Pair trading and multicurrency arbitrage. The showdown. - page 266
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Here from the phone such a formula and yours can be used? Yours is not even a logarithm which.
The formula looks like this:
Relative change (%) = (( Current price - Beginning price) / Beginning price) * 100%
Read how datasets normalise values before feeding them into a neural network. The principle is the same. Only here 100 units are equalised as 100% to make the analysis clearer.
Okay.
This is all well and good, but de facto it is practically counter-trend trading on a pullback. "found a big move, took a small pullback and quickly-quickly got out before it turned around:-)"
and you want to take the daily movement itself. It's several times bigger.
Who is strong in spirit and powerful neurons?
there is a data-set that you can play with and try to get something out of it, look for patterns or even try to predict.
Anyway, I've sketched out a script that collects statistics like this:
2021.07.13,19:15,2,[JPY,CAD,USD,CHF,USX,GEOM,EUR,GBP,NZD,AUD],0.01018922,[0.00370719,0.00192560,0.00000000,0.00000000,-0.00008696,-0.00159321,-0.00467351,-0.00502914,-0.00553771,-0.00648203],[6.78633498,0.32234353,0.00000000,-0.17753904,0.14599499,0.66345219,0.24016776,0.47000055,-0.52091434,-0.42027918]
2021.07.14,03:15,3,[NZD,EUR,AUD,GBP,GEOM,USX,USD,CHF,CAD,JPY],0.02559251,[0.01747391,0.00648013,0.00580617,0.00498748,0.00260831,0.00019551,0.00000000,0.00000000,-0.00028810,-0.00811860],[-0.52645204,0.23549425,-0.42676122,0.46497141,0.66185898,0.14590803,0.00000000,-0.17753904,0.32426913,6.79004217]
2021.07.14,19:15,3,[NZD,AUD,EUR,GBP,GEOM,CAD,USX,CHF,USD,JPY],0.00272100,[0.00157996,0.00144790,0.00090246,0.00048947,0.00034863,0.00024201,0.00002690,0.00000000,0.00000000,-0.00114105],[-0.50897814,-0.42095505,0.24197438,0.46995889,0.66446729,0.32398103,0.14610354,-0.17753904,0.00000000,6.78192357]
fields:
1 date
2 time
3 day of the week
4 currencies in order of size of movement, you can say the rating of currencies.
5 total swing (maximum movement minus minimum)
6 movements themselves in the same positional order as the rating
7 and prices for the specified time.
along with the usual currencies are given:
GEOM - average value. Since everything here is in logarithms, it is a geometric mean.
USX - indicator of "USD influence" usx=1.0/diag(eur,gbp,....) ; you can use it instead of usd. (the rate of dollar to dollar is 1, and its logarithm is 0, and it is not always convenient when the constant is 0).
Prices and movements are given in log2 scale, if you need to convert them to the usual form (though why?) price=MathPow(2.0, log2price); and movement in % percents = (MathPow(2.0,log2diver) - 1)*100.0.
Basically any pattern found is useful.
If it is possible to predict at least the sign of the GEOM,USX movement it is very cool.
If you manage to predict the order in the ranking of any pair, consider that you have won; You guessed for example GBP AUD, and indeed GBP in the ranking is to the left (higher) than AUD.
Well, if you determine exactly who will be on the edges of the ranking, you can consider yourself the owner of forex :-))
---
datasets in the trailer
script too.
USX - indicator of "USD influence" usx=1.0/diag(eur,gbp,....) ; you can use it instead of usd. (the rate of dollar to dollar is 1, and its logarithm is 0, and it is not always convenient when the constant is 0).
Prices and movements are given in log2 scale, if you need to convert them to the usual form (though why?) price=MathPow(2.0, log2price); and movement in % percents = (MathPow(2.0,log2diver) - 1)*100.0.
Basically any pattern found is useful.
If it is possible to predict at least the sign of the GEOM,USX movement it is very cool.
If you manage to predict the order in the ranking of any pair, consider that you have won; You guessed for example GBP AUD, and indeed GBP in the ranking is to the left (higher) than AUD.
Well, if you determine exactly who will be on the edges of the ranking, you can consider yourself the owner of forex :-))
---
datasets in the trailer
script too.
very interesting all this! and new interpretations! and conditions! it would be even more to understand it all at once.....
very all this is very interesting! and new interpretations! and conditions! it would be even more at once to understand it all.....
The interpretations are the same. In dataset exactly the same as it was on the screenshots: all currencies, rating and divergences for the period of time.
the interpretations are the same. In the dataset exactly the same as in the screenshots: all currencies, rating and discrepancies for the period of time.
thanks for the clarification - I am smoking the material - I am writing an indicator (edit) on relative movement - it is already written in the database at me.....
the interpretations are the same. In the dataset exactly the same as in the screenshots: all currencies, rating and discrepancies for the period of time.
By the way, the formula for the non-zero Renov triangle:
- for any point in time (even for now) take the cross aaabbb and aaausd bbbusd
- we will look for multipliers of lots k1, k2, k3 so that buy k1 aaabbb, sell k2 aaausd, buy k3 bbbusd more or less steadily rotate around 0, but were not 0 in the spot initially.
- from f-l price crosses : aaabbb = aaausd / bbbusd or in logarithms ln(aaabbb) = ln(aaausd) - ln(bbbusd).
- linearise ln() in the vicinity of the cross aaabbb.
- that is, we assume (or statistically obtain) that aaabbb in the near future can be from aaabbb_min to aaabbb_max
- we take the corresponding segment of the logarithmic function and conduct a linear regression. We get a certain f(x) = k1*x + b
- actually k1 will be the multiplier of lots at aaabbb.
- repeat for majors.
that's all. we got lots k1,k2,k3 which are slightly different from the empty lot.
PS/ it was written in my sleep, everything is accurate to the inverse function :-) i.e. linearise exp() , the same formula ... or take coefficients 1/k1 1/k2.
PPS/ you can do the same with any compositions, not necessarily a triangle, you can make a multi-facet.
by the way, the formula for a non-zero ren triangle:
- for any moment in time (even for now) take the cross aaabbb and aaausd bbbusd
- we will look for multipliers of lots k1, k2, k3 so that buy k1 aaabbb, sell k2 aaausd, buy k3 bbbusd more or less steadily rotate around 0, but were not 0 in the spot initially.
- from f-l cross prices: aaabbb = aaausd / bbbusd or in logarithms ln(aaabbb) = ln(aaausd) - ln(bbbusd).
- linearise ln() in the vicinity of the cross aaabbb.
- i.e. we assume (or statistically obtain) that aaabbb in the near future can be from aaabbb_min to aaabbb_max
- we take the corresponding segment of the logarithmic function and conduct linear regression. We get a certain f(x) = k1*x + b
- actually k1 will be the multiplier of lots at aaabbb.
- repeat for majors.
that's all. we got lots k1,k2,k3, which are slightly different from the empty lot.
PS/ it was written in my sleep, everything is accurate to the inverse function :-) i.e. linearise exp() , the same formula ... or take coefficients 1/k1 1/k2.
PPS/ you can do the same with any compositions, not necessarily a triangle, you can make a multi-facet.
Super! Wow - you have dug!!!! where!!!! I'll reread it tonight after work.... carefully... under the record....)
Also GRAAL formula!!!!! )
"- we take the corresponding segment of the logarithmic function and run a linear regression. We get a certain f(x) = k1*x + b" - that's yes! his equity there is constantly growing at an angle and has a straight line...
By the way, Friday's forecast OK...except that no positions were opened for the weekend.
on Friday the maximum run-up during the active market day, from 3 to 19, showed :
- up NZD (on the screenshot the white line at the top of the first bundle).
- down JPY (purple line at the bottom of the first bundle).
according to the principle "the maximally extended pair should at least pullback" the forecast was: NZD to sell, JPY to buy.
The second bundle is the result of the divergence from Friday 19:00. JPY whistled up, NZD down...if positions had been opened, it would have been possible to close with profit.
By the way, Friday's forecast is OK...except there were no positions open for the weekend.
on Friday the maximum spread for an active market day, from 3 to 19, showed :
- up NZD (on the screenshot the white line at the top of the first bundle).
- down JPY (purple line at the bottom of the first bundle)
according to the principle "the maximally extended pair should at least pullback" the forecast was: NZD to sell, JPY to buy.
The second bundle is the result of the divergence from Friday 19:00. JPY whistled up, NZD down...if positions had been opened, it would have been possible to close with profit.
Super! No kidding! For joining in and explaining!
like, subscribe, comment! )