed reference point of the same index line. This reference point could be: (A) a point in the past X bars away (floating) -> with X chosen reasonably high could be robust enough to work with, adding minor distortion (B) a fixed date & time in the past (non-floating) -> possibly the most stable one of these four solutions. Date & time could be chosen reflecting the trading horizon. (C) an averaged point in the past X bars away (floating) -> more robust regarding the influence of past fluctuations as compared to (A) (D) current average of the index line -> resulting in MACD-like oscillations: very dynamic, but prune to short term distortions. Just a brainstorm for now. Out of these options (better ones may exist) I would prefer (B) for a first try. Regarding that reference point our recent discussion about resetting the indicator comes to my mind. s.
Ok, now I think that we are on the same “page”…
Since we are dealing with “STRENGTH”, why not use RSI (one of some other options)?
It’s values are already “normalized”, and the same for all currency-pairs.
This can be done by RSI(t)-RSI(t-1), or even RSI(fast)-RSI(slow) on the same bar. RSI(fast)/RSI(slow) …. I think that you already know the other options along this lines…
BTW: using some kind of RSI, gives me 1-2 bars earlier signal compared to previous versions.
Another option, along different line, is to use EXTERNAL reference.
You can compare each currency-pair movement to that external reference. USDx comes to mind as such reference, that will also “normalize” the DELTA.
USDx is basically a “basket” of currencies, but it’s possible to create another “basket”, although USDx is already internationally recognized.
G.
Ok, now I think that we are on the same “page”…
Great to hear that!
Since we are dealing with “STRENGTH”, why not use RSI (one of some other options)? It’s values are already “normalized”, and the same for all currency-pairs.
Sure, RSI is an option. At first sight, I’m having the same doubts like in case of Stoch: this kind of ‘hard’ normalization with fixed limits tends to distort the original values and the strenghts of movements. For very wide calculation windows, this issue is minimized, though.
I’ll give it a try tomorrow. Above method (B) is already implemented, looks ok. Need to debug & clean the code, possibly can post tomorrow.
A description of the corresponding EUR index: https://www.theice.com/publicdocs/rulebooks/futures_us/24_ICE_Futures_EURO_Index.pdf .
For comparison, that Euro Index is also implemented.
On the pic, as an appetizer for tomorrow:
Daily TF, top window showing USDx, bottom window EURx, both of them calculating crosses of our 8 major currencies. Reference point is set about 1 month after release of the EURCHF peg early this year, see blue dashed line.
simplex
A good trader is a realist who wants to grab a chunk from the body of a trend, leaving top- and bottom-fishing to people on an ego trip. (Dr. Alexander Elder)
Ok, based on @gg53 ‘s ideas posted earlier in this thread, here’s my 1st currency strength indicator based on USDx and EURx index baskets. A sincere thank you to gg53 for posting the idea!
For a description of the indexes, please refer to the links published in my earlier post.
Source code is provided, so anybody feel free to check out the algorithm – any feedback much appreciated!
It is obvious, how much easier it is to visually spot major moves with this indicator as compared to my earlier CS experiments.
In this version, normalization is done by calculating the current displacement of the 8 index values relative to a their respective values given at a constant point in time at midnight, X days ago. I will check out RSI normalization soon.
You may choose between a USDx and a EURx basis: both will provide similar signals, yet with different emphasis for different currencies. From a fundamental point of view it might be obvious to choose heavy weight USD centric view, but for certain instruments or trading styles EUR centric view might be useful. Just check it out!
The rest of the input variables should be rather self explanatory.
Have fun & take care!
simplex
A good trader is a realist who wants to grab a chunk from the body of a trend, leaving top- and bottom-fishing to people on an ego trip. (Dr. Alexander Elder)
Good job, simplex !!
I didn’t test it yet, but it looks good.
It will be beneficial to ADD the USDx (USD index) line (maybe in BOLD, or other line style, and different color) to that indicator.
It is an old habbit of mine to NEVER trade AGAINST it’s general direction…
G.
It looks “strange” on M1 (default params, MAX zoom).
Lines are almost “flat” (compared to previous short history on the same chart).
G.
I think that the problem is historical scaling.
Your indie is drawing history before the “anchor” or start date, and if there are huge moves in that past – the whole scaling of current session is “shrinked” accordingly – up to a point that medium/large moves are hardly noticed.
G.
Proper trade Entry (“By the Book”):
In the attached you can see the NZD (white) is very strong, and the USD, JPY, CHF falling bellow the “0” line – marked by the RED vertical dotted line.
That’s our “be prepared” for trade entry, but not just yet… We wait for proper timing to go LONG on NZDUSD, NZDCHF & NZDJPY.
The GREEN dotted line signals LONG. Why?
Our bias is LONG based on the Currency-Strength. We wait for the price to be at its lowest (“buy cheap”, daddy said…).
To identify “buy cheap” entry, and minimize DD, we use either the Gadi_OBV, or Tick_Volume Trough, or the “Spaghetti” Trough.
When we go LONG – we want to buy Troughs. Vice versa for SHORTS (SELL Peaks).
I showed ALL of them just for illustration. In practice you can use only one or two for Entry timing confirmation.
…And yes, you can trade ALL three A/M currency pairs…
Green pips, and thanks to simplex.
G.
Ok, based on @gg53 ‘s ideas posted earlier in this thread, here’s my 1st currency strength indicator based on USDx and EURx index baskets.
Thanks simplex , testing right now.
Good job, simplex !!
Thank you!
It will be beneficial to ADD the USDx (USD index) line (maybe in BOLD, or other line style, and different color) to that indicator.
It’s already there! When USDx basis is chosen, USD line represents pure USDx, with the same normalization applied as the other lines. The same is true for EURx and EUR line, when EURx basis is chosen.
It looks “strange” on M1 (default params, MAX zoom). Lines are almost “flat” (compared to previous short history on the same chart).
That’s absolutely normal, given the nature of the normalization chosen. For M1, try ‘reference point’ = 1 day. Possibly modifying normalization (maybe RSI ?) is the next step to undertake.
I think that the problem is historical scaling. Your indie is drawing history before the “anchor” or start date, and if there are huge moves in that past – the whole scaling of current session is “shrinked” accordingly – up to a point that medium/large moves are hardly noticed.
This is also natural: depends on screen zoom chosen for your chart window. RSI normalization, like suggested in your earlier posts, would overcome this issue, while wiping out any information about the absolute strength of recent moves. That’s the price to be paid for every ‘hard’ normalization pressing original values in rigid limited scales. You already posted suggestions about properly finding points in time for resetting – I’m also thinking about implementing a modified reset mechanism based on those. Takes a bit of time, though.
I would assume that utilizing two instances of the indicator might be most useful: one showing CS relative to a certain chosen point in time in the past for general direction, and one showing the RSI normalized version to emphasize short term entry opportunities. This could be checked once the RSI version is ready.
By the way: I noticed the time stamps of your recent posts: do you NEVER sleep?
s.
A good trader is a realist who wants to grab a chunk from the body of a trend, leaving top- and bottom-fishing to people on an ego trip. (Dr. Alexander Elder)
Good job, simplex !!
Thank you!
It will be beneficial to ADD the USDx (USD index) line (maybe in BOLD, or other line style, and different color) to that indicator.
It’s already there! When USDx basis is chosen, USD line represents pure USDx, with the same normalization applied as the other lines. The same is true for EURx and EUR line, when EURx basis is chosen.
It looks “strange” on M1 (default params, MAX zoom). Lines are almost “flat” (compared to previous short history on the same chart).
That’s absolutely normal, given the nature of the normalization chosen. For M1, try ‘reference point’ = 1 day. Possibly modifying normalization (maybe RSI ?) is the next step to undertake.
I think that the problem is historical scaling. Your indie is drawing history before the “anchor” or start date, and if there are huge moves in that past – the whole scaling of current session is “shrinked” accordingly – up to a point that medium/large moves are hardly noticed.
This is also natural: depends on screen zoom chosen for your chart window. RSI normalization, like suggested in your earlier posts, would overcome this issue, while wiping out any information about the absolute strength of recent moves. That’s the price to be paid for every ‘hard’ normalization pressing original values in rigid limited scales. You already posted suggestions about properly finding points in time for resetting – I’m also thinking about implementing a modified reset mechanism based on those. Takes a bit of time, though. I would assume that utilizing two instances of the indicator might be most useful: one showing CS relative to a certain chosen point in time in the past for general direction, and one showing the RSI normalized version to emphasize short term entry opportunities. This could be checked once the RSI version is ready. By the way: I noticed the time stamps of your recent posts: do you NEVER sleep? s.
Scaling: In this case the scaling is a problem. All below “0” are squeezed and the “0” is not in the center. Although matematically correct, it’s visually misleading.
USDx: In my indicator the USD is calculated the same as the other currencies, and it’s a bit different from the USDx.
Sleeping: It’s “running from reality, and a waste of time…. ;-}
G.
Proper trade Entry (“By the Book”)
Good summary!In the attached you can see the NZD (white) is very strong, and the USD, JPY, CHF falling bellow the “0” line – marked by the RED vertical dotted line. That’s our “be prepared” for trade entry, but not just yet… We wait for proper timing to go LONG on NZDUSD, NZDCHF & NZDJPY.
Please play with that reference point parameter in CIX. It’s like modifying RSI length and may change the point of view. See reset discussion in my previous post. And you already posted your opinion about the zero line in this kind of CS algorithm: it is meaningless as a rigid borderline, I’m absolutely with you regarding this detail. What counts is the relative position of all lines against each other (mid term bias), and their short term movements.
Back to coder’s work now: reset and normalization are waiting to be enhanced further.
s.
A good trader is a realist who wants to grab a chunk from the body of a trend, leaving top- and bottom-fishing to people on an ego trip. (Dr. Alexander Elder)
All below “0” are squeezed and the “0” is not in the center.
I confirm that. This is index and algorithm specific. Check out EURx, and other currencies will be sqeezed – that’s why I implemented this one. Those currency baskets provide specific views of the FX market: US centered vs. EU centered.
USDx: In my indicator the USD is calculated the same as the other currencies, and it’s a bit different from the USDx.
Ok, I’ll check that.
Sleeping: It’s “running from reality, and a waste of time…. ;-}
Interesting! So you assume that what our brain filters out of the information our senses will provide while we’re awake is really real? And our dreams (if any) are not?
Re ‘waste of time’: every once in a while I wake up in the morning before the alarm clock gets busy, having an idea in my mind that solves one of my current coding problems. I love those mornings! Today’s not one of those, but nice autumn weather evens it out!
s.
A good trader is a realist who wants to grab a chunk from the body of a trend, leaving top- and bottom-fishing to people on an ego trip. (Dr. Alexander Elder)
USDx: In my indicator the USD is calculated the same as the other currencies, and it’s a bit different from the USDx.
Hello G.!
I checked it, my code looks correct to me.
This is the formula you provided:
`Usdx=50.14348112 * MathPow(iClose(“EURUSD”,0,i),-0.576) * MathPow(iClose(“USDJPY”,0,i),0.136)
* MathPow(iClose(“GBPUSD”,0,i),-0.119) * MathPow(iClose(“USDCAD”,0,i),0.091)
* MathPow(iClose(“USDSEK”,0,i),0.042) * MathPow(iClose(“USDCHF”,0,i),0.036);`When extracting EURx from this one, we simply multiply by
iClose(“EURUSD”,0,i), when extracting CADx we would divide byiClose(“USDCAD”,0,i). At least, this is the arithmetics I’m using at the moment, and it looks ok.For USDx, there’s nothing to do:
USDx = USDx, seefunction getCurrencyIndexValue(const int _tf, const int sh, const int currIndex)in my code.So looking at the pure math, for a USD centered index the pure index exactly represents the USD indicator value, and the same is true for the EUR centered index vs. EUR indicator value. Both have to be the same, otherwise there would be an issue to care for. Or maybe your indicator uses a different formula to extract YYYx value from USDx – I simply don’t know. Yet at the moment I can’t imagine why a different extraction of currency crosses should be used.
s.
A good trader is a realist who wants to grab a chunk from the body of a trend, leaving top- and bottom-fishing to people on an ego trip. (Dr. Alexander Elder)
Good summary!
- You must be logged in to reply to this topic.